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 BPF_CGROUP_INET4_POST_BIND,
156 BPF_CGROUP_INET6_POST_BIND,
157 __MAX_BPF_ATTACH_TYPE
160 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
162 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
164 * NONE(default): No further bpf programs allowed in the subtree.
166 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
167 * the program in this cgroup yields to sub-cgroup program.
169 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
170 * that cgroup program gets run in addition to the program in this cgroup.
172 * Only one program is allowed to be attached to a cgroup with
173 * NONE or BPF_F_ALLOW_OVERRIDE flag.
174 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
175 * release old program and attach the new one. Attach flags has to match.
177 * Multiple programs are allowed to be attached to a cgroup with
178 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
179 * (those that were attached first, run first)
180 * The programs of sub-cgroup are executed first, then programs of
181 * this cgroup and then programs of parent cgroup.
182 * When children program makes decision (like picking TCP CA or sock bind)
183 * parent program has a chance to override it.
185 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
186 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
188 * cgrp1 (MULTI progs A, B) ->
189 * cgrp2 (OVERRIDE prog C) ->
190 * cgrp3 (MULTI prog D) ->
191 * cgrp4 (OVERRIDE prog E) ->
192 * cgrp5 (NONE prog F)
193 * the event in cgrp5 triggers execution of F,D,A,B in that order.
194 * if prog F is detached, the execution is E,D,A,B
195 * if prog F and D are detached, the execution is E,A,B
196 * if prog F, E and D are detached, the execution is C,A,B
198 * All eligible programs are executed regardless of return code from
201 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
202 #define BPF_F_ALLOW_MULTI (1U << 1)
204 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
205 * verifier will perform strict alignment checking as if the kernel
206 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
207 * and NET_IP_ALIGN defined to 2.
209 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
211 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
212 #define BPF_PSEUDO_MAP_FD 1
214 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
215 * offset to another bpf function
217 #define BPF_PSEUDO_CALL 1
219 /* flags for BPF_MAP_UPDATE_ELEM command */
220 #define BPF_ANY 0 /* create new element or update existing */
221 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
222 #define BPF_EXIST 2 /* update existing element */
224 /* flags for BPF_MAP_CREATE command */
225 #define BPF_F_NO_PREALLOC (1U << 0)
226 /* Instead of having one common LRU list in the
227 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
228 * which can scale and perform better.
229 * Note, the LRU nodes (including free nodes) cannot be moved
230 * across different LRU lists.
232 #define BPF_F_NO_COMMON_LRU (1U << 1)
233 /* Specify numa node during map creation */
234 #define BPF_F_NUMA_NODE (1U << 2)
236 /* flags for BPF_PROG_QUERY */
237 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
239 #define BPF_OBJ_NAME_LEN 16U
241 /* Flags for accessing BPF object */
242 #define BPF_F_RDONLY (1U << 3)
243 #define BPF_F_WRONLY (1U << 4)
245 /* Flag for stack_map, store build_id+offset instead of pointer */
246 #define BPF_F_STACK_BUILD_ID (1U << 5)
248 enum bpf_stack_build_id_status {
249 /* user space need an empty entry to identify end of a trace */
250 BPF_STACK_BUILD_ID_EMPTY = 0,
251 /* with valid build_id and offset */
252 BPF_STACK_BUILD_ID_VALID = 1,
253 /* couldn't get build_id, fallback to ip */
254 BPF_STACK_BUILD_ID_IP = 2,
257 #define BPF_BUILD_ID_SIZE 20
258 struct bpf_stack_build_id {
260 unsigned char build_id[BPF_BUILD_ID_SIZE];
268 struct { /* anonymous struct used by BPF_MAP_CREATE command */
269 __u32 map_type; /* one of enum bpf_map_type */
270 __u32 key_size; /* size of key in bytes */
271 __u32 value_size; /* size of value in bytes */
272 __u32 max_entries; /* max number of entries in a map */
273 __u32 map_flags; /* BPF_MAP_CREATE related
274 * flags defined above.
276 __u32 inner_map_fd; /* fd pointing to the inner map */
277 __u32 numa_node; /* numa node (effective only if
278 * BPF_F_NUMA_NODE is set).
280 char map_name[BPF_OBJ_NAME_LEN];
281 __u32 map_ifindex; /* ifindex of netdev to create on */
284 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
289 __aligned_u64 next_key;
294 struct { /* anonymous struct used by BPF_PROG_LOAD command */
295 __u32 prog_type; /* one of enum bpf_prog_type */
298 __aligned_u64 license;
299 __u32 log_level; /* verbosity level of verifier */
300 __u32 log_size; /* size of user buffer */
301 __aligned_u64 log_buf; /* user supplied buffer */
302 __u32 kern_version; /* checked when prog_type=kprobe */
304 char prog_name[BPF_OBJ_NAME_LEN];
305 __u32 prog_ifindex; /* ifindex of netdev to prep for */
306 /* For some prog types expected attach type must be known at
307 * load time to verify attach type specific parts of prog
308 * (context accesses, allowed helpers, etc).
310 __u32 expected_attach_type;
313 struct { /* anonymous struct used by BPF_OBJ_* commands */
314 __aligned_u64 pathname;
319 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
320 __u32 target_fd; /* container object to attach to */
321 __u32 attach_bpf_fd; /* eBPF program to attach */
326 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
331 __aligned_u64 data_in;
332 __aligned_u64 data_out;
337 struct { /* anonymous struct used by BPF_*_GET_*_ID */
347 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
353 struct { /* anonymous struct used by BPF_PROG_QUERY command */
354 __u32 target_fd; /* container object to query */
358 __aligned_u64 prog_ids;
366 } __attribute__((aligned(8)));
368 /* BPF helper function descriptions:
370 * void *bpf_map_lookup_elem(&map, &key)
371 * Return: Map value or NULL
373 * int bpf_map_update_elem(&map, &key, &value, flags)
374 * Return: 0 on success or negative error
376 * int bpf_map_delete_elem(&map, &key)
377 * Return: 0 on success or negative error
379 * int bpf_probe_read(void *dst, int size, void *src)
380 * Return: 0 on success or negative error
382 * u64 bpf_ktime_get_ns(void)
383 * Return: current ktime
385 * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
386 * Return: length of buffer written or negative error
388 * u32 bpf_prandom_u32(void)
389 * Return: random value
391 * u32 bpf_raw_smp_processor_id(void)
392 * Return: SMP processor ID
394 * int bpf_skb_store_bytes(skb, offset, from, len, flags)
395 * store bytes into packet
396 * @skb: pointer to skb
397 * @offset: offset within packet from skb->mac_header
398 * @from: pointer where to copy bytes from
399 * @len: number of bytes to store into packet
400 * @flags: bit 0 - if true, recompute skb->csum
401 * other bits - reserved
402 * Return: 0 on success or negative error
404 * int bpf_l3_csum_replace(skb, offset, from, to, flags)
405 * recompute IP checksum
406 * @skb: pointer to skb
407 * @offset: offset within packet where IP checksum is located
408 * @from: old value of header field
409 * @to: new value of header field
410 * @flags: bits 0-3 - size of header field
411 * other bits - reserved
412 * Return: 0 on success or negative error
414 * int bpf_l4_csum_replace(skb, offset, from, to, flags)
415 * recompute TCP/UDP checksum
416 * @skb: pointer to skb
417 * @offset: offset within packet where TCP/UDP checksum is located
418 * @from: old value of header field
419 * @to: new value of header field
420 * @flags: bits 0-3 - size of header field
421 * bit 4 - is pseudo header
422 * other bits - reserved
423 * Return: 0 on success or negative error
425 * int bpf_tail_call(ctx, prog_array_map, index)
426 * jump into another BPF program
427 * @ctx: context pointer passed to next program
428 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
429 * @index: 32-bit index inside array that selects specific program to run
430 * Return: 0 on success or negative error
432 * int bpf_clone_redirect(skb, ifindex, flags)
433 * redirect to another netdev
434 * @skb: pointer to skb
435 * @ifindex: ifindex of the net device
436 * @flags: bit 0 - if set, redirect to ingress instead of egress
437 * other bits - reserved
438 * Return: 0 on success or negative error
440 * u64 bpf_get_current_pid_tgid(void)
441 * Return: current->tgid << 32 | current->pid
443 * u64 bpf_get_current_uid_gid(void)
444 * Return: current_gid << 32 | current_uid
446 * int bpf_get_current_comm(char *buf, int size_of_buf)
447 * stores current->comm into buf
448 * Return: 0 on success or negative error
450 * u32 bpf_get_cgroup_classid(skb)
451 * retrieve a proc's classid
452 * @skb: pointer to skb
453 * Return: classid if != 0
455 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
456 * Return: 0 on success or negative error
458 * int bpf_skb_vlan_pop(skb)
459 * Return: 0 on success or negative error
461 * int bpf_skb_get_tunnel_key(skb, key, size, flags)
462 * int bpf_skb_set_tunnel_key(skb, key, size, flags)
463 * retrieve or populate tunnel metadata
464 * @skb: pointer to skb
465 * @key: pointer to 'struct bpf_tunnel_key'
466 * @size: size of 'struct bpf_tunnel_key'
467 * @flags: room for future extensions
468 * Return: 0 on success or negative error
470 * u64 bpf_perf_event_read(map, flags)
471 * read perf event counter value
472 * @map: pointer to perf_event_array map
473 * @flags: index of event in the map or bitmask flags
474 * Return: value of perf event counter read or error code
476 * int bpf_redirect(ifindex, flags)
477 * redirect to another netdev
478 * @ifindex: ifindex of the net device
481 * bit 0 - if set, redirect to ingress instead of egress
482 * other bits - reserved
484 * all bits - reserved
485 * Return: cls_bpf: TC_ACT_REDIRECT on success or TC_ACT_SHOT on error
486 * xdp_bfp: XDP_REDIRECT on success or XDP_ABORT on error
487 * int bpf_redirect_map(map, key, flags)
488 * redirect to endpoint in map
489 * @map: pointer to dev map
490 * @key: index in map to lookup
492 * Return: XDP_REDIRECT on success or XDP_ABORT on error
494 * u32 bpf_get_route_realm(skb)
495 * retrieve a dst's tclassid
496 * @skb: pointer to skb
497 * Return: realm if != 0
499 * int bpf_perf_event_output(ctx, map, flags, data, size)
500 * output perf raw sample
501 * @ctx: struct pt_regs*
502 * @map: pointer to perf_event_array map
503 * @flags: index of event in the map or bitmask flags
504 * @data: data on stack to be output as raw data
505 * @size: size of data
506 * Return: 0 on success or negative error
508 * int bpf_get_stackid(ctx, map, flags)
509 * walk user or kernel stack and return id
510 * @ctx: struct pt_regs*
511 * @map: pointer to stack_trace map
512 * @flags: bits 0-7 - numer of stack frames to skip
513 * bit 8 - collect user stack instead of kernel
514 * bit 9 - compare stacks by hash only
515 * bit 10 - if two different stacks hash into the same stackid
517 * other bits - reserved
518 * Return: >= 0 stackid on success or negative error
520 * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
521 * calculate csum diff
522 * @from: raw from buffer
523 * @from_size: length of from buffer
525 * @to_size: length of to buffer
526 * @seed: optional seed
527 * Return: csum result or negative error code
529 * int bpf_skb_get_tunnel_opt(skb, opt, size)
530 * retrieve tunnel options metadata
531 * @skb: pointer to skb
532 * @opt: pointer to raw tunnel option data
533 * @size: size of @opt
534 * Return: option size
536 * int bpf_skb_set_tunnel_opt(skb, opt, size)
537 * populate tunnel options metadata
538 * @skb: pointer to skb
539 * @opt: pointer to raw tunnel option data
540 * @size: size of @opt
541 * Return: 0 on success or negative error
543 * int bpf_skb_change_proto(skb, proto, flags)
544 * Change protocol of the skb. Currently supported is v4 -> v6,
545 * v6 -> v4 transitions. The helper will also resize the skb. eBPF
546 * program is expected to fill the new headers via skb_store_bytes
547 * and lX_csum_replace.
548 * @skb: pointer to skb
549 * @proto: new skb->protocol type
551 * Return: 0 on success or negative error
553 * int bpf_skb_change_type(skb, type)
554 * Change packet type of skb.
555 * @skb: pointer to skb
556 * @type: new skb->pkt_type type
557 * Return: 0 on success or negative error
559 * int bpf_skb_under_cgroup(skb, map, index)
560 * Check cgroup2 membership of skb
561 * @skb: pointer to skb
562 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
563 * @index: index of the cgroup in the bpf_map
565 * == 0 skb failed the cgroup2 descendant test
566 * == 1 skb succeeded the cgroup2 descendant test
569 * u32 bpf_get_hash_recalc(skb)
570 * Retrieve and possibly recalculate skb->hash.
571 * @skb: pointer to skb
574 * u64 bpf_get_current_task(void)
575 * Returns current task_struct
578 * int bpf_probe_write_user(void *dst, void *src, int len)
579 * safely attempt to write to a location
580 * @dst: destination address in userspace
581 * @src: source address on stack
582 * @len: number of bytes to copy
583 * Return: 0 on success or negative error
585 * int bpf_current_task_under_cgroup(map, index)
586 * Check cgroup2 membership of current task
587 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
588 * @index: index of the cgroup in the bpf_map
590 * == 0 current failed the cgroup2 descendant test
591 * == 1 current succeeded the cgroup2 descendant test
594 * int bpf_skb_change_tail(skb, len, flags)
595 * The helper will resize the skb to the given new size, to be used f.e.
596 * with control messages.
597 * @skb: pointer to skb
598 * @len: new skb length
600 * Return: 0 on success or negative error
602 * int bpf_skb_pull_data(skb, len)
603 * The helper will pull in non-linear data in case the skb is non-linear
604 * and not all of len are part of the linear section. Only needed for
605 * read/write with direct packet access.
606 * @skb: pointer to skb
607 * @len: len to make read/writeable
608 * Return: 0 on success or negative error
610 * s64 bpf_csum_update(skb, csum)
611 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
612 * @skb: pointer to skb
614 * Return: csum on success or negative error
616 * void bpf_set_hash_invalid(skb)
617 * Invalidate current skb->hash.
618 * @skb: pointer to skb
620 * int bpf_get_numa_node_id()
621 * Return: Id of current NUMA node.
623 * int bpf_skb_change_head()
624 * Grows headroom of skb and adjusts MAC header offset accordingly.
625 * Will extends/reallocae as required automatically.
626 * May change skb data pointer and will thus invalidate any check
627 * performed for direct packet access.
628 * @skb: pointer to skb
629 * @len: length of header to be pushed in front
630 * @flags: Flags (unused for now)
631 * Return: 0 on success or negative error
633 * int bpf_xdp_adjust_head(xdp_md, delta)
634 * Adjust the xdp_md.data by delta
635 * @xdp_md: pointer to xdp_md
636 * @delta: An positive/negative integer to be added to xdp_md.data
637 * Return: 0 on success or negative on error
639 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
640 * Copy a NUL terminated string from unsafe address. In case the string
641 * length is smaller than size, the target is not padded with further NUL
642 * bytes. In case the string length is larger than size, just count-1
643 * bytes are copied and the last byte is set to NUL.
644 * @dst: destination address
645 * @size: maximum number of bytes to copy, including the trailing NUL
646 * @unsafe_ptr: unsafe address
648 * > 0 length of the string including the trailing NUL on success
651 * u64 bpf_get_socket_cookie(skb)
652 * Get the cookie for the socket stored inside sk_buff.
653 * @skb: pointer to skb
654 * Return: 8 Bytes non-decreasing number on success or 0 if the socket
655 * field is missing inside sk_buff
657 * u32 bpf_get_socket_uid(skb)
658 * Get the owner uid of the socket stored inside sk_buff.
659 * @skb: pointer to skb
660 * Return: uid of the socket owner on success or overflowuid if failed.
662 * u32 bpf_set_hash(skb, hash)
663 * Set full skb->hash.
664 * @skb: pointer to skb
667 * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
668 * Calls setsockopt. Not all opts are available, only those with
669 * integer optvals plus TCP_CONGESTION.
670 * Supported levels: SOL_SOCKET and IPPROTO_TCP
671 * @bpf_socket: pointer to bpf_socket
672 * @level: SOL_SOCKET or IPPROTO_TCP
673 * @optname: option name
674 * @optval: pointer to option value
675 * @optlen: length of optval in bytes
676 * Return: 0 or negative error
678 * int bpf_getsockopt(bpf_socket, level, optname, optval, optlen)
679 * Calls getsockopt. Not all opts are available.
680 * Supported levels: IPPROTO_TCP
681 * @bpf_socket: pointer to bpf_socket
682 * @level: IPPROTO_TCP
683 * @optname: option name
684 * @optval: pointer to option value
685 * @optlen: length of optval in bytes
686 * Return: 0 or negative error
688 * int bpf_sock_ops_cb_flags_set(bpf_sock_ops, flags)
689 * Set callback flags for sock_ops
690 * @bpf_sock_ops: pointer to bpf_sock_ops_kern struct
691 * @flags: flags value
692 * Return: 0 for no error
693 * -EINVAL if there is no full tcp socket
694 * bits in flags that are not supported by current kernel
696 * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
697 * Grow or shrink room in sk_buff.
698 * @skb: pointer to skb
699 * @len_diff: (signed) amount of room to grow/shrink
700 * @mode: operation mode (enum bpf_adj_room_mode)
701 * @flags: reserved for future use
702 * Return: 0 on success or negative error code
704 * int bpf_sk_redirect_map(map, key, flags)
705 * Redirect skb to a sock in map using key as a lookup key for the
707 * @map: pointer to sockmap
708 * @key: key to lookup sock in map
709 * @flags: reserved for future use
712 * int bpf_sock_map_update(skops, map, key, flags)
713 * @skops: pointer to bpf_sock_ops
714 * @map: pointer to sockmap to update
715 * @key: key to insert/update sock in map
716 * @flags: same flags as map update elem
718 * int bpf_xdp_adjust_meta(xdp_md, delta)
719 * Adjust the xdp_md.data_meta by delta
720 * @xdp_md: pointer to xdp_md
721 * @delta: An positive/negative integer to be added to xdp_md.data_meta
722 * Return: 0 on success or negative on error
724 * int bpf_perf_event_read_value(map, flags, buf, buf_size)
725 * read perf event counter value and perf event enabled/running time
726 * @map: pointer to perf_event_array map
727 * @flags: index of event in the map or bitmask flags
729 * @buf_size: size of the buf
730 * Return: 0 on success or negative error code
732 * int bpf_perf_prog_read_value(ctx, buf, buf_size)
733 * read perf prog attached perf event counter and enabled/running time
734 * @ctx: pointer to ctx
736 * @buf_size: size of the buf
737 * Return : 0 on success or negative error code
739 * int bpf_override_return(pt_regs, rc)
740 * @pt_regs: pointer to struct pt_regs
741 * @rc: the return value to set
743 * int bpf_msg_redirect_map(map, key, flags)
744 * Redirect msg to a sock in map using key as a lookup key for the
746 * @map: pointer to sockmap
747 * @key: key to lookup sock in map
748 * @flags: reserved for future use
751 * int bpf_bind(ctx, addr, addr_len)
752 * Bind socket to address. Only binding to IP is supported, no port can be
754 * @ctx: pointer to context of type bpf_sock_addr
755 * @addr: pointer to struct sockaddr to bind socket to
756 * @addr_len: length of sockaddr structure
757 * Return: 0 on success or negative error code
759 #define __BPF_FUNC_MAPPER(FN) \
761 FN(map_lookup_elem), \
762 FN(map_update_elem), \
763 FN(map_delete_elem), \
767 FN(get_prandom_u32), \
768 FN(get_smp_processor_id), \
769 FN(skb_store_bytes), \
770 FN(l3_csum_replace), \
771 FN(l4_csum_replace), \
773 FN(clone_redirect), \
774 FN(get_current_pid_tgid), \
775 FN(get_current_uid_gid), \
776 FN(get_current_comm), \
777 FN(get_cgroup_classid), \
780 FN(skb_get_tunnel_key), \
781 FN(skb_set_tunnel_key), \
782 FN(perf_event_read), \
784 FN(get_route_realm), \
785 FN(perf_event_output), \
786 FN(skb_load_bytes), \
789 FN(skb_get_tunnel_opt), \
790 FN(skb_set_tunnel_opt), \
791 FN(skb_change_proto), \
792 FN(skb_change_type), \
793 FN(skb_under_cgroup), \
794 FN(get_hash_recalc), \
795 FN(get_current_task), \
796 FN(probe_write_user), \
797 FN(current_task_under_cgroup), \
798 FN(skb_change_tail), \
801 FN(set_hash_invalid), \
802 FN(get_numa_node_id), \
803 FN(skb_change_head), \
804 FN(xdp_adjust_head), \
805 FN(probe_read_str), \
806 FN(get_socket_cookie), \
807 FN(get_socket_uid), \
810 FN(skb_adjust_room), \
812 FN(sk_redirect_map), \
813 FN(sock_map_update), \
814 FN(xdp_adjust_meta), \
815 FN(perf_event_read_value), \
816 FN(perf_prog_read_value), \
818 FN(override_return), \
819 FN(sock_ops_cb_flags_set), \
820 FN(msg_redirect_map), \
821 FN(msg_apply_bytes), \
822 FN(msg_cork_bytes), \
826 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
827 * function eBPF program intends to call
829 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
831 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
836 /* All flags used by eBPF helper functions, placed here. */
838 /* BPF_FUNC_skb_store_bytes flags. */
839 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
840 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
842 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
843 * First 4 bits are for passing the header field size.
845 #define BPF_F_HDR_FIELD_MASK 0xfULL
847 /* BPF_FUNC_l4_csum_replace flags. */
848 #define BPF_F_PSEUDO_HDR (1ULL << 4)
849 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
850 #define BPF_F_MARK_ENFORCE (1ULL << 6)
852 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
853 #define BPF_F_INGRESS (1ULL << 0)
855 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
856 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
858 /* BPF_FUNC_get_stackid flags. */
859 #define BPF_F_SKIP_FIELD_MASK 0xffULL
860 #define BPF_F_USER_STACK (1ULL << 8)
861 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
862 #define BPF_F_REUSE_STACKID (1ULL << 10)
864 /* BPF_FUNC_skb_set_tunnel_key flags. */
865 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
866 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
868 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
869 * BPF_FUNC_perf_event_read_value flags.
871 #define BPF_F_INDEX_MASK 0xffffffffULL
872 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
873 /* BPF_FUNC_perf_event_output for sk_buff input context. */
874 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
876 /* Mode for BPF_FUNC_skb_adjust_room helper. */
877 enum bpf_adj_room_mode {
881 /* user accessible mirror of in-kernel sk_buff.
882 * new fields can only be added to the end of this structure
894 __u32 ingress_ifindex;
904 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
906 __u32 remote_ip4; /* Stored in network byte order */
907 __u32 local_ip4; /* Stored in network byte order */
908 __u32 remote_ip6[4]; /* Stored in network byte order */
909 __u32 local_ip6[4]; /* Stored in network byte order */
910 __u32 remote_port; /* Stored in network byte order */
911 __u32 local_port; /* stored in host byte order */
917 struct bpf_tunnel_key {
921 __u32 remote_ipv6[4];
929 /* Generic BPF return codes which all BPF program types may support.
930 * The values are binary compatible with their TC_ACT_* counter-part to
931 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
934 * XDP is handled seprately, see XDP_*.
942 /* >127 are reserved for prog type specific return codes */
952 __u32 src_ip4; /* Allows 1,2,4-byte read.
953 * Stored in network byte order.
955 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
956 * Stored in network byte order.
958 __u32 src_port; /* Allows 4-byte read.
959 * Stored in host byte order
963 #define XDP_PACKET_HEADROOM 256
965 /* User return codes for XDP prog type.
966 * A valid XDP program must return one of these defined values. All other
967 * return codes are reserved for future use. Unknown return codes will
968 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
978 /* user accessible metadata for XDP packet hook
979 * new fields must be added to the end of this structure
985 /* Below access go through struct xdp_rxq_info */
986 __u32 ingress_ifindex; /* rxq->dev->ifindex */
987 __u32 rx_queue_index; /* rxq->queue_index */
995 /* user accessible metadata for SK_MSG packet hook, new fields must
996 * be added to the end of this structure
1003 #define BPF_TAG_SIZE 8
1005 struct bpf_prog_info {
1008 __u8 tag[BPF_TAG_SIZE];
1009 __u32 jited_prog_len;
1010 __u32 xlated_prog_len;
1011 __aligned_u64 jited_prog_insns;
1012 __aligned_u64 xlated_prog_insns;
1013 __u64 load_time; /* ns since boottime */
1014 __u32 created_by_uid;
1016 __aligned_u64 map_ids;
1017 char name[BPF_OBJ_NAME_LEN];
1021 } __attribute__((aligned(8)));
1023 struct bpf_map_info {
1030 char name[BPF_OBJ_NAME_LEN];
1034 } __attribute__((aligned(8)));
1036 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
1037 * by user and intended to be used by socket (e.g. to bind to, depends on
1038 * attach attach type).
1040 struct bpf_sock_addr {
1041 __u32 user_family; /* Allows 4-byte read, but no write. */
1042 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
1043 * Stored in network byte order.
1045 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
1046 * Stored in network byte order.
1048 __u32 user_port; /* Allows 4-byte read and write.
1049 * Stored in network byte order
1051 __u32 family; /* Allows 4-byte read, but no write */
1052 __u32 type; /* Allows 4-byte read, but no write */
1053 __u32 protocol; /* Allows 4-byte read, but no write */
1056 /* User bpf_sock_ops struct to access socket values and specify request ops
1057 * and their replies.
1058 * Some of this fields are in network (bigendian) byte order and may need
1059 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
1060 * New fields can only be added at the end of this structure
1062 struct bpf_sock_ops {
1065 __u32 args[4]; /* Optionally passed to bpf program */
1066 __u32 reply; /* Returned by bpf program */
1067 __u32 replylong[4]; /* Optionally returned by bpf prog */
1070 __u32 remote_ip4; /* Stored in network byte order */
1071 __u32 local_ip4; /* Stored in network byte order */
1072 __u32 remote_ip6[4]; /* Stored in network byte order */
1073 __u32 local_ip6[4]; /* Stored in network byte order */
1074 __u32 remote_port; /* Stored in network byte order */
1075 __u32 local_port; /* stored in host byte order */
1076 __u32 is_fullsock; /* Some TCP fields are only valid if
1077 * there is a full socket. If not, the
1078 * fields read as zero.
1081 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
1082 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
1091 __u32 rate_delivered;
1092 __u32 rate_interval_us;
1095 __u32 total_retrans;
1099 __u32 data_segs_out;
1103 __u64 bytes_received;
1107 /* Definitions for bpf_sock_ops_cb_flags */
1108 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
1109 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
1110 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
1111 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
1112 * supported cb flags
1115 /* List of known BPF sock_ops operators.
1116 * New entries can only be added at the end
1120 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
1121 * -1 if default value should be used
1123 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
1124 * window (in packets) or -1 if default
1125 * value should be used
1127 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
1128 * active connection is initialized
1130 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
1131 * active connection is
1134 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
1135 * passive connection is
1138 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
1141 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
1142 * based on the path and may be
1143 * dependent on the congestion control
1144 * algorithm. In general it indicates
1145 * a congestion threshold. RTTs above
1146 * this indicate congestion
1148 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
1149 * Arg1: value of icsk_retransmits
1150 * Arg2: value of icsk_rto
1151 * Arg3: whether RTO has expired
1153 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
1154 * Arg1: sequence number of 1st byte
1156 * Arg3: return value of
1157 * tcp_transmit_skb (0 => success)
1159 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
1165 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
1166 * changes between the TCP and BPF versions. Ideally this should never happen.
1167 * If it does, we need to add code to convert them before calling
1168 * the BPF sock_ops function.
1171 BPF_TCP_ESTABLISHED = 1,
1181 BPF_TCP_CLOSING, /* Now a valid state */
1182 BPF_TCP_NEW_SYN_RECV,
1184 BPF_TCP_MAX_STATES /* Leave at the end! */
1187 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
1188 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
1190 struct bpf_perf_event_value {
1196 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
1197 #define BPF_DEVCG_ACC_READ (1ULL << 1)
1198 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
1200 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
1201 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
1203 struct bpf_cgroup_dev_ctx {
1204 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
1210 struct bpf_raw_tracepoint_args {
1214 #endif /* _UAPI__LINUX_BPF_H__ */