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,
139 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
142 enum bpf_attach_type {
143 BPF_CGROUP_INET_INGRESS,
144 BPF_CGROUP_INET_EGRESS,
145 BPF_CGROUP_INET_SOCK_CREATE,
147 BPF_SK_SKB_STREAM_PARSER,
148 BPF_SK_SKB_STREAM_VERDICT,
151 BPF_CGROUP_INET4_BIND,
152 BPF_CGROUP_INET6_BIND,
153 BPF_CGROUP_INET4_CONNECT,
154 BPF_CGROUP_INET6_CONNECT,
155 __MAX_BPF_ATTACH_TYPE
158 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
160 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
162 * NONE(default): No further bpf programs allowed in the subtree.
164 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
165 * the program in this cgroup yields to sub-cgroup program.
167 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
168 * that cgroup program gets run in addition to the program in this cgroup.
170 * Only one program is allowed to be attached to a cgroup with
171 * NONE or BPF_F_ALLOW_OVERRIDE flag.
172 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
173 * release old program and attach the new one. Attach flags has to match.
175 * Multiple programs are allowed to be attached to a cgroup with
176 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
177 * (those that were attached first, run first)
178 * The programs of sub-cgroup are executed first, then programs of
179 * this cgroup and then programs of parent cgroup.
180 * When children program makes decision (like picking TCP CA or sock bind)
181 * parent program has a chance to override it.
183 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
184 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
186 * cgrp1 (MULTI progs A, B) ->
187 * cgrp2 (OVERRIDE prog C) ->
188 * cgrp3 (MULTI prog D) ->
189 * cgrp4 (OVERRIDE prog E) ->
190 * cgrp5 (NONE prog F)
191 * the event in cgrp5 triggers execution of F,D,A,B in that order.
192 * if prog F is detached, the execution is E,D,A,B
193 * if prog F and D are detached, the execution is E,A,B
194 * if prog F, E and D are detached, the execution is C,A,B
196 * All eligible programs are executed regardless of return code from
199 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
200 #define BPF_F_ALLOW_MULTI (1U << 1)
202 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
203 * verifier will perform strict alignment checking as if the kernel
204 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
205 * and NET_IP_ALIGN defined to 2.
207 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
209 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
210 #define BPF_PSEUDO_MAP_FD 1
212 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
213 * offset to another bpf function
215 #define BPF_PSEUDO_CALL 1
217 /* flags for BPF_MAP_UPDATE_ELEM command */
218 #define BPF_ANY 0 /* create new element or update existing */
219 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
220 #define BPF_EXIST 2 /* update existing element */
222 /* flags for BPF_MAP_CREATE command */
223 #define BPF_F_NO_PREALLOC (1U << 0)
224 /* Instead of having one common LRU list in the
225 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
226 * which can scale and perform better.
227 * Note, the LRU nodes (including free nodes) cannot be moved
228 * across different LRU lists.
230 #define BPF_F_NO_COMMON_LRU (1U << 1)
231 /* Specify numa node during map creation */
232 #define BPF_F_NUMA_NODE (1U << 2)
234 /* flags for BPF_PROG_QUERY */
235 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
237 #define BPF_OBJ_NAME_LEN 16U
239 /* Flags for accessing BPF object */
240 #define BPF_F_RDONLY (1U << 3)
241 #define BPF_F_WRONLY (1U << 4)
243 /* Flag for stack_map, store build_id+offset instead of pointer */
244 #define BPF_F_STACK_BUILD_ID (1U << 5)
246 enum bpf_stack_build_id_status {
247 /* user space need an empty entry to identify end of a trace */
248 BPF_STACK_BUILD_ID_EMPTY = 0,
249 /* with valid build_id and offset */
250 BPF_STACK_BUILD_ID_VALID = 1,
251 /* couldn't get build_id, fallback to ip */
252 BPF_STACK_BUILD_ID_IP = 2,
255 #define BPF_BUILD_ID_SIZE 20
256 struct bpf_stack_build_id {
258 unsigned char build_id[BPF_BUILD_ID_SIZE];
266 struct { /* anonymous struct used by BPF_MAP_CREATE command */
267 __u32 map_type; /* one of enum bpf_map_type */
268 __u32 key_size; /* size of key in bytes */
269 __u32 value_size; /* size of value in bytes */
270 __u32 max_entries; /* max number of entries in a map */
271 __u32 map_flags; /* BPF_MAP_CREATE related
272 * flags defined above.
274 __u32 inner_map_fd; /* fd pointing to the inner map */
275 __u32 numa_node; /* numa node (effective only if
276 * BPF_F_NUMA_NODE is set).
278 char map_name[BPF_OBJ_NAME_LEN];
279 __u32 map_ifindex; /* ifindex of netdev to create on */
282 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
287 __aligned_u64 next_key;
292 struct { /* anonymous struct used by BPF_PROG_LOAD command */
293 __u32 prog_type; /* one of enum bpf_prog_type */
296 __aligned_u64 license;
297 __u32 log_level; /* verbosity level of verifier */
298 __u32 log_size; /* size of user buffer */
299 __aligned_u64 log_buf; /* user supplied buffer */
300 __u32 kern_version; /* checked when prog_type=kprobe */
302 char prog_name[BPF_OBJ_NAME_LEN];
303 __u32 prog_ifindex; /* ifindex of netdev to prep for */
304 /* For some prog types expected attach type must be known at
305 * load time to verify attach type specific parts of prog
306 * (context accesses, allowed helpers, etc).
308 __u32 expected_attach_type;
311 struct { /* anonymous struct used by BPF_OBJ_* commands */
312 __aligned_u64 pathname;
317 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
318 __u32 target_fd; /* container object to attach to */
319 __u32 attach_bpf_fd; /* eBPF program to attach */
324 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
329 __aligned_u64 data_in;
330 __aligned_u64 data_out;
335 struct { /* anonymous struct used by BPF_*_GET_*_ID */
345 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
351 struct { /* anonymous struct used by BPF_PROG_QUERY command */
352 __u32 target_fd; /* container object to query */
356 __aligned_u64 prog_ids;
364 } __attribute__((aligned(8)));
366 /* BPF helper function descriptions:
368 * void *bpf_map_lookup_elem(&map, &key)
369 * Return: Map value or NULL
371 * int bpf_map_update_elem(&map, &key, &value, flags)
372 * Return: 0 on success or negative error
374 * int bpf_map_delete_elem(&map, &key)
375 * Return: 0 on success or negative error
377 * int bpf_probe_read(void *dst, int size, void *src)
378 * Return: 0 on success or negative error
380 * u64 bpf_ktime_get_ns(void)
381 * Return: current ktime
383 * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
384 * Return: length of buffer written or negative error
386 * u32 bpf_prandom_u32(void)
387 * Return: random value
389 * u32 bpf_raw_smp_processor_id(void)
390 * Return: SMP processor ID
392 * int bpf_skb_store_bytes(skb, offset, from, len, flags)
393 * store bytes into packet
394 * @skb: pointer to skb
395 * @offset: offset within packet from skb->mac_header
396 * @from: pointer where to copy bytes from
397 * @len: number of bytes to store into packet
398 * @flags: bit 0 - if true, recompute skb->csum
399 * other bits - reserved
400 * Return: 0 on success or negative error
402 * int bpf_l3_csum_replace(skb, offset, from, to, flags)
403 * recompute IP checksum
404 * @skb: pointer to skb
405 * @offset: offset within packet where IP checksum is located
406 * @from: old value of header field
407 * @to: new value of header field
408 * @flags: bits 0-3 - size of header field
409 * other bits - reserved
410 * Return: 0 on success or negative error
412 * int bpf_l4_csum_replace(skb, offset, from, to, flags)
413 * recompute TCP/UDP checksum
414 * @skb: pointer to skb
415 * @offset: offset within packet where TCP/UDP checksum is located
416 * @from: old value of header field
417 * @to: new value of header field
418 * @flags: bits 0-3 - size of header field
419 * bit 4 - is pseudo header
420 * other bits - reserved
421 * Return: 0 on success or negative error
423 * int bpf_tail_call(ctx, prog_array_map, index)
424 * jump into another BPF program
425 * @ctx: context pointer passed to next program
426 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
427 * @index: 32-bit index inside array that selects specific program to run
428 * Return: 0 on success or negative error
430 * int bpf_clone_redirect(skb, ifindex, flags)
431 * redirect to another netdev
432 * @skb: pointer to skb
433 * @ifindex: ifindex of the net device
434 * @flags: bit 0 - if set, redirect to ingress instead of egress
435 * other bits - reserved
436 * Return: 0 on success or negative error
438 * u64 bpf_get_current_pid_tgid(void)
439 * Return: current->tgid << 32 | current->pid
441 * u64 bpf_get_current_uid_gid(void)
442 * Return: current_gid << 32 | current_uid
444 * int bpf_get_current_comm(char *buf, int size_of_buf)
445 * stores current->comm into buf
446 * Return: 0 on success or negative error
448 * u32 bpf_get_cgroup_classid(skb)
449 * retrieve a proc's classid
450 * @skb: pointer to skb
451 * Return: classid if != 0
453 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
454 * Return: 0 on success or negative error
456 * int bpf_skb_vlan_pop(skb)
457 * Return: 0 on success or negative error
459 * int bpf_skb_get_tunnel_key(skb, key, size, flags)
460 * int bpf_skb_set_tunnel_key(skb, key, size, flags)
461 * retrieve or populate tunnel metadata
462 * @skb: pointer to skb
463 * @key: pointer to 'struct bpf_tunnel_key'
464 * @size: size of 'struct bpf_tunnel_key'
465 * @flags: room for future extensions
466 * Return: 0 on success or negative error
468 * u64 bpf_perf_event_read(map, flags)
469 * read perf event counter value
470 * @map: pointer to perf_event_array map
471 * @flags: index of event in the map or bitmask flags
472 * Return: value of perf event counter read or error code
474 * int bpf_redirect(ifindex, flags)
475 * redirect to another netdev
476 * @ifindex: ifindex of the net device
479 * bit 0 - if set, redirect to ingress instead of egress
480 * other bits - reserved
482 * all bits - reserved
483 * Return: cls_bpf: TC_ACT_REDIRECT on success or TC_ACT_SHOT on error
484 * xdp_bfp: XDP_REDIRECT on success or XDP_ABORT on error
485 * int bpf_redirect_map(map, key, flags)
486 * redirect to endpoint in map
487 * @map: pointer to dev map
488 * @key: index in map to lookup
490 * Return: XDP_REDIRECT on success or XDP_ABORT on error
492 * u32 bpf_get_route_realm(skb)
493 * retrieve a dst's tclassid
494 * @skb: pointer to skb
495 * Return: realm if != 0
497 * int bpf_perf_event_output(ctx, map, flags, data, size)
498 * output perf raw sample
499 * @ctx: struct pt_regs*
500 * @map: pointer to perf_event_array map
501 * @flags: index of event in the map or bitmask flags
502 * @data: data on stack to be output as raw data
503 * @size: size of data
504 * Return: 0 on success or negative error
506 * int bpf_get_stackid(ctx, map, flags)
507 * walk user or kernel stack and return id
508 * @ctx: struct pt_regs*
509 * @map: pointer to stack_trace map
510 * @flags: bits 0-7 - numer of stack frames to skip
511 * bit 8 - collect user stack instead of kernel
512 * bit 9 - compare stacks by hash only
513 * bit 10 - if two different stacks hash into the same stackid
515 * other bits - reserved
516 * Return: >= 0 stackid on success or negative error
518 * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
519 * calculate csum diff
520 * @from: raw from buffer
521 * @from_size: length of from buffer
523 * @to_size: length of to buffer
524 * @seed: optional seed
525 * Return: csum result or negative error code
527 * int bpf_skb_get_tunnel_opt(skb, opt, size)
528 * retrieve tunnel options metadata
529 * @skb: pointer to skb
530 * @opt: pointer to raw tunnel option data
531 * @size: size of @opt
532 * Return: option size
534 * int bpf_skb_set_tunnel_opt(skb, opt, size)
535 * populate tunnel options metadata
536 * @skb: pointer to skb
537 * @opt: pointer to raw tunnel option data
538 * @size: size of @opt
539 * Return: 0 on success or negative error
541 * int bpf_skb_change_proto(skb, proto, flags)
542 * Change protocol of the skb. Currently supported is v4 -> v6,
543 * v6 -> v4 transitions. The helper will also resize the skb. eBPF
544 * program is expected to fill the new headers via skb_store_bytes
545 * and lX_csum_replace.
546 * @skb: pointer to skb
547 * @proto: new skb->protocol type
549 * Return: 0 on success or negative error
551 * int bpf_skb_change_type(skb, type)
552 * Change packet type of skb.
553 * @skb: pointer to skb
554 * @type: new skb->pkt_type type
555 * Return: 0 on success or negative error
557 * int bpf_skb_under_cgroup(skb, map, index)
558 * Check cgroup2 membership of skb
559 * @skb: pointer to skb
560 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
561 * @index: index of the cgroup in the bpf_map
563 * == 0 skb failed the cgroup2 descendant test
564 * == 1 skb succeeded the cgroup2 descendant test
567 * u32 bpf_get_hash_recalc(skb)
568 * Retrieve and possibly recalculate skb->hash.
569 * @skb: pointer to skb
572 * u64 bpf_get_current_task(void)
573 * Returns current task_struct
576 * int bpf_probe_write_user(void *dst, void *src, int len)
577 * safely attempt to write to a location
578 * @dst: destination address in userspace
579 * @src: source address on stack
580 * @len: number of bytes to copy
581 * Return: 0 on success or negative error
583 * int bpf_current_task_under_cgroup(map, index)
584 * Check cgroup2 membership of current task
585 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
586 * @index: index of the cgroup in the bpf_map
588 * == 0 current failed the cgroup2 descendant test
589 * == 1 current succeeded the cgroup2 descendant test
592 * int bpf_skb_change_tail(skb, len, flags)
593 * The helper will resize the skb to the given new size, to be used f.e.
594 * with control messages.
595 * @skb: pointer to skb
596 * @len: new skb length
598 * Return: 0 on success or negative error
600 * int bpf_skb_pull_data(skb, len)
601 * The helper will pull in non-linear data in case the skb is non-linear
602 * and not all of len are part of the linear section. Only needed for
603 * read/write with direct packet access.
604 * @skb: pointer to skb
605 * @len: len to make read/writeable
606 * Return: 0 on success or negative error
608 * s64 bpf_csum_update(skb, csum)
609 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
610 * @skb: pointer to skb
612 * Return: csum on success or negative error
614 * void bpf_set_hash_invalid(skb)
615 * Invalidate current skb->hash.
616 * @skb: pointer to skb
618 * int bpf_get_numa_node_id()
619 * Return: Id of current NUMA node.
621 * int bpf_skb_change_head()
622 * Grows headroom of skb and adjusts MAC header offset accordingly.
623 * Will extends/reallocae as required automatically.
624 * May change skb data pointer and will thus invalidate any check
625 * performed for direct packet access.
626 * @skb: pointer to skb
627 * @len: length of header to be pushed in front
628 * @flags: Flags (unused for now)
629 * Return: 0 on success or negative error
631 * int bpf_xdp_adjust_head(xdp_md, delta)
632 * Adjust the xdp_md.data by delta
633 * @xdp_md: pointer to xdp_md
634 * @delta: An positive/negative integer to be added to xdp_md.data
635 * Return: 0 on success or negative on error
637 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
638 * Copy a NUL terminated string from unsafe address. In case the string
639 * length is smaller than size, the target is not padded with further NUL
640 * bytes. In case the string length is larger than size, just count-1
641 * bytes are copied and the last byte is set to NUL.
642 * @dst: destination address
643 * @size: maximum number of bytes to copy, including the trailing NUL
644 * @unsafe_ptr: unsafe address
646 * > 0 length of the string including the trailing NUL on success
649 * u64 bpf_get_socket_cookie(skb)
650 * Get the cookie for the socket stored inside sk_buff.
651 * @skb: pointer to skb
652 * Return: 8 Bytes non-decreasing number on success or 0 if the socket
653 * field is missing inside sk_buff
655 * u32 bpf_get_socket_uid(skb)
656 * Get the owner uid of the socket stored inside sk_buff.
657 * @skb: pointer to skb
658 * Return: uid of the socket owner on success or overflowuid if failed.
660 * u32 bpf_set_hash(skb, hash)
661 * Set full skb->hash.
662 * @skb: pointer to skb
665 * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
666 * Calls setsockopt. Not all opts are available, only those with
667 * integer optvals plus TCP_CONGESTION.
668 * Supported levels: SOL_SOCKET and IPPROTO_TCP
669 * @bpf_socket: pointer to bpf_socket
670 * @level: SOL_SOCKET or IPPROTO_TCP
671 * @optname: option name
672 * @optval: pointer to option value
673 * @optlen: length of optval in bytes
674 * Return: 0 or negative error
676 * int bpf_getsockopt(bpf_socket, level, optname, optval, optlen)
677 * Calls getsockopt. Not all opts are available.
678 * Supported levels: IPPROTO_TCP
679 * @bpf_socket: pointer to bpf_socket
680 * @level: IPPROTO_TCP
681 * @optname: option name
682 * @optval: pointer to option value
683 * @optlen: length of optval in bytes
684 * Return: 0 or negative error
686 * int bpf_sock_ops_cb_flags_set(bpf_sock_ops, flags)
687 * Set callback flags for sock_ops
688 * @bpf_sock_ops: pointer to bpf_sock_ops_kern struct
689 * @flags: flags value
690 * Return: 0 for no error
691 * -EINVAL if there is no full tcp socket
692 * bits in flags that are not supported by current kernel
694 * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
695 * Grow or shrink room in sk_buff.
696 * @skb: pointer to skb
697 * @len_diff: (signed) amount of room to grow/shrink
698 * @mode: operation mode (enum bpf_adj_room_mode)
699 * @flags: reserved for future use
700 * Return: 0 on success or negative error code
702 * int bpf_sk_redirect_map(map, key, flags)
703 * Redirect skb to a sock in map using key as a lookup key for the
705 * @map: pointer to sockmap
706 * @key: key to lookup sock in map
707 * @flags: reserved for future use
710 * int bpf_sock_map_update(skops, map, key, flags)
711 * @skops: pointer to bpf_sock_ops
712 * @map: pointer to sockmap to update
713 * @key: key to insert/update sock in map
714 * @flags: same flags as map update elem
716 * int bpf_xdp_adjust_meta(xdp_md, delta)
717 * Adjust the xdp_md.data_meta by delta
718 * @xdp_md: pointer to xdp_md
719 * @delta: An positive/negative integer to be added to xdp_md.data_meta
720 * Return: 0 on success or negative on error
722 * int bpf_perf_event_read_value(map, flags, buf, buf_size)
723 * read perf event counter value and perf event enabled/running time
724 * @map: pointer to perf_event_array map
725 * @flags: index of event in the map or bitmask flags
727 * @buf_size: size of the buf
728 * Return: 0 on success or negative error code
730 * int bpf_perf_prog_read_value(ctx, buf, buf_size)
731 * read perf prog attached perf event counter and enabled/running time
732 * @ctx: pointer to ctx
734 * @buf_size: size of the buf
735 * Return : 0 on success or negative error code
737 * int bpf_override_return(pt_regs, rc)
738 * @pt_regs: pointer to struct pt_regs
739 * @rc: the return value to set
741 * int bpf_msg_redirect_map(map, key, flags)
742 * Redirect msg to a sock in map using key as a lookup key for the
744 * @map: pointer to sockmap
745 * @key: key to lookup sock in map
746 * @flags: reserved for future use
749 * int bpf_bind(ctx, addr, addr_len)
750 * Bind socket to address. Only binding to IP is supported, no port can be
752 * @ctx: pointer to context of type bpf_sock_addr
753 * @addr: pointer to struct sockaddr to bind socket to
754 * @addr_len: length of sockaddr structure
755 * Return: 0 on success or negative error code
757 #define __BPF_FUNC_MAPPER(FN) \
759 FN(map_lookup_elem), \
760 FN(map_update_elem), \
761 FN(map_delete_elem), \
765 FN(get_prandom_u32), \
766 FN(get_smp_processor_id), \
767 FN(skb_store_bytes), \
768 FN(l3_csum_replace), \
769 FN(l4_csum_replace), \
771 FN(clone_redirect), \
772 FN(get_current_pid_tgid), \
773 FN(get_current_uid_gid), \
774 FN(get_current_comm), \
775 FN(get_cgroup_classid), \
778 FN(skb_get_tunnel_key), \
779 FN(skb_set_tunnel_key), \
780 FN(perf_event_read), \
782 FN(get_route_realm), \
783 FN(perf_event_output), \
784 FN(skb_load_bytes), \
787 FN(skb_get_tunnel_opt), \
788 FN(skb_set_tunnel_opt), \
789 FN(skb_change_proto), \
790 FN(skb_change_type), \
791 FN(skb_under_cgroup), \
792 FN(get_hash_recalc), \
793 FN(get_current_task), \
794 FN(probe_write_user), \
795 FN(current_task_under_cgroup), \
796 FN(skb_change_tail), \
799 FN(set_hash_invalid), \
800 FN(get_numa_node_id), \
801 FN(skb_change_head), \
802 FN(xdp_adjust_head), \
803 FN(probe_read_str), \
804 FN(get_socket_cookie), \
805 FN(get_socket_uid), \
808 FN(skb_adjust_room), \
810 FN(sk_redirect_map), \
811 FN(sock_map_update), \
812 FN(xdp_adjust_meta), \
813 FN(perf_event_read_value), \
814 FN(perf_prog_read_value), \
816 FN(override_return), \
817 FN(sock_ops_cb_flags_set), \
818 FN(msg_redirect_map), \
819 FN(msg_apply_bytes), \
820 FN(msg_cork_bytes), \
824 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
825 * function eBPF program intends to call
827 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
829 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
834 /* All flags used by eBPF helper functions, placed here. */
836 /* BPF_FUNC_skb_store_bytes flags. */
837 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
838 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
840 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
841 * First 4 bits are for passing the header field size.
843 #define BPF_F_HDR_FIELD_MASK 0xfULL
845 /* BPF_FUNC_l4_csum_replace flags. */
846 #define BPF_F_PSEUDO_HDR (1ULL << 4)
847 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
848 #define BPF_F_MARK_ENFORCE (1ULL << 6)
850 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
851 #define BPF_F_INGRESS (1ULL << 0)
853 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
854 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
856 /* BPF_FUNC_get_stackid flags. */
857 #define BPF_F_SKIP_FIELD_MASK 0xffULL
858 #define BPF_F_USER_STACK (1ULL << 8)
859 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
860 #define BPF_F_REUSE_STACKID (1ULL << 10)
862 /* BPF_FUNC_skb_set_tunnel_key flags. */
863 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
864 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
865 #define BPF_F_SEQ_NUMBER (1ULL << 3)
867 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
868 * BPF_FUNC_perf_event_read_value flags.
870 #define BPF_F_INDEX_MASK 0xffffffffULL
871 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
872 /* BPF_FUNC_perf_event_output for sk_buff input context. */
873 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
875 /* Mode for BPF_FUNC_skb_adjust_room helper. */
876 enum bpf_adj_room_mode {
880 /* user accessible mirror of in-kernel sk_buff.
881 * new fields can only be added to the end of this structure
893 __u32 ingress_ifindex;
903 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
905 __u32 remote_ip4; /* Stored in network byte order */
906 __u32 local_ip4; /* Stored in network byte order */
907 __u32 remote_ip6[4]; /* Stored in network byte order */
908 __u32 local_ip6[4]; /* Stored in network byte order */
909 __u32 remote_port; /* Stored in network byte order */
910 __u32 local_port; /* stored in host byte order */
916 struct bpf_tunnel_key {
920 __u32 remote_ipv6[4];
928 /* Generic BPF return codes which all BPF program types may support.
929 * The values are binary compatible with their TC_ACT_* counter-part to
930 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
933 * XDP is handled seprately, see XDP_*.
941 /* >127 are reserved for prog type specific return codes */
953 #define XDP_PACKET_HEADROOM 256
955 /* User return codes for XDP prog type.
956 * A valid XDP program must return one of these defined values. All other
957 * return codes are reserved for future use. Unknown return codes will
958 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
968 /* user accessible metadata for XDP packet hook
969 * new fields must be added to the end of this structure
975 /* Below access go through struct xdp_rxq_info */
976 __u32 ingress_ifindex; /* rxq->dev->ifindex */
977 __u32 rx_queue_index; /* rxq->queue_index */
985 /* user accessible metadata for SK_MSG packet hook, new fields must
986 * be added to the end of this structure
993 #define BPF_TAG_SIZE 8
995 struct bpf_prog_info {
998 __u8 tag[BPF_TAG_SIZE];
999 __u32 jited_prog_len;
1000 __u32 xlated_prog_len;
1001 __aligned_u64 jited_prog_insns;
1002 __aligned_u64 xlated_prog_insns;
1003 __u64 load_time; /* ns since boottime */
1004 __u32 created_by_uid;
1006 __aligned_u64 map_ids;
1007 char name[BPF_OBJ_NAME_LEN];
1011 } __attribute__((aligned(8)));
1013 struct bpf_map_info {
1020 char name[BPF_OBJ_NAME_LEN];
1024 } __attribute__((aligned(8)));
1026 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
1027 * by user and intended to be used by socket (e.g. to bind to, depends on
1028 * attach attach type).
1030 struct bpf_sock_addr {
1031 __u32 user_family; /* Allows 4-byte read, but no write. */
1032 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
1033 * Stored in network byte order.
1035 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
1036 * Stored in network byte order.
1038 __u32 user_port; /* Allows 4-byte read and write.
1039 * Stored in network byte order
1041 __u32 family; /* Allows 4-byte read, but no write */
1042 __u32 type; /* Allows 4-byte read, but no write */
1043 __u32 protocol; /* Allows 4-byte read, but no write */
1046 /* User bpf_sock_ops struct to access socket values and specify request ops
1047 * and their replies.
1048 * Some of this fields are in network (bigendian) byte order and may need
1049 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
1050 * New fields can only be added at the end of this structure
1052 struct bpf_sock_ops {
1055 __u32 args[4]; /* Optionally passed to bpf program */
1056 __u32 reply; /* Returned by bpf program */
1057 __u32 replylong[4]; /* Optionally returned by bpf prog */
1060 __u32 remote_ip4; /* Stored in network byte order */
1061 __u32 local_ip4; /* Stored in network byte order */
1062 __u32 remote_ip6[4]; /* Stored in network byte order */
1063 __u32 local_ip6[4]; /* Stored in network byte order */
1064 __u32 remote_port; /* Stored in network byte order */
1065 __u32 local_port; /* stored in host byte order */
1066 __u32 is_fullsock; /* Some TCP fields are only valid if
1067 * there is a full socket. If not, the
1068 * fields read as zero.
1071 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
1072 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
1081 __u32 rate_delivered;
1082 __u32 rate_interval_us;
1085 __u32 total_retrans;
1089 __u32 data_segs_out;
1093 __u64 bytes_received;
1097 /* Definitions for bpf_sock_ops_cb_flags */
1098 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
1099 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
1100 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
1101 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
1102 * supported cb flags
1105 /* List of known BPF sock_ops operators.
1106 * New entries can only be added at the end
1110 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
1111 * -1 if default value should be used
1113 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
1114 * window (in packets) or -1 if default
1115 * value should be used
1117 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
1118 * active connection is initialized
1120 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
1121 * active connection is
1124 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
1125 * passive connection is
1128 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
1131 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
1132 * based on the path and may be
1133 * dependent on the congestion control
1134 * algorithm. In general it indicates
1135 * a congestion threshold. RTTs above
1136 * this indicate congestion
1138 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
1139 * Arg1: value of icsk_retransmits
1140 * Arg2: value of icsk_rto
1141 * Arg3: whether RTO has expired
1143 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
1144 * Arg1: sequence number of 1st byte
1146 * Arg3: return value of
1147 * tcp_transmit_skb (0 => success)
1149 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
1155 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
1156 * changes between the TCP and BPF versions. Ideally this should never happen.
1157 * If it does, we need to add code to convert them before calling
1158 * the BPF sock_ops function.
1161 BPF_TCP_ESTABLISHED = 1,
1171 BPF_TCP_CLOSING, /* Now a valid state */
1172 BPF_TCP_NEW_SYN_RECV,
1174 BPF_TCP_MAX_STATES /* Leave at the end! */
1177 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
1178 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
1180 struct bpf_perf_event_value {
1186 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
1187 #define BPF_DEVCG_ACC_READ (1ULL << 1)
1188 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
1190 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
1191 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
1193 struct bpf_cgroup_dev_ctx {
1194 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
1200 struct bpf_raw_tracepoint_args {
1204 #endif /* _UAPI__LINUX_BPF_H__ */