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,
105 BPF_MAP_TYPE_PROG_ARRAY,
106 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
107 BPF_MAP_TYPE_PERCPU_HASH,
108 BPF_MAP_TYPE_PERCPU_ARRAY,
109 BPF_MAP_TYPE_STACK_TRACE,
110 BPF_MAP_TYPE_CGROUP_ARRAY,
111 BPF_MAP_TYPE_LRU_HASH,
112 BPF_MAP_TYPE_LRU_PERCPU_HASH,
113 BPF_MAP_TYPE_LPM_TRIE,
114 BPF_MAP_TYPE_ARRAY_OF_MAPS,
115 BPF_MAP_TYPE_HASH_OF_MAPS,
117 BPF_MAP_TYPE_SOCKMAP,
122 BPF_PROG_TYPE_UNSPEC,
123 BPF_PROG_TYPE_SOCKET_FILTER,
124 BPF_PROG_TYPE_KPROBE,
125 BPF_PROG_TYPE_SCHED_CLS,
126 BPF_PROG_TYPE_SCHED_ACT,
127 BPF_PROG_TYPE_TRACEPOINT,
129 BPF_PROG_TYPE_PERF_EVENT,
130 BPF_PROG_TYPE_CGROUP_SKB,
131 BPF_PROG_TYPE_CGROUP_SOCK,
132 BPF_PROG_TYPE_LWT_IN,
133 BPF_PROG_TYPE_LWT_OUT,
134 BPF_PROG_TYPE_LWT_XMIT,
135 BPF_PROG_TYPE_SOCK_OPS,
136 BPF_PROG_TYPE_SK_SKB,
137 BPF_PROG_TYPE_CGROUP_DEVICE,
138 BPF_PROG_TYPE_SK_MSG,
139 BPF_PROG_TYPE_RAW_TRACEPOINT,
140 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
143 enum bpf_attach_type {
144 BPF_CGROUP_INET_INGRESS,
145 BPF_CGROUP_INET_EGRESS,
146 BPF_CGROUP_INET_SOCK_CREATE,
148 BPF_SK_SKB_STREAM_PARSER,
149 BPF_SK_SKB_STREAM_VERDICT,
152 BPF_CGROUP_INET4_BIND,
153 BPF_CGROUP_INET6_BIND,
154 BPF_CGROUP_INET4_CONNECT,
155 BPF_CGROUP_INET6_CONNECT,
156 BPF_CGROUP_INET4_POST_BIND,
157 BPF_CGROUP_INET6_POST_BIND,
158 __MAX_BPF_ATTACH_TYPE
161 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
163 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
165 * NONE(default): No further bpf programs allowed in the subtree.
167 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
168 * the program in this cgroup yields to sub-cgroup program.
170 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
171 * that cgroup program gets run in addition to the program in this cgroup.
173 * Only one program is allowed to be attached to a cgroup with
174 * NONE or BPF_F_ALLOW_OVERRIDE flag.
175 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
176 * release old program and attach the new one. Attach flags has to match.
178 * Multiple programs are allowed to be attached to a cgroup with
179 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
180 * (those that were attached first, run first)
181 * The programs of sub-cgroup are executed first, then programs of
182 * this cgroup and then programs of parent cgroup.
183 * When children program makes decision (like picking TCP CA or sock bind)
184 * parent program has a chance to override it.
186 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
187 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
189 * cgrp1 (MULTI progs A, B) ->
190 * cgrp2 (OVERRIDE prog C) ->
191 * cgrp3 (MULTI prog D) ->
192 * cgrp4 (OVERRIDE prog E) ->
193 * cgrp5 (NONE prog F)
194 * the event in cgrp5 triggers execution of F,D,A,B in that order.
195 * if prog F is detached, the execution is E,D,A,B
196 * if prog F and D are detached, the execution is E,A,B
197 * if prog F, E and D are detached, the execution is C,A,B
199 * All eligible programs are executed regardless of return code from
202 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
203 #define BPF_F_ALLOW_MULTI (1U << 1)
205 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
206 * verifier will perform strict alignment checking as if the kernel
207 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
208 * and NET_IP_ALIGN defined to 2.
210 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
212 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
213 #define BPF_PSEUDO_MAP_FD 1
215 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
216 * offset to another bpf function
218 #define BPF_PSEUDO_CALL 1
220 /* flags for BPF_MAP_UPDATE_ELEM command */
221 #define BPF_ANY 0 /* create new element or update existing */
222 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
223 #define BPF_EXIST 2 /* update existing element */
225 /* flags for BPF_MAP_CREATE command */
226 #define BPF_F_NO_PREALLOC (1U << 0)
227 /* Instead of having one common LRU list in the
228 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
229 * which can scale and perform better.
230 * Note, the LRU nodes (including free nodes) cannot be moved
231 * across different LRU lists.
233 #define BPF_F_NO_COMMON_LRU (1U << 1)
234 /* Specify numa node during map creation */
235 #define BPF_F_NUMA_NODE (1U << 2)
237 /* flags for BPF_PROG_QUERY */
238 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
240 #define BPF_OBJ_NAME_LEN 16U
242 /* Flags for accessing BPF object */
243 #define BPF_F_RDONLY (1U << 3)
244 #define BPF_F_WRONLY (1U << 4)
246 /* Flag for stack_map, store build_id+offset instead of pointer */
247 #define BPF_F_STACK_BUILD_ID (1U << 5)
249 enum bpf_stack_build_id_status {
250 /* user space need an empty entry to identify end of a trace */
251 BPF_STACK_BUILD_ID_EMPTY = 0,
252 /* with valid build_id and offset */
253 BPF_STACK_BUILD_ID_VALID = 1,
254 /* couldn't get build_id, fallback to ip */
255 BPF_STACK_BUILD_ID_IP = 2,
258 #define BPF_BUILD_ID_SIZE 20
259 struct bpf_stack_build_id {
261 unsigned char build_id[BPF_BUILD_ID_SIZE];
269 struct { /* anonymous struct used by BPF_MAP_CREATE command */
270 __u32 map_type; /* one of enum bpf_map_type */
271 __u32 key_size; /* size of key in bytes */
272 __u32 value_size; /* size of value in bytes */
273 __u32 max_entries; /* max number of entries in a map */
274 __u32 map_flags; /* BPF_MAP_CREATE related
275 * flags defined above.
277 __u32 inner_map_fd; /* fd pointing to the inner map */
278 __u32 numa_node; /* numa node (effective only if
279 * BPF_F_NUMA_NODE is set).
281 char map_name[BPF_OBJ_NAME_LEN];
282 __u32 map_ifindex; /* ifindex of netdev to create on */
283 __u32 btf_fd; /* fd pointing to a BTF type data */
284 __u32 btf_key_id; /* BTF type_id of the key */
285 __u32 btf_value_id; /* BTF type_id of the value */
288 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
293 __aligned_u64 next_key;
298 struct { /* anonymous struct used by BPF_PROG_LOAD command */
299 __u32 prog_type; /* one of enum bpf_prog_type */
302 __aligned_u64 license;
303 __u32 log_level; /* verbosity level of verifier */
304 __u32 log_size; /* size of user buffer */
305 __aligned_u64 log_buf; /* user supplied buffer */
306 __u32 kern_version; /* checked when prog_type=kprobe */
308 char prog_name[BPF_OBJ_NAME_LEN];
309 __u32 prog_ifindex; /* ifindex of netdev to prep for */
310 /* For some prog types expected attach type must be known at
311 * load time to verify attach type specific parts of prog
312 * (context accesses, allowed helpers, etc).
314 __u32 expected_attach_type;
317 struct { /* anonymous struct used by BPF_OBJ_* commands */
318 __aligned_u64 pathname;
323 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
324 __u32 target_fd; /* container object to attach to */
325 __u32 attach_bpf_fd; /* eBPF program to attach */
330 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
335 __aligned_u64 data_in;
336 __aligned_u64 data_out;
341 struct { /* anonymous struct used by BPF_*_GET_*_ID */
351 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
357 struct { /* anonymous struct used by BPF_PROG_QUERY command */
358 __u32 target_fd; /* container object to query */
362 __aligned_u64 prog_ids;
371 struct { /* anonymous struct for BPF_BTF_LOAD */
373 __aligned_u64 btf_log_buf;
378 } __attribute__((aligned(8)));
380 /* BPF helper function descriptions:
382 * void *bpf_map_lookup_elem(&map, &key)
383 * Return: Map value or NULL
385 * int bpf_map_update_elem(&map, &key, &value, flags)
386 * Return: 0 on success or negative error
388 * int bpf_map_delete_elem(&map, &key)
389 * Return: 0 on success or negative error
391 * int bpf_probe_read(void *dst, int size, void *src)
392 * Return: 0 on success or negative error
394 * u64 bpf_ktime_get_ns(void)
395 * Return: current ktime
397 * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
398 * Return: length of buffer written or negative error
400 * u32 bpf_prandom_u32(void)
401 * Return: random value
403 * u32 bpf_raw_smp_processor_id(void)
404 * Return: SMP processor ID
406 * int bpf_skb_store_bytes(skb, offset, from, len, flags)
407 * store bytes into packet
408 * @skb: pointer to skb
409 * @offset: offset within packet from skb->mac_header
410 * @from: pointer where to copy bytes from
411 * @len: number of bytes to store into packet
412 * @flags: bit 0 - if true, recompute skb->csum
413 * other bits - reserved
414 * Return: 0 on success or negative error
416 * int bpf_l3_csum_replace(skb, offset, from, to, flags)
417 * recompute IP checksum
418 * @skb: pointer to skb
419 * @offset: offset within packet where IP checksum is located
420 * @from: old value of header field
421 * @to: new value of header field
422 * @flags: bits 0-3 - size of header field
423 * other bits - reserved
424 * Return: 0 on success or negative error
426 * int bpf_l4_csum_replace(skb, offset, from, to, flags)
427 * recompute TCP/UDP checksum
428 * @skb: pointer to skb
429 * @offset: offset within packet where TCP/UDP checksum is located
430 * @from: old value of header field
431 * @to: new value of header field
432 * @flags: bits 0-3 - size of header field
433 * bit 4 - is pseudo header
434 * other bits - reserved
435 * Return: 0 on success or negative error
437 * int bpf_tail_call(ctx, prog_array_map, index)
438 * jump into another BPF program
439 * @ctx: context pointer passed to next program
440 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
441 * @index: 32-bit index inside array that selects specific program to run
442 * Return: 0 on success or negative error
444 * int bpf_clone_redirect(skb, ifindex, flags)
445 * redirect to another netdev
446 * @skb: pointer to skb
447 * @ifindex: ifindex of the net device
448 * @flags: bit 0 - if set, redirect to ingress instead of egress
449 * other bits - reserved
450 * Return: 0 on success or negative error
452 * u64 bpf_get_current_pid_tgid(void)
453 * Return: current->tgid << 32 | current->pid
455 * u64 bpf_get_current_uid_gid(void)
456 * Return: current_gid << 32 | current_uid
458 * int bpf_get_current_comm(char *buf, int size_of_buf)
459 * stores current->comm into buf
460 * Return: 0 on success or negative error
462 * u32 bpf_get_cgroup_classid(skb)
463 * retrieve a proc's classid
464 * @skb: pointer to skb
465 * Return: classid if != 0
467 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
468 * Return: 0 on success or negative error
470 * int bpf_skb_vlan_pop(skb)
471 * Return: 0 on success or negative error
473 * int bpf_skb_get_tunnel_key(skb, key, size, flags)
474 * int bpf_skb_set_tunnel_key(skb, key, size, flags)
475 * retrieve or populate tunnel metadata
476 * @skb: pointer to skb
477 * @key: pointer to 'struct bpf_tunnel_key'
478 * @size: size of 'struct bpf_tunnel_key'
479 * @flags: room for future extensions
480 * Return: 0 on success or negative error
482 * u64 bpf_perf_event_read(map, flags)
483 * read perf event counter value
484 * @map: pointer to perf_event_array map
485 * @flags: index of event in the map or bitmask flags
486 * Return: value of perf event counter read or error code
488 * int bpf_redirect(ifindex, flags)
489 * redirect to another netdev
490 * @ifindex: ifindex of the net device
493 * bit 0 - if set, redirect to ingress instead of egress
494 * other bits - reserved
496 * all bits - reserved
497 * Return: cls_bpf: TC_ACT_REDIRECT on success or TC_ACT_SHOT on error
498 * xdp_bfp: XDP_REDIRECT on success or XDP_ABORT on error
499 * int bpf_redirect_map(map, key, flags)
500 * redirect to endpoint in map
501 * @map: pointer to dev map
502 * @key: index in map to lookup
504 * Return: XDP_REDIRECT on success or XDP_ABORT on error
506 * u32 bpf_get_route_realm(skb)
507 * retrieve a dst's tclassid
508 * @skb: pointer to skb
509 * Return: realm if != 0
511 * int bpf_perf_event_output(ctx, map, flags, data, size)
512 * output perf raw sample
513 * @ctx: struct pt_regs*
514 * @map: pointer to perf_event_array map
515 * @flags: index of event in the map or bitmask flags
516 * @data: data on stack to be output as raw data
517 * @size: size of data
518 * Return: 0 on success or negative error
520 * int bpf_get_stackid(ctx, map, flags)
521 * walk user or kernel stack and return id
522 * @ctx: struct pt_regs*
523 * @map: pointer to stack_trace map
524 * @flags: bits 0-7 - numer of stack frames to skip
525 * bit 8 - collect user stack instead of kernel
526 * bit 9 - compare stacks by hash only
527 * bit 10 - if two different stacks hash into the same stackid
529 * other bits - reserved
530 * Return: >= 0 stackid on success or negative error
532 * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
533 * calculate csum diff
534 * @from: raw from buffer
535 * @from_size: length of from buffer
537 * @to_size: length of to buffer
538 * @seed: optional seed
539 * Return: csum result or negative error code
541 * int bpf_skb_get_tunnel_opt(skb, opt, size)
542 * retrieve tunnel options metadata
543 * @skb: pointer to skb
544 * @opt: pointer to raw tunnel option data
545 * @size: size of @opt
546 * Return: option size
548 * int bpf_skb_set_tunnel_opt(skb, opt, size)
549 * populate tunnel options metadata
550 * @skb: pointer to skb
551 * @opt: pointer to raw tunnel option data
552 * @size: size of @opt
553 * Return: 0 on success or negative error
555 * int bpf_skb_change_proto(skb, proto, flags)
556 * Change protocol of the skb. Currently supported is v4 -> v6,
557 * v6 -> v4 transitions. The helper will also resize the skb. eBPF
558 * program is expected to fill the new headers via skb_store_bytes
559 * and lX_csum_replace.
560 * @skb: pointer to skb
561 * @proto: new skb->protocol type
563 * Return: 0 on success or negative error
565 * int bpf_skb_change_type(skb, type)
566 * Change packet type of skb.
567 * @skb: pointer to skb
568 * @type: new skb->pkt_type type
569 * Return: 0 on success or negative error
571 * int bpf_skb_under_cgroup(skb, map, index)
572 * Check cgroup2 membership of skb
573 * @skb: pointer to skb
574 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
575 * @index: index of the cgroup in the bpf_map
577 * == 0 skb failed the cgroup2 descendant test
578 * == 1 skb succeeded the cgroup2 descendant test
581 * u32 bpf_get_hash_recalc(skb)
582 * Retrieve and possibly recalculate skb->hash.
583 * @skb: pointer to skb
586 * u64 bpf_get_current_task(void)
587 * Returns current task_struct
590 * int bpf_probe_write_user(void *dst, void *src, int len)
591 * safely attempt to write to a location
592 * @dst: destination address in userspace
593 * @src: source address on stack
594 * @len: number of bytes to copy
595 * Return: 0 on success or negative error
597 * int bpf_current_task_under_cgroup(map, index)
598 * Check cgroup2 membership of current task
599 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
600 * @index: index of the cgroup in the bpf_map
602 * == 0 current failed the cgroup2 descendant test
603 * == 1 current succeeded the cgroup2 descendant test
606 * int bpf_skb_change_tail(skb, len, flags)
607 * The helper will resize the skb to the given new size, to be used f.e.
608 * with control messages.
609 * @skb: pointer to skb
610 * @len: new skb length
612 * Return: 0 on success or negative error
614 * int bpf_skb_pull_data(skb, len)
615 * The helper will pull in non-linear data in case the skb is non-linear
616 * and not all of len are part of the linear section. Only needed for
617 * read/write with direct packet access.
618 * @skb: pointer to skb
619 * @len: len to make read/writeable
620 * Return: 0 on success or negative error
622 * s64 bpf_csum_update(skb, csum)
623 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
624 * @skb: pointer to skb
626 * Return: csum on success or negative error
628 * void bpf_set_hash_invalid(skb)
629 * Invalidate current skb->hash.
630 * @skb: pointer to skb
632 * int bpf_get_numa_node_id()
633 * Return: Id of current NUMA node.
635 * int bpf_skb_change_head()
636 * Grows headroom of skb and adjusts MAC header offset accordingly.
637 * Will extends/reallocae as required automatically.
638 * May change skb data pointer and will thus invalidate any check
639 * performed for direct packet access.
640 * @skb: pointer to skb
641 * @len: length of header to be pushed in front
642 * @flags: Flags (unused for now)
643 * Return: 0 on success or negative error
645 * int bpf_xdp_adjust_head(xdp_md, delta)
646 * Adjust the xdp_md.data by delta
647 * @xdp_md: pointer to xdp_md
648 * @delta: An positive/negative integer to be added to xdp_md.data
649 * Return: 0 on success or negative on error
651 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
652 * Copy a NUL terminated string from unsafe address. In case the string
653 * length is smaller than size, the target is not padded with further NUL
654 * bytes. In case the string length is larger than size, just count-1
655 * bytes are copied and the last byte is set to NUL.
656 * @dst: destination address
657 * @size: maximum number of bytes to copy, including the trailing NUL
658 * @unsafe_ptr: unsafe address
660 * > 0 length of the string including the trailing NUL on success
663 * u64 bpf_get_socket_cookie(skb)
664 * Get the cookie for the socket stored inside sk_buff.
665 * @skb: pointer to skb
666 * Return: 8 Bytes non-decreasing number on success or 0 if the socket
667 * field is missing inside sk_buff
669 * u32 bpf_get_socket_uid(skb)
670 * Get the owner uid of the socket stored inside sk_buff.
671 * @skb: pointer to skb
672 * Return: uid of the socket owner on success or overflowuid if failed.
674 * u32 bpf_set_hash(skb, hash)
675 * Set full skb->hash.
676 * @skb: pointer to skb
679 * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
680 * Calls setsockopt. Not all opts are available, only those with
681 * integer optvals plus TCP_CONGESTION.
682 * Supported levels: SOL_SOCKET and IPPROTO_TCP
683 * @bpf_socket: pointer to bpf_socket
684 * @level: SOL_SOCKET or IPPROTO_TCP
685 * @optname: option name
686 * @optval: pointer to option value
687 * @optlen: length of optval in bytes
688 * Return: 0 or negative error
690 * int bpf_getsockopt(bpf_socket, level, optname, optval, optlen)
691 * Calls getsockopt. Not all opts are available.
692 * Supported levels: IPPROTO_TCP
693 * @bpf_socket: pointer to bpf_socket
694 * @level: IPPROTO_TCP
695 * @optname: option name
696 * @optval: pointer to option value
697 * @optlen: length of optval in bytes
698 * Return: 0 or negative error
700 * int bpf_sock_ops_cb_flags_set(bpf_sock_ops, flags)
701 * Set callback flags for sock_ops
702 * @bpf_sock_ops: pointer to bpf_sock_ops_kern struct
703 * @flags: flags value
704 * Return: 0 for no error
705 * -EINVAL if there is no full tcp socket
706 * bits in flags that are not supported by current kernel
708 * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
709 * Grow or shrink room in sk_buff.
710 * @skb: pointer to skb
711 * @len_diff: (signed) amount of room to grow/shrink
712 * @mode: operation mode (enum bpf_adj_room_mode)
713 * @flags: reserved for future use
714 * Return: 0 on success or negative error code
716 * int bpf_sk_redirect_map(map, key, flags)
717 * Redirect skb to a sock in map using key as a lookup key for the
719 * @map: pointer to sockmap
720 * @key: key to lookup sock in map
721 * @flags: reserved for future use
724 * int bpf_sock_map_update(skops, map, key, flags)
725 * @skops: pointer to bpf_sock_ops
726 * @map: pointer to sockmap to update
727 * @key: key to insert/update sock in map
728 * @flags: same flags as map update elem
730 * int bpf_xdp_adjust_meta(xdp_md, delta)
731 * Adjust the xdp_md.data_meta by delta
732 * @xdp_md: pointer to xdp_md
733 * @delta: An positive/negative integer to be added to xdp_md.data_meta
734 * Return: 0 on success or negative on error
736 * int bpf_perf_event_read_value(map, flags, buf, buf_size)
737 * read perf event counter value and perf event enabled/running time
738 * @map: pointer to perf_event_array map
739 * @flags: index of event in the map or bitmask flags
741 * @buf_size: size of the buf
742 * Return: 0 on success or negative error code
744 * int bpf_perf_prog_read_value(ctx, buf, buf_size)
745 * read perf prog attached perf event counter and enabled/running time
746 * @ctx: pointer to ctx
748 * @buf_size: size of the buf
749 * Return : 0 on success or negative error code
751 * int bpf_override_return(pt_regs, rc)
752 * @pt_regs: pointer to struct pt_regs
753 * @rc: the return value to set
755 * int bpf_msg_redirect_map(map, key, flags)
756 * Redirect msg to a sock in map using key as a lookup key for the
758 * @map: pointer to sockmap
759 * @key: key to lookup sock in map
760 * @flags: reserved for future use
763 * int bpf_bind(ctx, addr, addr_len)
764 * Bind socket to address. Only binding to IP is supported, no port can be
766 * @ctx: pointer to context of type bpf_sock_addr
767 * @addr: pointer to struct sockaddr to bind socket to
768 * @addr_len: length of sockaddr structure
769 * Return: 0 on success or negative error code
771 * int bpf_xdp_adjust_tail(xdp_md, delta)
772 * Adjust the xdp_md.data_end by delta. Only shrinking of packet's
774 * @xdp_md: pointer to xdp_md
775 * @delta: A negative integer to be added to xdp_md.data_end
776 * Return: 0 on success or negative on error
778 #define __BPF_FUNC_MAPPER(FN) \
780 FN(map_lookup_elem), \
781 FN(map_update_elem), \
782 FN(map_delete_elem), \
786 FN(get_prandom_u32), \
787 FN(get_smp_processor_id), \
788 FN(skb_store_bytes), \
789 FN(l3_csum_replace), \
790 FN(l4_csum_replace), \
792 FN(clone_redirect), \
793 FN(get_current_pid_tgid), \
794 FN(get_current_uid_gid), \
795 FN(get_current_comm), \
796 FN(get_cgroup_classid), \
799 FN(skb_get_tunnel_key), \
800 FN(skb_set_tunnel_key), \
801 FN(perf_event_read), \
803 FN(get_route_realm), \
804 FN(perf_event_output), \
805 FN(skb_load_bytes), \
808 FN(skb_get_tunnel_opt), \
809 FN(skb_set_tunnel_opt), \
810 FN(skb_change_proto), \
811 FN(skb_change_type), \
812 FN(skb_under_cgroup), \
813 FN(get_hash_recalc), \
814 FN(get_current_task), \
815 FN(probe_write_user), \
816 FN(current_task_under_cgroup), \
817 FN(skb_change_tail), \
820 FN(set_hash_invalid), \
821 FN(get_numa_node_id), \
822 FN(skb_change_head), \
823 FN(xdp_adjust_head), \
824 FN(probe_read_str), \
825 FN(get_socket_cookie), \
826 FN(get_socket_uid), \
829 FN(skb_adjust_room), \
831 FN(sk_redirect_map), \
832 FN(sock_map_update), \
833 FN(xdp_adjust_meta), \
834 FN(perf_event_read_value), \
835 FN(perf_prog_read_value), \
837 FN(override_return), \
838 FN(sock_ops_cb_flags_set), \
839 FN(msg_redirect_map), \
840 FN(msg_apply_bytes), \
841 FN(msg_cork_bytes), \
846 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
847 * function eBPF program intends to call
849 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
851 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
856 /* All flags used by eBPF helper functions, placed here. */
858 /* BPF_FUNC_skb_store_bytes flags. */
859 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
860 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
862 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
863 * First 4 bits are for passing the header field size.
865 #define BPF_F_HDR_FIELD_MASK 0xfULL
867 /* BPF_FUNC_l4_csum_replace flags. */
868 #define BPF_F_PSEUDO_HDR (1ULL << 4)
869 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
870 #define BPF_F_MARK_ENFORCE (1ULL << 6)
872 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
873 #define BPF_F_INGRESS (1ULL << 0)
875 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
876 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
878 /* BPF_FUNC_get_stackid flags. */
879 #define BPF_F_SKIP_FIELD_MASK 0xffULL
880 #define BPF_F_USER_STACK (1ULL << 8)
881 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
882 #define BPF_F_REUSE_STACKID (1ULL << 10)
884 /* BPF_FUNC_skb_set_tunnel_key flags. */
885 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
886 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
887 #define BPF_F_SEQ_NUMBER (1ULL << 3)
889 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
890 * BPF_FUNC_perf_event_read_value flags.
892 #define BPF_F_INDEX_MASK 0xffffffffULL
893 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
894 /* BPF_FUNC_perf_event_output for sk_buff input context. */
895 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
897 /* Mode for BPF_FUNC_skb_adjust_room helper. */
898 enum bpf_adj_room_mode {
902 /* user accessible mirror of in-kernel sk_buff.
903 * new fields can only be added to the end of this structure
915 __u32 ingress_ifindex;
925 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
927 __u32 remote_ip4; /* Stored in network byte order */
928 __u32 local_ip4; /* Stored in network byte order */
929 __u32 remote_ip6[4]; /* Stored in network byte order */
930 __u32 local_ip6[4]; /* Stored in network byte order */
931 __u32 remote_port; /* Stored in network byte order */
932 __u32 local_port; /* stored in host byte order */
938 struct bpf_tunnel_key {
942 __u32 remote_ipv6[4];
950 /* Generic BPF return codes which all BPF program types may support.
951 * The values are binary compatible with their TC_ACT_* counter-part to
952 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
955 * XDP is handled seprately, see XDP_*.
963 /* >127 are reserved for prog type specific return codes */
973 __u32 src_ip4; /* Allows 1,2,4-byte read.
974 * Stored in network byte order.
976 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
977 * Stored in network byte order.
979 __u32 src_port; /* Allows 4-byte read.
980 * Stored in host byte order
984 #define XDP_PACKET_HEADROOM 256
986 /* User return codes for XDP prog type.
987 * A valid XDP program must return one of these defined values. All other
988 * return codes are reserved for future use. Unknown return codes will
989 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
999 /* user accessible metadata for XDP packet hook
1000 * new fields must be added to the end of this structure
1006 /* Below access go through struct xdp_rxq_info */
1007 __u32 ingress_ifindex; /* rxq->dev->ifindex */
1008 __u32 rx_queue_index; /* rxq->queue_index */
1016 /* user accessible metadata for SK_MSG packet hook, new fields must
1017 * be added to the end of this structure
1024 #define BPF_TAG_SIZE 8
1026 struct bpf_prog_info {
1029 __u8 tag[BPF_TAG_SIZE];
1030 __u32 jited_prog_len;
1031 __u32 xlated_prog_len;
1032 __aligned_u64 jited_prog_insns;
1033 __aligned_u64 xlated_prog_insns;
1034 __u64 load_time; /* ns since boottime */
1035 __u32 created_by_uid;
1037 __aligned_u64 map_ids;
1038 char name[BPF_OBJ_NAME_LEN];
1042 } __attribute__((aligned(8)));
1044 struct bpf_map_info {
1051 char name[BPF_OBJ_NAME_LEN];
1055 } __attribute__((aligned(8)));
1057 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
1058 * by user and intended to be used by socket (e.g. to bind to, depends on
1059 * attach attach type).
1061 struct bpf_sock_addr {
1062 __u32 user_family; /* Allows 4-byte read, but no write. */
1063 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
1064 * Stored in network byte order.
1066 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
1067 * Stored in network byte order.
1069 __u32 user_port; /* Allows 4-byte read and write.
1070 * Stored in network byte order
1072 __u32 family; /* Allows 4-byte read, but no write */
1073 __u32 type; /* Allows 4-byte read, but no write */
1074 __u32 protocol; /* Allows 4-byte read, but no write */
1077 /* User bpf_sock_ops struct to access socket values and specify request ops
1078 * and their replies.
1079 * Some of this fields are in network (bigendian) byte order and may need
1080 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
1081 * New fields can only be added at the end of this structure
1083 struct bpf_sock_ops {
1086 __u32 args[4]; /* Optionally passed to bpf program */
1087 __u32 reply; /* Returned by bpf program */
1088 __u32 replylong[4]; /* Optionally returned by bpf prog */
1091 __u32 remote_ip4; /* Stored in network byte order */
1092 __u32 local_ip4; /* Stored in network byte order */
1093 __u32 remote_ip6[4]; /* Stored in network byte order */
1094 __u32 local_ip6[4]; /* Stored in network byte order */
1095 __u32 remote_port; /* Stored in network byte order */
1096 __u32 local_port; /* stored in host byte order */
1097 __u32 is_fullsock; /* Some TCP fields are only valid if
1098 * there is a full socket. If not, the
1099 * fields read as zero.
1102 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
1103 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
1112 __u32 rate_delivered;
1113 __u32 rate_interval_us;
1116 __u32 total_retrans;
1120 __u32 data_segs_out;
1124 __u64 bytes_received;
1128 /* Definitions for bpf_sock_ops_cb_flags */
1129 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
1130 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
1131 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
1132 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
1133 * supported cb flags
1136 /* List of known BPF sock_ops operators.
1137 * New entries can only be added at the end
1141 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
1142 * -1 if default value should be used
1144 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
1145 * window (in packets) or -1 if default
1146 * value should be used
1148 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
1149 * active connection is initialized
1151 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
1152 * active connection is
1155 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
1156 * passive connection is
1159 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
1162 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
1163 * based on the path and may be
1164 * dependent on the congestion control
1165 * algorithm. In general it indicates
1166 * a congestion threshold. RTTs above
1167 * this indicate congestion
1169 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
1170 * Arg1: value of icsk_retransmits
1171 * Arg2: value of icsk_rto
1172 * Arg3: whether RTO has expired
1174 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
1175 * Arg1: sequence number of 1st byte
1177 * Arg3: return value of
1178 * tcp_transmit_skb (0 => success)
1180 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
1186 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
1187 * changes between the TCP and BPF versions. Ideally this should never happen.
1188 * If it does, we need to add code to convert them before calling
1189 * the BPF sock_ops function.
1192 BPF_TCP_ESTABLISHED = 1,
1202 BPF_TCP_CLOSING, /* Now a valid state */
1203 BPF_TCP_NEW_SYN_RECV,
1205 BPF_TCP_MAX_STATES /* Leave at the end! */
1208 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
1209 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
1211 struct bpf_perf_event_value {
1217 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
1218 #define BPF_DEVCG_ACC_READ (1ULL << 1)
1219 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
1221 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
1222 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
1224 struct bpf_cgroup_dev_ctx {
1225 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
1231 struct bpf_raw_tracepoint_args {
1235 #endif /* _UAPI__LINUX_BPF_H__ */