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_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_XADD 0xc0 /* exclusive add */
25 #define BPF_MOV 0xb0 /* mov reg to reg */
26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28 /* change endianness of a register */
29 #define BPF_END 0xd0 /* flags for endianness conversion: */
30 #define BPF_TO_LE 0x00 /* convert to little-endian */
31 #define BPF_TO_BE 0x08 /* convert to big-endian */
32 #define BPF_FROM_LE BPF_TO_LE
33 #define BPF_FROM_BE BPF_TO_BE
36 #define BPF_JNE 0x50 /* jump != */
37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
43 #define BPF_CALL 0x80 /* function call */
44 #define BPF_EXIT 0x90 /* function return */
46 /* Register numbers */
62 /* BPF has 10 general purpose 64-bit registers and stack frame. */
63 #define MAX_BPF_REG __MAX_BPF_REG
66 __u8 code
; /* opcode */
67 __u8 dst_reg
:4; /* dest register */
68 __u8 src_reg
:4; /* source register */
69 __s16 off
; /* signed offset */
70 __s32 imm
; /* signed immediate constant */
73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74 struct bpf_lpm_trie_key
{
75 __u32 prefixlen
; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data
[0]; /* Arbitrary size */
79 struct bpf_cgroup_storage_key
{
80 __u64 cgroup_inode_id
; /* cgroup inode id */
81 __u32 attach_type
; /* program attach type */
84 /* BPF syscall commands, see bpf(2) man-page for details. */
99 BPF_PROG_GET_FD_BY_ID
,
100 BPF_MAP_GET_FD_BY_ID
,
101 BPF_OBJ_GET_INFO_BY_FD
,
103 BPF_RAW_TRACEPOINT_OPEN
,
105 BPF_BTF_GET_FD_BY_ID
,
107 BPF_MAP_LOOKUP_AND_DELETE_ELEM
,
115 BPF_MAP_TYPE_PROG_ARRAY
,
116 BPF_MAP_TYPE_PERF_EVENT_ARRAY
,
117 BPF_MAP_TYPE_PERCPU_HASH
,
118 BPF_MAP_TYPE_PERCPU_ARRAY
,
119 BPF_MAP_TYPE_STACK_TRACE
,
120 BPF_MAP_TYPE_CGROUP_ARRAY
,
121 BPF_MAP_TYPE_LRU_HASH
,
122 BPF_MAP_TYPE_LRU_PERCPU_HASH
,
123 BPF_MAP_TYPE_LPM_TRIE
,
124 BPF_MAP_TYPE_ARRAY_OF_MAPS
,
125 BPF_MAP_TYPE_HASH_OF_MAPS
,
127 BPF_MAP_TYPE_SOCKMAP
,
130 BPF_MAP_TYPE_SOCKHASH
,
131 BPF_MAP_TYPE_CGROUP_STORAGE
,
132 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY
,
133 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
,
136 BPF_MAP_TYPE_SK_STORAGE
,
139 /* Note that tracing related programs such as
140 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
141 * are not subject to a stable API since kernel internal data
142 * structures can change from release to release and may
143 * therefore break existing tracing BPF programs. Tracing BPF
144 * programs correspond to /a/ specific kernel which is to be
145 * analyzed, and not /a/ specific kernel /and/ all future ones.
148 BPF_PROG_TYPE_UNSPEC
,
149 BPF_PROG_TYPE_SOCKET_FILTER
,
150 BPF_PROG_TYPE_KPROBE
,
151 BPF_PROG_TYPE_SCHED_CLS
,
152 BPF_PROG_TYPE_SCHED_ACT
,
153 BPF_PROG_TYPE_TRACEPOINT
,
155 BPF_PROG_TYPE_PERF_EVENT
,
156 BPF_PROG_TYPE_CGROUP_SKB
,
157 BPF_PROG_TYPE_CGROUP_SOCK
,
158 BPF_PROG_TYPE_LWT_IN
,
159 BPF_PROG_TYPE_LWT_OUT
,
160 BPF_PROG_TYPE_LWT_XMIT
,
161 BPF_PROG_TYPE_SOCK_OPS
,
162 BPF_PROG_TYPE_SK_SKB
,
163 BPF_PROG_TYPE_CGROUP_DEVICE
,
164 BPF_PROG_TYPE_SK_MSG
,
165 BPF_PROG_TYPE_RAW_TRACEPOINT
,
166 BPF_PROG_TYPE_CGROUP_SOCK_ADDR
,
167 BPF_PROG_TYPE_LWT_SEG6LOCAL
,
168 BPF_PROG_TYPE_LIRC_MODE2
,
169 BPF_PROG_TYPE_SK_REUSEPORT
,
170 BPF_PROG_TYPE_FLOW_DISSECTOR
,
171 BPF_PROG_TYPE_CGROUP_SYSCTL
,
172 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE
,
175 enum bpf_attach_type
{
176 BPF_CGROUP_INET_INGRESS
,
177 BPF_CGROUP_INET_EGRESS
,
178 BPF_CGROUP_INET_SOCK_CREATE
,
180 BPF_SK_SKB_STREAM_PARSER
,
181 BPF_SK_SKB_STREAM_VERDICT
,
184 BPF_CGROUP_INET4_BIND
,
185 BPF_CGROUP_INET6_BIND
,
186 BPF_CGROUP_INET4_CONNECT
,
187 BPF_CGROUP_INET6_CONNECT
,
188 BPF_CGROUP_INET4_POST_BIND
,
189 BPF_CGROUP_INET6_POST_BIND
,
190 BPF_CGROUP_UDP4_SENDMSG
,
191 BPF_CGROUP_UDP6_SENDMSG
,
195 __MAX_BPF_ATTACH_TYPE
198 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
200 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
202 * NONE(default): No further bpf programs allowed in the subtree.
204 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
205 * the program in this cgroup yields to sub-cgroup program.
207 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
208 * that cgroup program gets run in addition to the program in this cgroup.
210 * Only one program is allowed to be attached to a cgroup with
211 * NONE or BPF_F_ALLOW_OVERRIDE flag.
212 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
213 * release old program and attach the new one. Attach flags has to match.
215 * Multiple programs are allowed to be attached to a cgroup with
216 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
217 * (those that were attached first, run first)
218 * The programs of sub-cgroup are executed first, then programs of
219 * this cgroup and then programs of parent cgroup.
220 * When children program makes decision (like picking TCP CA or sock bind)
221 * parent program has a chance to override it.
223 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
224 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
226 * cgrp1 (MULTI progs A, B) ->
227 * cgrp2 (OVERRIDE prog C) ->
228 * cgrp3 (MULTI prog D) ->
229 * cgrp4 (OVERRIDE prog E) ->
230 * cgrp5 (NONE prog F)
231 * the event in cgrp5 triggers execution of F,D,A,B in that order.
232 * if prog F is detached, the execution is E,D,A,B
233 * if prog F and D are detached, the execution is E,A,B
234 * if prog F, E and D are detached, the execution is C,A,B
236 * All eligible programs are executed regardless of return code from
239 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
240 #define BPF_F_ALLOW_MULTI (1U << 1)
242 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
243 * verifier will perform strict alignment checking as if the kernel
244 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
245 * and NET_IP_ALIGN defined to 2.
247 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
249 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
250 * verifier will allow any alignment whatsoever. On platforms
251 * with strict alignment requirements for loads ands stores (such
252 * as sparc and mips) the verifier validates that all loads and
253 * stores provably follow this requirement. This flag turns that
254 * checking and enforcement off.
256 * It is mostly used for testing when we want to validate the
257 * context and memory access aspects of the verifier, but because
258 * of an unaligned access the alignment check would trigger before
259 * the one we are interested in.
261 #define BPF_F_ANY_ALIGNMENT (1U << 1)
263 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
266 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
267 * insn[0].imm: map fd map fd
268 * insn[1].imm: 0 offset into value
271 * ldimm64 rewrite: address of map address of map[0]+offset
272 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
274 #define BPF_PSEUDO_MAP_FD 1
275 #define BPF_PSEUDO_MAP_VALUE 2
277 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
278 * offset to another bpf function
280 #define BPF_PSEUDO_CALL 1
282 /* flags for BPF_MAP_UPDATE_ELEM command */
283 #define BPF_ANY 0 /* create new element or update existing */
284 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
285 #define BPF_EXIST 2 /* update existing element */
286 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
288 /* flags for BPF_MAP_CREATE command */
289 #define BPF_F_NO_PREALLOC (1U << 0)
290 /* Instead of having one common LRU list in the
291 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
292 * which can scale and perform better.
293 * Note, the LRU nodes (including free nodes) cannot be moved
294 * across different LRU lists.
296 #define BPF_F_NO_COMMON_LRU (1U << 1)
297 /* Specify numa node during map creation */
298 #define BPF_F_NUMA_NODE (1U << 2)
300 #define BPF_OBJ_NAME_LEN 16U
302 /* Flags for accessing BPF object from syscall side. */
303 #define BPF_F_RDONLY (1U << 3)
304 #define BPF_F_WRONLY (1U << 4)
306 /* Flag for stack_map, store build_id+offset instead of pointer */
307 #define BPF_F_STACK_BUILD_ID (1U << 5)
309 /* Zero-initialize hash function seed. This should only be used for testing. */
310 #define BPF_F_ZERO_SEED (1U << 6)
312 /* Flags for accessing BPF object from program side. */
313 #define BPF_F_RDONLY_PROG (1U << 7)
314 #define BPF_F_WRONLY_PROG (1U << 8)
316 /* flags for BPF_PROG_QUERY */
317 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
319 enum bpf_stack_build_id_status
{
320 /* user space need an empty entry to identify end of a trace */
321 BPF_STACK_BUILD_ID_EMPTY
= 0,
322 /* with valid build_id and offset */
323 BPF_STACK_BUILD_ID_VALID
= 1,
324 /* couldn't get build_id, fallback to ip */
325 BPF_STACK_BUILD_ID_IP
= 2,
328 #define BPF_BUILD_ID_SIZE 20
329 struct bpf_stack_build_id
{
331 unsigned char build_id
[BPF_BUILD_ID_SIZE
];
339 struct { /* anonymous struct used by BPF_MAP_CREATE command */
340 __u32 map_type
; /* one of enum bpf_map_type */
341 __u32 key_size
; /* size of key in bytes */
342 __u32 value_size
; /* size of value in bytes */
343 __u32 max_entries
; /* max number of entries in a map */
344 __u32 map_flags
; /* BPF_MAP_CREATE related
345 * flags defined above.
347 __u32 inner_map_fd
; /* fd pointing to the inner map */
348 __u32 numa_node
; /* numa node (effective only if
349 * BPF_F_NUMA_NODE is set).
351 char map_name
[BPF_OBJ_NAME_LEN
];
352 __u32 map_ifindex
; /* ifindex of netdev to create on */
353 __u32 btf_fd
; /* fd pointing to a BTF type data */
354 __u32 btf_key_type_id
; /* BTF type_id of the key */
355 __u32 btf_value_type_id
; /* BTF type_id of the value */
358 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
363 __aligned_u64 next_key
;
368 struct { /* anonymous struct used by BPF_PROG_LOAD command */
369 __u32 prog_type
; /* one of enum bpf_prog_type */
372 __aligned_u64 license
;
373 __u32 log_level
; /* verbosity level of verifier */
374 __u32 log_size
; /* size of user buffer */
375 __aligned_u64 log_buf
; /* user supplied buffer */
376 __u32 kern_version
; /* not used */
378 char prog_name
[BPF_OBJ_NAME_LEN
];
379 __u32 prog_ifindex
; /* ifindex of netdev to prep for */
380 /* For some prog types expected attach type must be known at
381 * load time to verify attach type specific parts of prog
382 * (context accesses, allowed helpers, etc).
384 __u32 expected_attach_type
;
385 __u32 prog_btf_fd
; /* fd pointing to BTF type data */
386 __u32 func_info_rec_size
; /* userspace bpf_func_info size */
387 __aligned_u64 func_info
; /* func info */
388 __u32 func_info_cnt
; /* number of bpf_func_info records */
389 __u32 line_info_rec_size
; /* userspace bpf_line_info size */
390 __aligned_u64 line_info
; /* line info */
391 __u32 line_info_cnt
; /* number of bpf_line_info records */
394 struct { /* anonymous struct used by BPF_OBJ_* commands */
395 __aligned_u64 pathname
;
400 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
401 __u32 target_fd
; /* container object to attach to */
402 __u32 attach_bpf_fd
; /* eBPF program to attach */
407 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
410 __u32 data_size_in
; /* input: len of data_in */
411 __u32 data_size_out
; /* input/output: len of data_out
412 * returns ENOSPC if data_out
415 __aligned_u64 data_in
;
416 __aligned_u64 data_out
;
419 __u32 ctx_size_in
; /* input: len of ctx_in */
420 __u32 ctx_size_out
; /* input/output: len of ctx_out
421 * returns ENOSPC if ctx_out
424 __aligned_u64 ctx_in
;
425 __aligned_u64 ctx_out
;
428 struct { /* anonymous struct used by BPF_*_GET_*_ID */
439 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
445 struct { /* anonymous struct used by BPF_PROG_QUERY command */
446 __u32 target_fd
; /* container object to query */
450 __aligned_u64 prog_ids
;
459 struct { /* anonymous struct for BPF_BTF_LOAD */
461 __aligned_u64 btf_log_buf
;
468 __u32 pid
; /* input: pid */
469 __u32 fd
; /* input: fd */
470 __u32 flags
; /* input: flags */
471 __u32 buf_len
; /* input/output: buf len */
472 __aligned_u64 buf
; /* input/output:
473 * tp_name for tracepoint
475 * filename for uprobe
477 __u32 prog_id
; /* output: prod_id */
478 __u32 fd_type
; /* output: BPF_FD_TYPE_* */
479 __u64 probe_offset
; /* output: probe_offset */
480 __u64 probe_addr
; /* output: probe_addr */
482 } __attribute__((aligned(8)));
484 /* The description below is an attempt at providing documentation to eBPF
485 * developers about the multiple available eBPF helper functions. It can be
486 * parsed and used to produce a manual page. The workflow is the following,
487 * and requires the rst2man utility:
489 * $ ./scripts/bpf_helpers_doc.py \
490 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
491 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
492 * $ man /tmp/bpf-helpers.7
494 * Note that in order to produce this external documentation, some RST
495 * formatting is used in the descriptions to get "bold" and "italics" in
496 * manual pages. Also note that the few trailing white spaces are
497 * intentional, removing them would break paragraphs for rst2man.
499 * Start of BPF helper function descriptions:
501 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
503 * Perform a lookup in *map* for an entry associated to *key*.
505 * Map value associated to *key*, or **NULL** if no entry was
508 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
510 * Add or update the value of the entry associated to *key* in
511 * *map* with *value*. *flags* is one of:
514 * The entry for *key* must not exist in the map.
516 * The entry for *key* must already exist in the map.
518 * No condition on the existence of the entry for *key*.
520 * Flag value **BPF_NOEXIST** cannot be used for maps of types
521 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
522 * elements always exist), the helper would return an error.
524 * 0 on success, or a negative error in case of failure.
526 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
528 * Delete entry with *key* from *map*.
530 * 0 on success, or a negative error in case of failure.
532 * int bpf_probe_read(void *dst, u32 size, const void *src)
534 * For tracing programs, safely attempt to read *size* bytes from
535 * address *src* and store the data in *dst*.
537 * 0 on success, or a negative error in case of failure.
539 * u64 bpf_ktime_get_ns(void)
541 * Return the time elapsed since system boot, in nanoseconds.
545 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
547 * This helper is a "printk()-like" facility for debugging. It
548 * prints a message defined by format *fmt* (of size *fmt_size*)
549 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
550 * available. It can take up to three additional **u64**
551 * arguments (as an eBPF helpers, the total number of arguments is
554 * Each time the helper is called, it appends a line to the trace.
555 * The format of the trace is customizable, and the exact output
556 * one will get depends on the options set in
557 * *\/sys/kernel/debug/tracing/trace_options* (see also the
558 * *README* file under the same directory). However, it usually
559 * defaults to something like:
563 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
567 * * ``telnet`` is the name of the current task.
568 * * ``470`` is the PID of the current task.
569 * * ``001`` is the CPU number on which the task is
571 * * In ``.N..``, each character refers to a set of
572 * options (whether irqs are enabled, scheduling
573 * options, whether hard/softirqs are running, level of
574 * preempt_disabled respectively). **N** means that
575 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
577 * * ``419421.045894`` is a timestamp.
578 * * ``0x00000001`` is a fake value used by BPF for the
579 * instruction pointer register.
580 * * ``<formatted msg>`` is the message formatted with
583 * The conversion specifiers supported by *fmt* are similar, but
584 * more limited than for printk(). They are **%d**, **%i**,
585 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
586 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
587 * of field, padding with zeroes, etc.) is available, and the
588 * helper will return **-EINVAL** (but print nothing) if it
589 * encounters an unknown specifier.
591 * Also, note that **bpf_trace_printk**\ () is slow, and should
592 * only be used for debugging purposes. For this reason, a notice
593 * bloc (spanning several lines) is printed to kernel logs and
594 * states that the helper should not be used "for production use"
595 * the first time this helper is used (or more precisely, when
596 * **trace_printk**\ () buffers are allocated). For passing values
597 * to user space, perf events should be preferred.
599 * The number of bytes written to the buffer, or a negative error
600 * in case of failure.
602 * u32 bpf_get_prandom_u32(void)
604 * Get a pseudo-random number.
606 * From a security point of view, this helper uses its own
607 * pseudo-random internal state, and cannot be used to infer the
608 * seed of other random functions in the kernel. However, it is
609 * essential to note that the generator used by the helper is not
610 * cryptographically secure.
612 * A random 32-bit unsigned value.
614 * u32 bpf_get_smp_processor_id(void)
616 * Get the SMP (symmetric multiprocessing) processor id. Note that
617 * all programs run with preemption disabled, which means that the
618 * SMP processor id is stable during all the execution of the
621 * The SMP id of the processor running the program.
623 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
625 * Store *len* bytes from address *from* into the packet
626 * associated to *skb*, at *offset*. *flags* are a combination of
627 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
628 * checksum for the packet after storing the bytes) and
629 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
630 * **->swhash** and *skb*\ **->l4hash** to 0).
632 * A call to this helper is susceptible to change the underlying
633 * packet buffer. Therefore, at load time, all checks on pointers
634 * previously done by the verifier are invalidated and must be
635 * performed again, if the helper is used in combination with
636 * direct packet access.
638 * 0 on success, or a negative error in case of failure.
640 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
642 * Recompute the layer 3 (e.g. IP) checksum for the packet
643 * associated to *skb*. Computation is incremental, so the helper
644 * must know the former value of the header field that was
645 * modified (*from*), the new value of this field (*to*), and the
646 * number of bytes (2 or 4) for this field, stored in *size*.
647 * Alternatively, it is possible to store the difference between
648 * the previous and the new values of the header field in *to*, by
649 * setting *from* and *size* to 0. For both methods, *offset*
650 * indicates the location of the IP checksum within the packet.
652 * This helper works in combination with **bpf_csum_diff**\ (),
653 * which does not update the checksum in-place, but offers more
654 * flexibility and can handle sizes larger than 2 or 4 for the
655 * checksum to update.
657 * A call to this helper is susceptible to change the underlying
658 * packet buffer. Therefore, at load time, all checks on pointers
659 * previously done by the verifier are invalidated and must be
660 * performed again, if the helper is used in combination with
661 * direct packet access.
663 * 0 on success, or a negative error in case of failure.
665 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
667 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
668 * packet associated to *skb*. Computation is incremental, so the
669 * helper must know the former value of the header field that was
670 * modified (*from*), the new value of this field (*to*), and the
671 * number of bytes (2 or 4) for this field, stored on the lowest
672 * four bits of *flags*. Alternatively, it is possible to store
673 * the difference between the previous and the new values of the
674 * header field in *to*, by setting *from* and the four lowest
675 * bits of *flags* to 0. For both methods, *offset* indicates the
676 * location of the IP checksum within the packet. In addition to
677 * the size of the field, *flags* can be added (bitwise OR) actual
678 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
679 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
680 * for updates resulting in a null checksum the value is set to
681 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
682 * the checksum is to be computed against a pseudo-header.
684 * This helper works in combination with **bpf_csum_diff**\ (),
685 * which does not update the checksum in-place, but offers more
686 * flexibility and can handle sizes larger than 2 or 4 for the
687 * checksum to update.
689 * A call to this helper is susceptible to change the underlying
690 * packet buffer. Therefore, at load time, all checks on pointers
691 * previously done by the verifier are invalidated and must be
692 * performed again, if the helper is used in combination with
693 * direct packet access.
695 * 0 on success, or a negative error in case of failure.
697 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
699 * This special helper is used to trigger a "tail call", or in
700 * other words, to jump into another eBPF program. The same stack
701 * frame is used (but values on stack and in registers for the
702 * caller are not accessible to the callee). This mechanism allows
703 * for program chaining, either for raising the maximum number of
704 * available eBPF instructions, or to execute given programs in
705 * conditional blocks. For security reasons, there is an upper
706 * limit to the number of successive tail calls that can be
709 * Upon call of this helper, the program attempts to jump into a
710 * program referenced at index *index* in *prog_array_map*, a
711 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
712 * *ctx*, a pointer to the context.
714 * If the call succeeds, the kernel immediately runs the first
715 * instruction of the new program. This is not a function call,
716 * and it never returns to the previous program. If the call
717 * fails, then the helper has no effect, and the caller continues
718 * to run its subsequent instructions. A call can fail if the
719 * destination program for the jump does not exist (i.e. *index*
720 * is superior to the number of entries in *prog_array_map*), or
721 * if the maximum number of tail calls has been reached for this
722 * chain of programs. This limit is defined in the kernel by the
723 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
724 * which is currently set to 32.
726 * 0 on success, or a negative error in case of failure.
728 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
730 * Clone and redirect the packet associated to *skb* to another
731 * net device of index *ifindex*. Both ingress and egress
732 * interfaces can be used for redirection. The **BPF_F_INGRESS**
733 * value in *flags* is used to make the distinction (ingress path
734 * is selected if the flag is present, egress path otherwise).
735 * This is the only flag supported for now.
737 * In comparison with **bpf_redirect**\ () helper,
738 * **bpf_clone_redirect**\ () has the associated cost of
739 * duplicating the packet buffer, but this can be executed out of
740 * the eBPF program. Conversely, **bpf_redirect**\ () is more
741 * efficient, but it is handled through an action code where the
742 * redirection happens only after the eBPF program has returned.
744 * A call to this helper is susceptible to change the underlying
745 * packet buffer. Therefore, at load time, all checks on pointers
746 * previously done by the verifier are invalidated and must be
747 * performed again, if the helper is used in combination with
748 * direct packet access.
750 * 0 on success, or a negative error in case of failure.
752 * u64 bpf_get_current_pid_tgid(void)
754 * A 64-bit integer containing the current tgid and pid, and
756 * *current_task*\ **->tgid << 32 \|**
757 * *current_task*\ **->pid**.
759 * u64 bpf_get_current_uid_gid(void)
761 * A 64-bit integer containing the current GID and UID, and
762 * created as such: *current_gid* **<< 32 \|** *current_uid*.
764 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
766 * Copy the **comm** attribute of the current task into *buf* of
767 * *size_of_buf*. The **comm** attribute contains the name of
768 * the executable (excluding the path) for the current task. The
769 * *size_of_buf* must be strictly positive. On success, the
770 * helper makes sure that the *buf* is NUL-terminated. On failure,
771 * it is filled with zeroes.
773 * 0 on success, or a negative error in case of failure.
775 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
777 * Retrieve the classid for the current task, i.e. for the net_cls
778 * cgroup to which *skb* belongs.
780 * This helper can be used on TC egress path, but not on ingress.
782 * The net_cls cgroup provides an interface to tag network packets
783 * based on a user-provided identifier for all traffic coming from
784 * the tasks belonging to the related cgroup. See also the related
785 * kernel documentation, available from the Linux sources in file
786 * *Documentation/cgroup-v1/net_cls.txt*.
788 * The Linux kernel has two versions for cgroups: there are
789 * cgroups v1 and cgroups v2. Both are available to users, who can
790 * use a mixture of them, but note that the net_cls cgroup is for
791 * cgroup v1 only. This makes it incompatible with BPF programs
792 * run on cgroups, which is a cgroup-v2-only feature (a socket can
793 * only hold data for one version of cgroups at a time).
795 * This helper is only available is the kernel was compiled with
796 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
797 * "**y**" or to "**m**".
799 * The classid, or 0 for the default unconfigured classid.
801 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
803 * Push a *vlan_tci* (VLAN tag control information) of protocol
804 * *vlan_proto* to the packet associated to *skb*, then update
805 * the checksum. Note that if *vlan_proto* is different from
806 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
807 * be **ETH_P_8021Q**.
809 * A call to this helper is susceptible to change the underlying
810 * packet buffer. Therefore, at load time, all checks on pointers
811 * previously done by the verifier are invalidated and must be
812 * performed again, if the helper is used in combination with
813 * direct packet access.
815 * 0 on success, or a negative error in case of failure.
817 * int bpf_skb_vlan_pop(struct sk_buff *skb)
819 * Pop a VLAN header from the packet associated to *skb*.
821 * A call to this helper is susceptible to change the underlying
822 * packet buffer. Therefore, at load time, all checks on pointers
823 * previously done by the verifier are invalidated and must be
824 * performed again, if the helper is used in combination with
825 * direct packet access.
827 * 0 on success, or a negative error in case of failure.
829 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
831 * Get tunnel metadata. This helper takes a pointer *key* to an
832 * empty **struct bpf_tunnel_key** of **size**, that will be
833 * filled with tunnel metadata for the packet associated to *skb*.
834 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
835 * indicates that the tunnel is based on IPv6 protocol instead of
838 * The **struct bpf_tunnel_key** is an object that generalizes the
839 * principal parameters used by various tunneling protocols into a
840 * single struct. This way, it can be used to easily make a
841 * decision based on the contents of the encapsulation header,
842 * "summarized" in this struct. In particular, it holds the IP
843 * address of the remote end (IPv4 or IPv6, depending on the case)
844 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
845 * this struct exposes the *key*\ **->tunnel_id**, which is
846 * generally mapped to a VNI (Virtual Network Identifier), making
847 * it programmable together with the **bpf_skb_set_tunnel_key**\
850 * Let's imagine that the following code is part of a program
851 * attached to the TC ingress interface, on one end of a GRE
852 * tunnel, and is supposed to filter out all messages coming from
853 * remote ends with IPv4 address other than 10.0.0.1:
858 * struct bpf_tunnel_key key = {};
860 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
862 * return TC_ACT_SHOT; // drop packet
864 * if (key.remote_ipv4 != 0x0a000001)
865 * return TC_ACT_SHOT; // drop packet
867 * return TC_ACT_OK; // accept packet
869 * This interface can also be used with all encapsulation devices
870 * that can operate in "collect metadata" mode: instead of having
871 * one network device per specific configuration, the "collect
872 * metadata" mode only requires a single device where the
873 * configuration can be extracted from this helper.
875 * This can be used together with various tunnels such as VXLan,
876 * Geneve, GRE or IP in IP (IPIP).
878 * 0 on success, or a negative error in case of failure.
880 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
882 * Populate tunnel metadata for packet associated to *skb.* The
883 * tunnel metadata is set to the contents of *key*, of *size*. The
884 * *flags* can be set to a combination of the following values:
886 * **BPF_F_TUNINFO_IPV6**
887 * Indicate that the tunnel is based on IPv6 protocol
889 * **BPF_F_ZERO_CSUM_TX**
890 * For IPv4 packets, add a flag to tunnel metadata
891 * indicating that checksum computation should be skipped
892 * and checksum set to zeroes.
893 * **BPF_F_DONT_FRAGMENT**
894 * Add a flag to tunnel metadata indicating that the
895 * packet should not be fragmented.
896 * **BPF_F_SEQ_NUMBER**
897 * Add a flag to tunnel metadata indicating that a
898 * sequence number should be added to tunnel header before
899 * sending the packet. This flag was added for GRE
900 * encapsulation, but might be used with other protocols
901 * as well in the future.
903 * Here is a typical usage on the transmit path:
907 * struct bpf_tunnel_key key;
909 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
910 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
912 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
913 * helper for additional information.
915 * 0 on success, or a negative error in case of failure.
917 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
919 * Read the value of a perf event counter. This helper relies on a
920 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
921 * the perf event counter is selected when *map* is updated with
922 * perf event file descriptors. The *map* is an array whose size
923 * is the number of available CPUs, and each cell contains a value
924 * relative to one CPU. The value to retrieve is indicated by
925 * *flags*, that contains the index of the CPU to look up, masked
926 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
927 * **BPF_F_CURRENT_CPU** to indicate that the value for the
928 * current CPU should be retrieved.
930 * Note that before Linux 4.13, only hardware perf event can be
933 * Also, be aware that the newer helper
934 * **bpf_perf_event_read_value**\ () is recommended over
935 * **bpf_perf_event_read**\ () in general. The latter has some ABI
936 * quirks where error and counter value are used as a return code
937 * (which is wrong to do since ranges may overlap). This issue is
938 * fixed with **bpf_perf_event_read_value**\ (), which at the same
939 * time provides more features over the **bpf_perf_event_read**\
940 * () interface. Please refer to the description of
941 * **bpf_perf_event_read_value**\ () for details.
943 * The value of the perf event counter read from the map, or a
944 * negative error code in case of failure.
946 * int bpf_redirect(u32 ifindex, u64 flags)
948 * Redirect the packet to another net device of index *ifindex*.
949 * This helper is somewhat similar to **bpf_clone_redirect**\
950 * (), except that the packet is not cloned, which provides
951 * increased performance.
953 * Except for XDP, both ingress and egress interfaces can be used
954 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
955 * to make the distinction (ingress path is selected if the flag
956 * is present, egress path otherwise). Currently, XDP only
957 * supports redirection to the egress interface, and accepts no
960 * The same effect can be attained with the more generic
961 * **bpf_redirect_map**\ (), which requires specific maps to be
962 * used but offers better performance.
964 * For XDP, the helper returns **XDP_REDIRECT** on success or
965 * **XDP_ABORTED** on error. For other program types, the values
966 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
969 * u32 bpf_get_route_realm(struct sk_buff *skb)
971 * Retrieve the realm or the route, that is to say the
972 * **tclassid** field of the destination for the *skb*. The
973 * indentifier retrieved is a user-provided tag, similar to the
974 * one used with the net_cls cgroup (see description for
975 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
976 * held by a route (a destination entry), not by a task.
978 * Retrieving this identifier works with the clsact TC egress hook
979 * (see also **tc-bpf(8)**), or alternatively on conventional
980 * classful egress qdiscs, but not on TC ingress path. In case of
981 * clsact TC egress hook, this has the advantage that, internally,
982 * the destination entry has not been dropped yet in the transmit
983 * path. Therefore, the destination entry does not need to be
984 * artificially held via **netif_keep_dst**\ () for a classful
985 * qdisc until the *skb* is freed.
987 * This helper is available only if the kernel was compiled with
988 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
990 * The realm of the route for the packet associated to *skb*, or 0
993 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
995 * Write raw *data* blob into a special BPF perf event held by
996 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
997 * event must have the following attributes: **PERF_SAMPLE_RAW**
998 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
999 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1001 * The *flags* are used to indicate the index in *map* for which
1002 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1003 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1004 * to indicate that the index of the current CPU core should be
1007 * The value to write, of *size*, is passed through eBPF stack and
1008 * pointed by *data*.
1010 * The context of the program *ctx* needs also be passed to the
1013 * On user space, a program willing to read the values needs to
1014 * call **perf_event_open**\ () on the perf event (either for
1015 * one or for all CPUs) and to store the file descriptor into the
1016 * *map*. This must be done before the eBPF program can send data
1017 * into it. An example is available in file
1018 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1019 * tree (the eBPF program counterpart is in
1020 * *samples/bpf/trace_output_kern.c*).
1022 * **bpf_perf_event_output**\ () achieves better performance
1023 * than **bpf_trace_printk**\ () for sharing data with user
1024 * space, and is much better suitable for streaming data from eBPF
1027 * Note that this helper is not restricted to tracing use cases
1028 * and can be used with programs attached to TC or XDP as well,
1029 * where it allows for passing data to user space listeners. Data
1032 * * Only custom structs,
1033 * * Only the packet payload, or
1034 * * A combination of both.
1036 * 0 on success, or a negative error in case of failure.
1038 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1040 * This helper was provided as an easy way to load data from a
1041 * packet. It can be used to load *len* bytes from *offset* from
1042 * the packet associated to *skb*, into the buffer pointed by
1045 * Since Linux 4.7, usage of this helper has mostly been replaced
1046 * by "direct packet access", enabling packet data to be
1047 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1048 * pointing respectively to the first byte of packet data and to
1049 * the byte after the last byte of packet data. However, it
1050 * remains useful if one wishes to read large quantities of data
1051 * at once from a packet into the eBPF stack.
1053 * 0 on success, or a negative error in case of failure.
1055 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1057 * Walk a user or a kernel stack and return its id. To achieve
1058 * this, the helper needs *ctx*, which is a pointer to the context
1059 * on which the tracing program is executed, and a pointer to a
1060 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1062 * The last argument, *flags*, holds the number of stack frames to
1063 * skip (from 0 to 255), masked with
1064 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1065 * a combination of the following flags:
1067 * **BPF_F_USER_STACK**
1068 * Collect a user space stack instead of a kernel stack.
1069 * **BPF_F_FAST_STACK_CMP**
1070 * Compare stacks by hash only.
1071 * **BPF_F_REUSE_STACKID**
1072 * If two different stacks hash into the same *stackid*,
1073 * discard the old one.
1075 * The stack id retrieved is a 32 bit long integer handle which
1076 * can be further combined with other data (including other stack
1077 * ids) and used as a key into maps. This can be useful for
1078 * generating a variety of graphs (such as flame graphs or off-cpu
1081 * For walking a stack, this helper is an improvement over
1082 * **bpf_probe_read**\ (), which can be used with unrolled loops
1083 * but is not efficient and consumes a lot of eBPF instructions.
1084 * Instead, **bpf_get_stackid**\ () can collect up to
1085 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1086 * this limit can be controlled with the **sysctl** program, and
1087 * that it should be manually increased in order to profile long
1088 * user stacks (such as stacks for Java programs). To do so, use:
1092 * # sysctl kernel.perf_event_max_stack=<new value>
1094 * The positive or null stack id on success, or a negative error
1095 * in case of failure.
1097 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1099 * Compute a checksum difference, from the raw buffer pointed by
1100 * *from*, of length *from_size* (that must be a multiple of 4),
1101 * towards the raw buffer pointed by *to*, of size *to_size*
1102 * (same remark). An optional *seed* can be added to the value
1103 * (this can be cascaded, the seed may come from a previous call
1106 * This is flexible enough to be used in several ways:
1108 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1109 * checksum, it can be used when pushing new data.
1110 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1111 * checksum, it can be used when removing data from a packet.
1112 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1113 * can be used to compute a diff. Note that *from_size* and
1114 * *to_size* do not need to be equal.
1116 * This helper can be used in combination with
1117 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1118 * which one can feed in the difference computed with
1119 * **bpf_csum_diff**\ ().
1121 * The checksum result, or a negative error code in case of
1124 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1126 * Retrieve tunnel options metadata for the packet associated to
1127 * *skb*, and store the raw tunnel option data to the buffer *opt*
1130 * This helper can be used with encapsulation devices that can
1131 * operate in "collect metadata" mode (please refer to the related
1132 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1133 * more details). A particular example where this can be used is
1134 * in combination with the Geneve encapsulation protocol, where it
1135 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1136 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1137 * the eBPF program. This allows for full customization of these
1140 * The size of the option data retrieved.
1142 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1144 * Set tunnel options metadata for the packet associated to *skb*
1145 * to the option data contained in the raw buffer *opt* of *size*.
1147 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1148 * helper for additional information.
1150 * 0 on success, or a negative error in case of failure.
1152 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1154 * Change the protocol of the *skb* to *proto*. Currently
1155 * supported are transition from IPv4 to IPv6, and from IPv6 to
1156 * IPv4. The helper takes care of the groundwork for the
1157 * transition, including resizing the socket buffer. The eBPF
1158 * program is expected to fill the new headers, if any, via
1159 * **skb_store_bytes**\ () and to recompute the checksums with
1160 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1161 * (). The main case for this helper is to perform NAT64
1162 * operations out of an eBPF program.
1164 * Internally, the GSO type is marked as dodgy so that headers are
1165 * checked and segments are recalculated by the GSO/GRO engine.
1166 * The size for GSO target is adapted as well.
1168 * All values for *flags* are reserved for future usage, and must
1171 * A call to this helper is susceptible to change the underlying
1172 * packet buffer. Therefore, at load time, all checks on pointers
1173 * previously done by the verifier are invalidated and must be
1174 * performed again, if the helper is used in combination with
1175 * direct packet access.
1177 * 0 on success, or a negative error in case of failure.
1179 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1181 * Change the packet type for the packet associated to *skb*. This
1182 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1183 * the eBPF program does not have a write access to *skb*\
1184 * **->pkt_type** beside this helper. Using a helper here allows
1185 * for graceful handling of errors.
1187 * The major use case is to change incoming *skb*s to
1188 * **PACKET_HOST** in a programmatic way instead of having to
1189 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1192 * Note that *type* only allows certain values. At this time, they
1197 * **PACKET_BROADCAST**
1198 * Send packet to all.
1199 * **PACKET_MULTICAST**
1200 * Send packet to group.
1201 * **PACKET_OTHERHOST**
1202 * Send packet to someone else.
1204 * 0 on success, or a negative error in case of failure.
1206 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1208 * Check whether *skb* is a descendant of the cgroup2 held by
1209 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1211 * The return value depends on the result of the test, and can be:
1213 * * 0, if the *skb* failed the cgroup2 descendant test.
1214 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1215 * * A negative error code, if an error occurred.
1217 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1219 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1220 * not set, in particular if the hash was cleared due to mangling,
1221 * recompute this hash. Later accesses to the hash can be done
1222 * directly with *skb*\ **->hash**.
1224 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1225 * prototype with **bpf_skb_change_proto**\ (), or calling
1226 * **bpf_skb_store_bytes**\ () with the
1227 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1228 * the hash and to trigger a new computation for the next call to
1229 * **bpf_get_hash_recalc**\ ().
1233 * u64 bpf_get_current_task(void)
1235 * A pointer to the current task struct.
1237 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1239 * Attempt in a safe way to write *len* bytes from the buffer
1240 * *src* to *dst* in memory. It only works for threads that are in
1241 * user context, and *dst* must be a valid user space address.
1243 * This helper should not be used to implement any kind of
1244 * security mechanism because of TOC-TOU attacks, but rather to
1245 * debug, divert, and manipulate execution of semi-cooperative
1248 * Keep in mind that this feature is meant for experiments, and it
1249 * has a risk of crashing the system and running programs.
1250 * Therefore, when an eBPF program using this helper is attached,
1251 * a warning including PID and process name is printed to kernel
1254 * 0 on success, or a negative error in case of failure.
1256 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1258 * Check whether the probe is being run is the context of a given
1259 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1260 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1262 * The return value depends on the result of the test, and can be:
1264 * * 0, if the *skb* task belongs to the cgroup2.
1265 * * 1, if the *skb* task does not belong to the cgroup2.
1266 * * A negative error code, if an error occurred.
1268 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1270 * Resize (trim or grow) the packet associated to *skb* to the
1271 * new *len*. The *flags* are reserved for future usage, and must
1274 * The basic idea is that the helper performs the needed work to
1275 * change the size of the packet, then the eBPF program rewrites
1276 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1277 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1278 * and others. This helper is a slow path utility intended for
1279 * replies with control messages. And because it is targeted for
1280 * slow path, the helper itself can afford to be slow: it
1281 * implicitly linearizes, unclones and drops offloads from the
1284 * A call to this helper is susceptible to change the underlying
1285 * packet buffer. Therefore, at load time, all checks on pointers
1286 * previously done by the verifier are invalidated and must be
1287 * performed again, if the helper is used in combination with
1288 * direct packet access.
1290 * 0 on success, or a negative error in case of failure.
1292 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1294 * Pull in non-linear data in case the *skb* is non-linear and not
1295 * all of *len* are part of the linear section. Make *len* bytes
1296 * from *skb* readable and writable. If a zero value is passed for
1297 * *len*, then the whole length of the *skb* is pulled.
1299 * This helper is only needed for reading and writing with direct
1302 * For direct packet access, testing that offsets to access
1303 * are within packet boundaries (test on *skb*\ **->data_end**) is
1304 * susceptible to fail if offsets are invalid, or if the requested
1305 * data is in non-linear parts of the *skb*. On failure the
1306 * program can just bail out, or in the case of a non-linear
1307 * buffer, use a helper to make the data available. The
1308 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1309 * the data. Another one consists in using **bpf_skb_pull_data**
1310 * to pull in once the non-linear parts, then retesting and
1311 * eventually access the data.
1313 * At the same time, this also makes sure the *skb* is uncloned,
1314 * which is a necessary condition for direct write. As this needs
1315 * to be an invariant for the write part only, the verifier
1316 * detects writes and adds a prologue that is calling
1317 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1318 * the very beginning in case it is indeed cloned.
1320 * A call to this helper is susceptible to change the underlying
1321 * packet buffer. Therefore, at load time, all checks on pointers
1322 * previously done by the verifier are invalidated and must be
1323 * performed again, if the helper is used in combination with
1324 * direct packet access.
1326 * 0 on success, or a negative error in case of failure.
1328 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1330 * Add the checksum *csum* into *skb*\ **->csum** in case the
1331 * driver has supplied a checksum for the entire packet into that
1332 * field. Return an error otherwise. This helper is intended to be
1333 * used in combination with **bpf_csum_diff**\ (), in particular
1334 * when the checksum needs to be updated after data has been
1335 * written into the packet through direct packet access.
1337 * The checksum on success, or a negative error code in case of
1340 * void bpf_set_hash_invalid(struct sk_buff *skb)
1342 * Invalidate the current *skb*\ **->hash**. It can be used after
1343 * mangling on headers through direct packet access, in order to
1344 * indicate that the hash is outdated and to trigger a
1345 * recalculation the next time the kernel tries to access this
1346 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1348 * int bpf_get_numa_node_id(void)
1350 * Return the id of the current NUMA node. The primary use case
1351 * for this helper is the selection of sockets for the local NUMA
1352 * node, when the program is attached to sockets using the
1353 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1354 * but the helper is also available to other eBPF program types,
1355 * similarly to **bpf_get_smp_processor_id**\ ().
1357 * The id of current NUMA node.
1359 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1361 * Grows headroom of packet associated to *skb* and adjusts the
1362 * offset of the MAC header accordingly, adding *len* bytes of
1363 * space. It automatically extends and reallocates memory as
1366 * This helper can be used on a layer 3 *skb* to push a MAC header
1367 * for redirection into a layer 2 device.
1369 * All values for *flags* are reserved for future usage, and must
1372 * A call to this helper is susceptible to change the underlying
1373 * packet buffer. Therefore, at load time, all checks on pointers
1374 * previously done by the verifier are invalidated and must be
1375 * performed again, if the helper is used in combination with
1376 * direct packet access.
1378 * 0 on success, or a negative error in case of failure.
1380 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1382 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1383 * it is possible to use a negative value for *delta*. This helper
1384 * can be used to prepare the packet for pushing or popping
1387 * A call to this helper is susceptible to change the underlying
1388 * packet buffer. Therefore, at load time, all checks on pointers
1389 * previously done by the verifier are invalidated and must be
1390 * performed again, if the helper is used in combination with
1391 * direct packet access.
1393 * 0 on success, or a negative error in case of failure.
1395 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1397 * Copy a NUL terminated string from an unsafe address
1398 * *unsafe_ptr* to *dst*. The *size* should include the
1399 * terminating NUL byte. In case the string length is smaller than
1400 * *size*, the target is not padded with further NUL bytes. If the
1401 * string length is larger than *size*, just *size*-1 bytes are
1402 * copied and the last byte is set to NUL.
1404 * On success, the length of the copied string is returned. This
1405 * makes this helper useful in tracing programs for reading
1406 * strings, and more importantly to get its length at runtime. See
1407 * the following snippet:
1411 * SEC("kprobe/sys_open")
1412 * void bpf_sys_open(struct pt_regs *ctx)
1414 * char buf[PATHLEN]; // PATHLEN is defined to 256
1415 * int res = bpf_probe_read_str(buf, sizeof(buf),
1418 * // Consume buf, for example push it to
1419 * // userspace via bpf_perf_event_output(); we
1420 * // can use res (the string length) as event
1421 * // size, after checking its boundaries.
1424 * In comparison, using **bpf_probe_read()** helper here instead
1425 * to read the string would require to estimate the length at
1426 * compile time, and would often result in copying more memory
1429 * Another useful use case is when parsing individual process
1430 * arguments or individual environment variables navigating
1431 * *current*\ **->mm->arg_start** and *current*\
1432 * **->mm->env_start**: using this helper and the return value,
1433 * one can quickly iterate at the right offset of the memory area.
1435 * On success, the strictly positive length of the string,
1436 * including the trailing NUL character. On error, a negative
1439 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1441 * If the **struct sk_buff** pointed by *skb* has a known socket,
1442 * retrieve the cookie (generated by the kernel) of this socket.
1443 * If no cookie has been set yet, generate a new cookie. Once
1444 * generated, the socket cookie remains stable for the life of the
1445 * socket. This helper can be useful for monitoring per socket
1446 * networking traffic statistics as it provides a unique socket
1447 * identifier per namespace.
1449 * A 8-byte long non-decreasing number on success, or 0 if the
1450 * socket field is missing inside *skb*.
1452 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1454 * Equivalent to bpf_get_socket_cookie() helper that accepts
1455 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1457 * A 8-byte long non-decreasing number.
1459 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1461 * Equivalent to bpf_get_socket_cookie() helper that accepts
1462 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1464 * A 8-byte long non-decreasing number.
1466 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1468 * The owner UID of the socket associated to *skb*. If the socket
1469 * is **NULL**, or if it is not a full socket (i.e. if it is a
1470 * time-wait or a request socket instead), **overflowuid** value
1471 * is returned (note that **overflowuid** might also be the actual
1472 * UID value for the socket).
1474 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1476 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1481 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1483 * Emulate a call to **setsockopt()** on the socket associated to
1484 * *bpf_socket*, which must be a full socket. The *level* at
1485 * which the option resides and the name *optname* of the option
1486 * must be specified, see **setsockopt(2)** for more information.
1487 * The option value of length *optlen* is pointed by *optval*.
1489 * This helper actually implements a subset of **setsockopt()**.
1490 * It supports the following *level*\ s:
1492 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1493 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1494 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1495 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1496 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1497 * **TCP_BPF_SNDCWND_CLAMP**.
1498 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1499 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1501 * 0 on success, or a negative error in case of failure.
1503 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1505 * Grow or shrink the room for data in the packet associated to
1506 * *skb* by *len_diff*, and according to the selected *mode*.
1508 * There are two supported modes at this time:
1510 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1511 * (room space is added or removed below the layer 2 header).
1513 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1514 * (room space is added or removed below the layer 3 header).
1516 * The following flags are supported at this time:
1518 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1519 * Adjusting mss in this way is not allowed for datagrams.
1521 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1522 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1523 * Any new space is reserved to hold a tunnel header.
1524 * Configure skb offsets and other fields accordingly.
1526 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1527 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1528 * Use with ENCAP_L3 flags to further specify the tunnel type.
1530 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1531 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1532 * type; *len* is the length of the inner MAC header.
1534 * A call to this helper is susceptible to change the underlying
1535 * packet buffer. Therefore, at load time, all checks on pointers
1536 * previously done by the verifier are invalidated and must be
1537 * performed again, if the helper is used in combination with
1538 * direct packet access.
1540 * 0 on success, or a negative error in case of failure.
1542 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1544 * Redirect the packet to the endpoint referenced by *map* at
1545 * index *key*. Depending on its type, this *map* can contain
1546 * references to net devices (for forwarding packets through other
1547 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1548 * but this is only implemented for native XDP (with driver
1549 * support) as of this writing).
1551 * All values for *flags* are reserved for future usage, and must
1554 * When used to redirect packets to net devices, this helper
1555 * provides a high performance increase over **bpf_redirect**\ ().
1556 * This is due to various implementation details of the underlying
1557 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1558 * () tries to send packet as a "bulk" to the device.
1560 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1562 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1564 * Redirect the packet to the socket referenced by *map* (of type
1565 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1566 * egress interfaces can be used for redirection. The
1567 * **BPF_F_INGRESS** value in *flags* is used to make the
1568 * distinction (ingress path is selected if the flag is present,
1569 * egress path otherwise). This is the only flag supported for now.
1571 * **SK_PASS** on success, or **SK_DROP** on error.
1573 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1575 * Add an entry to, or update a *map* referencing sockets. The
1576 * *skops* is used as a new value for the entry associated to
1577 * *key*. *flags* is one of:
1580 * The entry for *key* must not exist in the map.
1582 * The entry for *key* must already exist in the map.
1584 * No condition on the existence of the entry for *key*.
1586 * If the *map* has eBPF programs (parser and verdict), those will
1587 * be inherited by the socket being added. If the socket is
1588 * already attached to eBPF programs, this results in an error.
1590 * 0 on success, or a negative error in case of failure.
1592 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1594 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1595 * *delta* (which can be positive or negative). Note that this
1596 * operation modifies the address stored in *xdp_md*\ **->data**,
1597 * so the latter must be loaded only after the helper has been
1600 * The use of *xdp_md*\ **->data_meta** is optional and programs
1601 * are not required to use it. The rationale is that when the
1602 * packet is processed with XDP (e.g. as DoS filter), it is
1603 * possible to push further meta data along with it before passing
1604 * to the stack, and to give the guarantee that an ingress eBPF
1605 * program attached as a TC classifier on the same device can pick
1606 * this up for further post-processing. Since TC works with socket
1607 * buffers, it remains possible to set from XDP the **mark** or
1608 * **priority** pointers, or other pointers for the socket buffer.
1609 * Having this scratch space generic and programmable allows for
1610 * more flexibility as the user is free to store whatever meta
1613 * A call to this helper is susceptible to change the underlying
1614 * packet buffer. Therefore, at load time, all checks on pointers
1615 * previously done by the verifier are invalidated and must be
1616 * performed again, if the helper is used in combination with
1617 * direct packet access.
1619 * 0 on success, or a negative error in case of failure.
1621 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1623 * Read the value of a perf event counter, and store it into *buf*
1624 * of size *buf_size*. This helper relies on a *map* of type
1625 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1626 * counter is selected when *map* is updated with perf event file
1627 * descriptors. The *map* is an array whose size is the number of
1628 * available CPUs, and each cell contains a value relative to one
1629 * CPU. The value to retrieve is indicated by *flags*, that
1630 * contains the index of the CPU to look up, masked with
1631 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1632 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1633 * current CPU should be retrieved.
1635 * This helper behaves in a way close to
1636 * **bpf_perf_event_read**\ () helper, save that instead of
1637 * just returning the value observed, it fills the *buf*
1638 * structure. This allows for additional data to be retrieved: in
1639 * particular, the enabled and running times (in *buf*\
1640 * **->enabled** and *buf*\ **->running**, respectively) are
1641 * copied. In general, **bpf_perf_event_read_value**\ () is
1642 * recommended over **bpf_perf_event_read**\ (), which has some
1643 * ABI issues and provides fewer functionalities.
1645 * These values are interesting, because hardware PMU (Performance
1646 * Monitoring Unit) counters are limited resources. When there are
1647 * more PMU based perf events opened than available counters,
1648 * kernel will multiplex these events so each event gets certain
1649 * percentage (but not all) of the PMU time. In case that
1650 * multiplexing happens, the number of samples or counter value
1651 * will not reflect the case compared to when no multiplexing
1652 * occurs. This makes comparison between different runs difficult.
1653 * Typically, the counter value should be normalized before
1654 * comparing to other experiments. The usual normalization is done
1659 * normalized_counter = counter * t_enabled / t_running
1661 * Where t_enabled is the time enabled for event and t_running is
1662 * the time running for event since last normalization. The
1663 * enabled and running times are accumulated since the perf event
1664 * open. To achieve scaling factor between two invocations of an
1665 * eBPF program, users can can use CPU id as the key (which is
1666 * typical for perf array usage model) to remember the previous
1667 * value and do the calculation inside the eBPF program.
1669 * 0 on success, or a negative error in case of failure.
1671 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1673 * For en eBPF program attached to a perf event, retrieve the
1674 * value of the event counter associated to *ctx* and store it in
1675 * the structure pointed by *buf* and of size *buf_size*. Enabled
1676 * and running times are also stored in the structure (see
1677 * description of helper **bpf_perf_event_read_value**\ () for
1680 * 0 on success, or a negative error in case of failure.
1682 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1684 * Emulate a call to **getsockopt()** on the socket associated to
1685 * *bpf_socket*, which must be a full socket. The *level* at
1686 * which the option resides and the name *optname* of the option
1687 * must be specified, see **getsockopt(2)** for more information.
1688 * The retrieved value is stored in the structure pointed by
1689 * *opval* and of length *optlen*.
1691 * This helper actually implements a subset of **getsockopt()**.
1692 * It supports the following *level*\ s:
1694 * * **IPPROTO_TCP**, which supports *optname*
1695 * **TCP_CONGESTION**.
1696 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1697 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1699 * 0 on success, or a negative error in case of failure.
1701 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1703 * Used for error injection, this helper uses kprobes to override
1704 * the return value of the probed function, and to set it to *rc*.
1705 * The first argument is the context *regs* on which the kprobe
1708 * This helper works by setting setting the PC (program counter)
1709 * to an override function which is run in place of the original
1710 * probed function. This means the probed function is not run at
1711 * all. The replacement function just returns with the required
1714 * This helper has security implications, and thus is subject to
1715 * restrictions. It is only available if the kernel was compiled
1716 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1717 * option, and in this case it only works on functions tagged with
1718 * **ALLOW_ERROR_INJECTION** in the kernel code.
1720 * Also, the helper is only available for the architectures having
1721 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1722 * x86 architecture is the only one to support this feature.
1726 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1728 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1729 * for the full TCP socket associated to *bpf_sock_ops* to
1732 * The primary use of this field is to determine if there should
1733 * be calls to eBPF programs of type
1734 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1735 * code. A program of the same type can change its value, per
1736 * connection and as necessary, when the connection is
1737 * established. This field is directly accessible for reading, but
1738 * this helper must be used for updates in order to return an
1739 * error if an eBPF program tries to set a callback that is not
1740 * supported in the current kernel.
1742 * *argval* is a flag array which can combine these flags:
1744 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1745 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1746 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1748 * Therefore, this function can be used to clear a callback flag by
1749 * setting the appropriate bit to zero. e.g. to disable the RTO
1752 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1753 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1755 * Here are some examples of where one could call such eBPF
1759 * * When a packet is retransmitted.
1760 * * When the connection terminates.
1761 * * When a packet is sent.
1762 * * When a packet is received.
1764 * Code **-EINVAL** if the socket is not a full TCP socket;
1765 * otherwise, a positive number containing the bits that could not
1766 * be set is returned (which comes down to 0 if all bits were set
1769 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1771 * This helper is used in programs implementing policies at the
1772 * socket level. If the message *msg* is allowed to pass (i.e. if
1773 * the verdict eBPF program returns **SK_PASS**), redirect it to
1774 * the socket referenced by *map* (of type
1775 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1776 * egress interfaces can be used for redirection. The
1777 * **BPF_F_INGRESS** value in *flags* is used to make the
1778 * distinction (ingress path is selected if the flag is present,
1779 * egress path otherwise). This is the only flag supported for now.
1781 * **SK_PASS** on success, or **SK_DROP** on error.
1783 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1785 * For socket policies, apply the verdict of the eBPF program to
1786 * the next *bytes* (number of bytes) of message *msg*.
1788 * For example, this helper can be used in the following cases:
1790 * * A single **sendmsg**\ () or **sendfile**\ () system call
1791 * contains multiple logical messages that the eBPF program is
1792 * supposed to read and for which it should apply a verdict.
1793 * * An eBPF program only cares to read the first *bytes* of a
1794 * *msg*. If the message has a large payload, then setting up
1795 * and calling the eBPF program repeatedly for all bytes, even
1796 * though the verdict is already known, would create unnecessary
1799 * When called from within an eBPF program, the helper sets a
1800 * counter internal to the BPF infrastructure, that is used to
1801 * apply the last verdict to the next *bytes*. If *bytes* is
1802 * smaller than the current data being processed from a
1803 * **sendmsg**\ () or **sendfile**\ () system call, the first
1804 * *bytes* will be sent and the eBPF program will be re-run with
1805 * the pointer for start of data pointing to byte number *bytes*
1806 * **+ 1**. If *bytes* is larger than the current data being
1807 * processed, then the eBPF verdict will be applied to multiple
1808 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1811 * Note that if a socket closes with the internal counter holding
1812 * a non-zero value, this is not a problem because data is not
1813 * being buffered for *bytes* and is sent as it is received.
1817 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1819 * For socket policies, prevent the execution of the verdict eBPF
1820 * program for message *msg* until *bytes* (byte number) have been
1823 * This can be used when one needs a specific number of bytes
1824 * before a verdict can be assigned, even if the data spans
1825 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1826 * case would be a user calling **sendmsg**\ () repeatedly with
1827 * 1-byte long message segments. Obviously, this is bad for
1828 * performance, but it is still valid. If the eBPF program needs
1829 * *bytes* bytes to validate a header, this helper can be used to
1830 * prevent the eBPF program to be called again until *bytes* have
1835 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1837 * For socket policies, pull in non-linear data from user space
1838 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1839 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1842 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1843 * *msg* it can only parse data that the (**data**, **data_end**)
1844 * pointers have already consumed. For **sendmsg**\ () hooks this
1845 * is likely the first scatterlist element. But for calls relying
1846 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1847 * be the range (**0**, **0**) because the data is shared with
1848 * user space and by default the objective is to avoid allowing
1849 * user space to modify data while (or after) eBPF verdict is
1850 * being decided. This helper can be used to pull in data and to
1851 * set the start and end pointer to given values. Data will be
1852 * copied if necessary (i.e. if data was not linear and if start
1853 * and end pointers do not point to the same chunk).
1855 * A call to this helper is susceptible to change the underlying
1856 * packet buffer. Therefore, at load time, all checks on pointers
1857 * previously done by the verifier are invalidated and must be
1858 * performed again, if the helper is used in combination with
1859 * direct packet access.
1861 * All values for *flags* are reserved for future usage, and must
1864 * 0 on success, or a negative error in case of failure.
1866 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1868 * Bind the socket associated to *ctx* to the address pointed by
1869 * *addr*, of length *addr_len*. This allows for making outgoing
1870 * connection from the desired IP address, which can be useful for
1871 * example when all processes inside a cgroup should use one
1872 * single IP address on a host that has multiple IP configured.
1874 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1875 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1876 * **AF_INET6**). Looking for a free port to bind to can be
1877 * expensive, therefore binding to port is not permitted by the
1878 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1879 * must be set to zero.
1881 * 0 on success, or a negative error in case of failure.
1883 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1885 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1886 * only possible to shrink the packet as of this writing,
1887 * therefore *delta* must be a negative integer.
1889 * A call to this helper is susceptible to change the underlying
1890 * packet buffer. Therefore, at load time, all checks on pointers
1891 * previously done by the verifier are invalidated and must be
1892 * performed again, if the helper is used in combination with
1893 * direct packet access.
1895 * 0 on success, or a negative error in case of failure.
1897 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1899 * Retrieve the XFRM state (IP transform framework, see also
1900 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1902 * The retrieved value is stored in the **struct bpf_xfrm_state**
1903 * pointed by *xfrm_state* and of length *size*.
1905 * All values for *flags* are reserved for future usage, and must
1908 * This helper is available only if the kernel was compiled with
1909 * **CONFIG_XFRM** configuration option.
1911 * 0 on success, or a negative error in case of failure.
1913 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1915 * Return a user or a kernel stack in bpf program provided buffer.
1916 * To achieve this, the helper needs *ctx*, which is a pointer
1917 * to the context on which the tracing program is executed.
1918 * To store the stacktrace, the bpf program provides *buf* with
1919 * a nonnegative *size*.
1921 * The last argument, *flags*, holds the number of stack frames to
1922 * skip (from 0 to 255), masked with
1923 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1924 * the following flags:
1926 * **BPF_F_USER_STACK**
1927 * Collect a user space stack instead of a kernel stack.
1928 * **BPF_F_USER_BUILD_ID**
1929 * Collect buildid+offset instead of ips for user stack,
1930 * only valid if **BPF_F_USER_STACK** is also specified.
1932 * **bpf_get_stack**\ () can collect up to
1933 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1934 * to sufficient large buffer size. Note that
1935 * this limit can be controlled with the **sysctl** program, and
1936 * that it should be manually increased in order to profile long
1937 * user stacks (such as stacks for Java programs). To do so, use:
1941 * # sysctl kernel.perf_event_max_stack=<new value>
1943 * A non-negative value equal to or less than *size* on success,
1944 * or a negative error in case of failure.
1946 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1948 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1949 * it provides an easy way to load *len* bytes from *offset*
1950 * from the packet associated to *skb*, into the buffer pointed
1951 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1952 * a fifth argument *start_header* exists in order to select a
1953 * base offset to start from. *start_header* can be one of:
1955 * **BPF_HDR_START_MAC**
1956 * Base offset to load data from is *skb*'s mac header.
1957 * **BPF_HDR_START_NET**
1958 * Base offset to load data from is *skb*'s network header.
1960 * In general, "direct packet access" is the preferred method to
1961 * access packet data, however, this helper is in particular useful
1962 * in socket filters where *skb*\ **->data** does not always point
1963 * to the start of the mac header and where "direct packet access"
1966 * 0 on success, or a negative error in case of failure.
1968 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1970 * Do FIB lookup in kernel tables using parameters in *params*.
1971 * If lookup is successful and result shows packet is to be
1972 * forwarded, the neighbor tables are searched for the nexthop.
1973 * If successful (ie., FIB lookup shows forwarding and nexthop
1974 * is resolved), the nexthop address is returned in ipv4_dst
1975 * or ipv6_dst based on family, smac is set to mac address of
1976 * egress device, dmac is set to nexthop mac address, rt_metric
1977 * is set to metric from route (IPv4/IPv6 only), and ifindex
1978 * is set to the device index of the nexthop from the FIB lookup.
1980 * *plen* argument is the size of the passed in struct.
1981 * *flags* argument can be a combination of one or more of the
1984 * **BPF_FIB_LOOKUP_DIRECT**
1985 * Do a direct table lookup vs full lookup using FIB
1987 * **BPF_FIB_LOOKUP_OUTPUT**
1988 * Perform lookup from an egress perspective (default is
1991 * *ctx* is either **struct xdp_md** for XDP programs or
1992 * **struct sk_buff** tc cls_act programs.
1994 * * < 0 if any input argument is invalid
1995 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1996 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1997 * packet is not forwarded or needs assist from full stack
1999 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
2001 * Add an entry to, or update a sockhash *map* referencing sockets.
2002 * The *skops* is used as a new value for the entry associated to
2003 * *key*. *flags* is one of:
2006 * The entry for *key* must not exist in the map.
2008 * The entry for *key* must already exist in the map.
2010 * No condition on the existence of the entry for *key*.
2012 * If the *map* has eBPF programs (parser and verdict), those will
2013 * be inherited by the socket being added. If the socket is
2014 * already attached to eBPF programs, this results in an error.
2016 * 0 on success, or a negative error in case of failure.
2018 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2020 * This helper is used in programs implementing policies at the
2021 * socket level. If the message *msg* is allowed to pass (i.e. if
2022 * the verdict eBPF program returns **SK_PASS**), redirect it to
2023 * the socket referenced by *map* (of type
2024 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2025 * egress interfaces can be used for redirection. The
2026 * **BPF_F_INGRESS** value in *flags* is used to make the
2027 * distinction (ingress path is selected if the flag is present,
2028 * egress path otherwise). This is the only flag supported for now.
2030 * **SK_PASS** on success, or **SK_DROP** on error.
2032 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2034 * This helper is used in programs implementing policies at the
2035 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2036 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2037 * to the socket referenced by *map* (of type
2038 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2039 * egress interfaces can be used for redirection. The
2040 * **BPF_F_INGRESS** value in *flags* is used to make the
2041 * distinction (ingress path is selected if the flag is present,
2042 * egress otherwise). This is the only flag supported for now.
2044 * **SK_PASS** on success, or **SK_DROP** on error.
2046 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2048 * Encapsulate the packet associated to *skb* within a Layer 3
2049 * protocol header. This header is provided in the buffer at
2050 * address *hdr*, with *len* its size in bytes. *type* indicates
2051 * the protocol of the header and can be one of:
2053 * **BPF_LWT_ENCAP_SEG6**
2054 * IPv6 encapsulation with Segment Routing Header
2055 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2056 * the IPv6 header is computed by the kernel.
2057 * **BPF_LWT_ENCAP_SEG6_INLINE**
2058 * Only works if *skb* contains an IPv6 packet. Insert a
2059 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2061 * **BPF_LWT_ENCAP_IP**
2062 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2063 * must be IPv4 or IPv6, followed by zero or more
2064 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2065 * total bytes in all prepended headers. Please note that
2066 * if **skb_is_gso**\ (*skb*) is true, no more than two
2067 * headers can be prepended, and the inner header, if
2068 * present, should be either GRE or UDP/GUE.
2070 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2071 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2072 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2073 * **BPF_PROG_TYPE_LWT_XMIT**.
2075 * A call to this helper is susceptible to change the underlying
2076 * packet buffer. Therefore, at load time, all checks on pointers
2077 * previously done by the verifier are invalidated and must be
2078 * performed again, if the helper is used in combination with
2079 * direct packet access.
2081 * 0 on success, or a negative error in case of failure.
2083 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2085 * Store *len* bytes from address *from* into the packet
2086 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2087 * inside the outermost IPv6 Segment Routing Header can be
2088 * modified through this helper.
2090 * A call to this helper is susceptible to change the underlying
2091 * packet buffer. Therefore, at load time, all checks on pointers
2092 * previously done by the verifier are invalidated and must be
2093 * performed again, if the helper is used in combination with
2094 * direct packet access.
2096 * 0 on success, or a negative error in case of failure.
2098 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2100 * Adjust the size allocated to TLVs in the outermost IPv6
2101 * Segment Routing Header contained in the packet associated to
2102 * *skb*, at position *offset* by *delta* bytes. Only offsets
2103 * after the segments are accepted. *delta* can be as well
2104 * positive (growing) as negative (shrinking).
2106 * A call to this helper is susceptible to change the underlying
2107 * packet buffer. Therefore, at load time, all checks on pointers
2108 * previously done by the verifier are invalidated and must be
2109 * performed again, if the helper is used in combination with
2110 * direct packet access.
2112 * 0 on success, or a negative error in case of failure.
2114 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2116 * Apply an IPv6 Segment Routing action of type *action* to the
2117 * packet associated to *skb*. Each action takes a parameter
2118 * contained at address *param*, and of length *param_len* bytes.
2119 * *action* can be one of:
2121 * **SEG6_LOCAL_ACTION_END_X**
2122 * End.X action: Endpoint with Layer-3 cross-connect.
2123 * Type of *param*: **struct in6_addr**.
2124 * **SEG6_LOCAL_ACTION_END_T**
2125 * End.T action: Endpoint with specific IPv6 table lookup.
2126 * Type of *param*: **int**.
2127 * **SEG6_LOCAL_ACTION_END_B6**
2128 * End.B6 action: Endpoint bound to an SRv6 policy.
2129 * Type of *param*: **struct ipv6_sr_hdr**.
2130 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2131 * End.B6.Encap action: Endpoint bound to an SRv6
2132 * encapsulation policy.
2133 * Type of *param*: **struct ipv6_sr_hdr**.
2135 * A call to this helper is susceptible to change the underlying
2136 * packet buffer. Therefore, at load time, all checks on pointers
2137 * previously done by the verifier are invalidated and must be
2138 * performed again, if the helper is used in combination with
2139 * direct packet access.
2141 * 0 on success, or a negative error in case of failure.
2143 * int bpf_rc_repeat(void *ctx)
2145 * This helper is used in programs implementing IR decoding, to
2146 * report a successfully decoded repeat key message. This delays
2147 * the generation of a key up event for previously generated
2150 * Some IR protocols like NEC have a special IR message for
2151 * repeating last button, for when a button is held down.
2153 * The *ctx* should point to the lirc sample as passed into
2156 * This helper is only available is the kernel was compiled with
2157 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2162 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2164 * This helper is used in programs implementing IR decoding, to
2165 * report a successfully decoded key press with *scancode*,
2166 * *toggle* value in the given *protocol*. The scancode will be
2167 * translated to a keycode using the rc keymap, and reported as
2168 * an input key down event. After a period a key up event is
2169 * generated. This period can be extended by calling either
2170 * **bpf_rc_keydown**\ () again with the same values, or calling
2171 * **bpf_rc_repeat**\ ().
2173 * Some protocols include a toggle bit, in case the button was
2174 * released and pressed again between consecutive scancodes.
2176 * The *ctx* should point to the lirc sample as passed into
2179 * The *protocol* is the decoded protocol number (see
2180 * **enum rc_proto** for some predefined values).
2182 * This helper is only available is the kernel was compiled with
2183 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2188 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2190 * Return the cgroup v2 id of the socket associated with the *skb*.
2191 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2192 * helper for cgroup v1 by providing a tag resp. identifier that
2193 * can be matched on or used for map lookups e.g. to implement
2194 * policy. The cgroup v2 id of a given path in the hierarchy is
2195 * exposed in user space through the f_handle API in order to get
2196 * to the same 64-bit id.
2198 * This helper can be used on TC egress path, but not on ingress,
2199 * and is available only if the kernel was compiled with the
2200 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2202 * The id is returned or 0 in case the id could not be retrieved.
2204 * u64 bpf_get_current_cgroup_id(void)
2206 * A 64-bit integer containing the current cgroup id based
2207 * on the cgroup within which the current task is running.
2209 * void *bpf_get_local_storage(void *map, u64 flags)
2211 * Get the pointer to the local storage area.
2212 * The type and the size of the local storage is defined
2213 * by the *map* argument.
2214 * The *flags* meaning is specific for each map type,
2215 * and has to be 0 for cgroup local storage.
2217 * Depending on the BPF program type, a local storage area
2218 * can be shared between multiple instances of the BPF program,
2219 * running simultaneously.
2221 * A user should care about the synchronization by himself.
2222 * For example, by using the **BPF_STX_XADD** instruction to alter
2225 * A pointer to the local storage area.
2227 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2229 * Select a **SO_REUSEPORT** socket from a
2230 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2231 * It checks the selected socket is matching the incoming
2232 * request in the socket buffer.
2234 * 0 on success, or a negative error in case of failure.
2236 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2238 * Return id of cgroup v2 that is ancestor of cgroup associated
2239 * with the *skb* at the *ancestor_level*. The root cgroup is at
2240 * *ancestor_level* zero and each step down the hierarchy
2241 * increments the level. If *ancestor_level* == level of cgroup
2242 * associated with *skb*, then return value will be same as that
2243 * of **bpf_skb_cgroup_id**\ ().
2245 * The helper is useful to implement policies based on cgroups
2246 * that are upper in hierarchy than immediate cgroup associated
2249 * The format of returned id and helper limitations are same as in
2250 * **bpf_skb_cgroup_id**\ ().
2252 * The id is returned or 0 in case the id could not be retrieved.
2254 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2256 * Look for TCP socket matching *tuple*, optionally in a child
2257 * network namespace *netns*. The return value must be checked,
2258 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2260 * The *ctx* should point to the context of the program, such as
2261 * the skb or socket (depending on the hook in use). This is used
2262 * to determine the base network namespace for the lookup.
2264 * *tuple_size* must be one of:
2266 * **sizeof**\ (*tuple*\ **->ipv4**)
2267 * Look for an IPv4 socket.
2268 * **sizeof**\ (*tuple*\ **->ipv6**)
2269 * Look for an IPv6 socket.
2271 * If the *netns* is a negative signed 32-bit integer, then the
2272 * socket lookup table in the netns associated with the *ctx* will
2273 * will be used. For the TC hooks, this is the netns of the device
2274 * in the skb. For socket hooks, this is the netns of the socket.
2275 * If *netns* is any other signed 32-bit value greater than or
2276 * equal to zero then it specifies the ID of the netns relative to
2277 * the netns associated with the *ctx*. *netns* values beyond the
2278 * range of 32-bit integers are reserved for future use.
2280 * All values for *flags* are reserved for future usage, and must
2283 * This helper is available only if the kernel was compiled with
2284 * **CONFIG_NET** configuration option.
2286 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2287 * For sockets with reuseport option, the **struct bpf_sock**
2288 * result is from *reuse*\ **->socks**\ [] using the hash of the
2291 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2293 * Look for UDP socket matching *tuple*, optionally in a child
2294 * network namespace *netns*. The return value must be checked,
2295 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2297 * The *ctx* should point to the context of the program, such as
2298 * the skb or socket (depending on the hook in use). This is used
2299 * to determine the base network namespace for the lookup.
2301 * *tuple_size* must be one of:
2303 * **sizeof**\ (*tuple*\ **->ipv4**)
2304 * Look for an IPv4 socket.
2305 * **sizeof**\ (*tuple*\ **->ipv6**)
2306 * Look for an IPv6 socket.
2308 * If the *netns* is a negative signed 32-bit integer, then the
2309 * socket lookup table in the netns associated with the *ctx* will
2310 * will be used. For the TC hooks, this is the netns of the device
2311 * in the skb. For socket hooks, this is the netns of the socket.
2312 * If *netns* is any other signed 32-bit value greater than or
2313 * equal to zero then it specifies the ID of the netns relative to
2314 * the netns associated with the *ctx*. *netns* values beyond the
2315 * range of 32-bit integers are reserved for future use.
2317 * All values for *flags* are reserved for future usage, and must
2320 * This helper is available only if the kernel was compiled with
2321 * **CONFIG_NET** configuration option.
2323 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2324 * For sockets with reuseport option, the **struct bpf_sock**
2325 * result is from *reuse*\ **->socks**\ [] using the hash of the
2328 * int bpf_sk_release(struct bpf_sock *sock)
2330 * Release the reference held by *sock*. *sock* must be a
2331 * non-**NULL** pointer that was returned from
2332 * **bpf_sk_lookup_xxx**\ ().
2334 * 0 on success, or a negative error in case of failure.
2336 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2338 * Push an element *value* in *map*. *flags* is one of:
2341 * If the queue/stack is full, the oldest element is
2342 * removed to make room for this.
2344 * 0 on success, or a negative error in case of failure.
2346 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2348 * Pop an element from *map*.
2350 * 0 on success, or a negative error in case of failure.
2352 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2354 * Get an element from *map* without removing it.
2356 * 0 on success, or a negative error in case of failure.
2358 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2360 * For socket policies, insert *len* bytes into *msg* at offset
2363 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2364 * *msg* it may want to insert metadata or options into the *msg*.
2365 * This can later be read and used by any of the lower layer BPF
2368 * This helper may fail if under memory pressure (a malloc
2369 * fails) in these cases BPF programs will get an appropriate
2370 * error and BPF programs will need to handle them.
2372 * 0 on success, or a negative error in case of failure.
2374 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2376 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2377 * This may result in **ENOMEM** errors under certain situations if
2378 * an allocation and copy are required due to a full ring buffer.
2379 * However, the helper will try to avoid doing the allocation
2380 * if possible. Other errors can occur if input parameters are
2381 * invalid either due to *start* byte not being valid part of *msg*
2382 * payload and/or *pop* value being to large.
2384 * 0 on success, or a negative error in case of failure.
2386 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2388 * This helper is used in programs implementing IR decoding, to
2389 * report a successfully decoded pointer movement.
2391 * The *ctx* should point to the lirc sample as passed into
2394 * This helper is only available is the kernel was compiled with
2395 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2400 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2402 * Acquire a spinlock represented by the pointer *lock*, which is
2403 * stored as part of a value of a map. Taking the lock allows to
2404 * safely update the rest of the fields in that value. The
2405 * spinlock can (and must) later be released with a call to
2406 * **bpf_spin_unlock**\ (\ *lock*\ ).
2408 * Spinlocks in BPF programs come with a number of restrictions
2411 * * **bpf_spin_lock** objects are only allowed inside maps of
2412 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2413 * list could be extended in the future).
2414 * * BTF description of the map is mandatory.
2415 * * The BPF program can take ONE lock at a time, since taking two
2416 * or more could cause dead locks.
2417 * * Only one **struct bpf_spin_lock** is allowed per map element.
2418 * * When the lock is taken, calls (either BPF to BPF or helpers)
2420 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2421 * allowed inside a spinlock-ed region.
2422 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2423 * the lock, on all execution paths, before it returns.
2424 * * The BPF program can access **struct bpf_spin_lock** only via
2425 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2426 * helpers. Loading or storing data into the **struct
2427 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2428 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2429 * of the map value must be a struct and have **struct
2430 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2431 * Nested lock inside another struct is not allowed.
2432 * * The **struct bpf_spin_lock** *lock* field in a map value must
2433 * be aligned on a multiple of 4 bytes in that value.
2434 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2435 * the **bpf_spin_lock** field to user space.
2436 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2437 * a BPF program, do not update the **bpf_spin_lock** field.
2438 * * **bpf_spin_lock** cannot be on the stack or inside a
2439 * networking packet (it can only be inside of a map values).
2440 * * **bpf_spin_lock** is available to root only.
2441 * * Tracing programs and socket filter programs cannot use
2442 * **bpf_spin_lock**\ () due to insufficient preemption checks
2443 * (but this may change in the future).
2444 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2448 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2450 * Release the *lock* previously locked by a call to
2451 * **bpf_spin_lock**\ (\ *lock*\ ).
2455 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2457 * This helper gets a **struct bpf_sock** pointer such
2458 * that all the fields in this **bpf_sock** can be accessed.
2460 * A **struct bpf_sock** pointer on success, or **NULL** in
2463 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2465 * This helper gets a **struct bpf_tcp_sock** pointer from a
2466 * **struct bpf_sock** pointer.
2468 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2471 * int bpf_skb_ecn_set_ce(struct sk_buf *skb)
2473 * Set ECN (Explicit Congestion Notification) field of IP header
2474 * to **CE** (Congestion Encountered) if current value is **ECT**
2475 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2478 * 1 if the **CE** flag is set (either by the current helper call
2479 * or because it was already present), 0 if it is not set.
2481 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2483 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2484 * **bpf_sk_release**\ () is unnecessary and not allowed.
2486 * A **struct bpf_sock** pointer on success, or **NULL** in
2489 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2491 * Look for TCP socket matching *tuple*, optionally in a child
2492 * network namespace *netns*. The return value must be checked,
2493 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2495 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2496 * that it also returns timewait or request sockets. Use
2497 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2500 * This helper is available only if the kernel was compiled with
2501 * **CONFIG_NET** configuration option.
2503 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2504 * For sockets with reuseport option, the **struct bpf_sock**
2505 * result is from *reuse*\ **->socks**\ [] using the hash of the
2508 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2510 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2511 * the listening socket in *sk*.
2513 * *iph* points to the start of the IPv4 or IPv6 header, while
2514 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2515 * **sizeof**\ (**struct ip6hdr**).
2517 * *th* points to the start of the TCP header, while *th_len*
2518 * contains **sizeof**\ (**struct tcphdr**).
2521 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2524 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2526 * Get name of sysctl in /proc/sys/ and copy it into provided by
2527 * program buffer *buf* of size *buf_len*.
2529 * The buffer is always NUL terminated, unless it's zero-sized.
2531 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2532 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2533 * only (e.g. "tcp_mem").
2535 * Number of character copied (not including the trailing NUL).
2537 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2538 * truncated name in this case).
2540 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2542 * Get current value of sysctl as it is presented in /proc/sys
2543 * (incl. newline, etc), and copy it as a string into provided
2544 * by program buffer *buf* of size *buf_len*.
2546 * The whole value is copied, no matter what file position user
2547 * space issued e.g. sys_read at.
2549 * The buffer is always NUL terminated, unless it's zero-sized.
2551 * Number of character copied (not including the trailing NUL).
2553 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2554 * truncated name in this case).
2556 * **-EINVAL** if current value was unavailable, e.g. because
2557 * sysctl is uninitialized and read returns -EIO for it.
2559 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2561 * Get new value being written by user space to sysctl (before
2562 * the actual write happens) and copy it as a string into
2563 * provided by program buffer *buf* of size *buf_len*.
2565 * User space may write new value at file position > 0.
2567 * The buffer is always NUL terminated, unless it's zero-sized.
2569 * Number of character copied (not including the trailing NUL).
2571 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2572 * truncated name in this case).
2574 * **-EINVAL** if sysctl is being read.
2576 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2578 * Override new value being written by user space to sysctl with
2579 * value provided by program in buffer *buf* of size *buf_len*.
2581 * *buf* should contain a string in same form as provided by user
2582 * space on sysctl write.
2584 * User space may write new value at file position > 0. To override
2585 * the whole sysctl value file position should be set to zero.
2589 * **-E2BIG** if the *buf_len* is too big.
2591 * **-EINVAL** if sysctl is being read.
2593 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2595 * Convert the initial part of the string from buffer *buf* of
2596 * size *buf_len* to a long integer according to the given base
2597 * and save the result in *res*.
2599 * The string may begin with an arbitrary amount of white space
2600 * (as determined by **isspace**\ (3)) followed by a single
2601 * optional '**-**' sign.
2603 * Five least significant bits of *flags* encode base, other bits
2604 * are currently unused.
2606 * Base must be either 8, 10, 16 or 0 to detect it automatically
2607 * similar to user space **strtol**\ (3).
2609 * Number of characters consumed on success. Must be positive but
2610 * no more than *buf_len*.
2612 * **-EINVAL** if no valid digits were found or unsupported base
2615 * **-ERANGE** if resulting value was out of range.
2617 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2619 * Convert the initial part of the string from buffer *buf* of
2620 * size *buf_len* to an unsigned long integer according to the
2621 * given base and save the result in *res*.
2623 * The string may begin with an arbitrary amount of white space
2624 * (as determined by **isspace**\ (3)).
2626 * Five least significant bits of *flags* encode base, other bits
2627 * are currently unused.
2629 * Base must be either 8, 10, 16 or 0 to detect it automatically
2630 * similar to user space **strtoul**\ (3).
2632 * Number of characters consumed on success. Must be positive but
2633 * no more than *buf_len*.
2635 * **-EINVAL** if no valid digits were found or unsupported base
2638 * **-ERANGE** if resulting value was out of range.
2640 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2642 * Get a bpf-local-storage from a *sk*.
2644 * Logically, it could be thought of getting the value from
2645 * a *map* with *sk* as the **key**. From this
2646 * perspective, the usage is not much different from
2647 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2648 * helper enforces the key must be a full socket and the map must
2649 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2651 * Underneath, the value is stored locally at *sk* instead of
2652 * the *map*. The *map* is used as the bpf-local-storage
2653 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2654 * searched against all bpf-local-storages residing at *sk*.
2656 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2657 * used such that a new bpf-local-storage will be
2658 * created if one does not exist. *value* can be used
2659 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2660 * the initial value of a bpf-local-storage. If *value* is
2661 * **NULL**, the new bpf-local-storage will be zero initialized.
2663 * A bpf-local-storage pointer is returned on success.
2665 * **NULL** if not found or there was an error in adding
2666 * a new bpf-local-storage.
2668 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2670 * Delete a bpf-local-storage from a *sk*.
2674 * **-ENOENT** if the bpf-local-storage cannot be found.
2676 #define __BPF_FUNC_MAPPER(FN) \
2678 FN(map_lookup_elem), \
2679 FN(map_update_elem), \
2680 FN(map_delete_elem), \
2684 FN(get_prandom_u32), \
2685 FN(get_smp_processor_id), \
2686 FN(skb_store_bytes), \
2687 FN(l3_csum_replace), \
2688 FN(l4_csum_replace), \
2690 FN(clone_redirect), \
2691 FN(get_current_pid_tgid), \
2692 FN(get_current_uid_gid), \
2693 FN(get_current_comm), \
2694 FN(get_cgroup_classid), \
2695 FN(skb_vlan_push), \
2697 FN(skb_get_tunnel_key), \
2698 FN(skb_set_tunnel_key), \
2699 FN(perf_event_read), \
2701 FN(get_route_realm), \
2702 FN(perf_event_output), \
2703 FN(skb_load_bytes), \
2706 FN(skb_get_tunnel_opt), \
2707 FN(skb_set_tunnel_opt), \
2708 FN(skb_change_proto), \
2709 FN(skb_change_type), \
2710 FN(skb_under_cgroup), \
2711 FN(get_hash_recalc), \
2712 FN(get_current_task), \
2713 FN(probe_write_user), \
2714 FN(current_task_under_cgroup), \
2715 FN(skb_change_tail), \
2716 FN(skb_pull_data), \
2718 FN(set_hash_invalid), \
2719 FN(get_numa_node_id), \
2720 FN(skb_change_head), \
2721 FN(xdp_adjust_head), \
2722 FN(probe_read_str), \
2723 FN(get_socket_cookie), \
2724 FN(get_socket_uid), \
2727 FN(skb_adjust_room), \
2729 FN(sk_redirect_map), \
2730 FN(sock_map_update), \
2731 FN(xdp_adjust_meta), \
2732 FN(perf_event_read_value), \
2733 FN(perf_prog_read_value), \
2735 FN(override_return), \
2736 FN(sock_ops_cb_flags_set), \
2737 FN(msg_redirect_map), \
2738 FN(msg_apply_bytes), \
2739 FN(msg_cork_bytes), \
2740 FN(msg_pull_data), \
2742 FN(xdp_adjust_tail), \
2743 FN(skb_get_xfrm_state), \
2745 FN(skb_load_bytes_relative), \
2747 FN(sock_hash_update), \
2748 FN(msg_redirect_hash), \
2749 FN(sk_redirect_hash), \
2750 FN(lwt_push_encap), \
2751 FN(lwt_seg6_store_bytes), \
2752 FN(lwt_seg6_adjust_srh), \
2753 FN(lwt_seg6_action), \
2756 FN(skb_cgroup_id), \
2757 FN(get_current_cgroup_id), \
2758 FN(get_local_storage), \
2759 FN(sk_select_reuseport), \
2760 FN(skb_ancestor_cgroup_id), \
2761 FN(sk_lookup_tcp), \
2762 FN(sk_lookup_udp), \
2764 FN(map_push_elem), \
2766 FN(map_peek_elem), \
2767 FN(msg_push_data), \
2769 FN(rc_pointer_rel), \
2774 FN(skb_ecn_set_ce), \
2775 FN(get_listener_sock), \
2776 FN(skc_lookup_tcp), \
2777 FN(tcp_check_syncookie), \
2778 FN(sysctl_get_name), \
2779 FN(sysctl_get_current_value), \
2780 FN(sysctl_get_new_value), \
2781 FN(sysctl_set_new_value), \
2784 FN(sk_storage_get), \
2785 FN(sk_storage_delete),
2787 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2788 * function eBPF program intends to call
2790 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2792 __BPF_FUNC_MAPPER(__BPF_ENUM_FN
)
2795 #undef __BPF_ENUM_FN
2797 /* All flags used by eBPF helper functions, placed here. */
2799 /* BPF_FUNC_skb_store_bytes flags. */
2800 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2801 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2803 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2804 * First 4 bits are for passing the header field size.
2806 #define BPF_F_HDR_FIELD_MASK 0xfULL
2808 /* BPF_FUNC_l4_csum_replace flags. */
2809 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2810 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2811 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2813 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2814 #define BPF_F_INGRESS (1ULL << 0)
2816 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2817 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2819 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2820 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2821 #define BPF_F_USER_STACK (1ULL << 8)
2822 /* flags used by BPF_FUNC_get_stackid only. */
2823 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2824 #define BPF_F_REUSE_STACKID (1ULL << 10)
2825 /* flags used by BPF_FUNC_get_stack only. */
2826 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2828 /* BPF_FUNC_skb_set_tunnel_key flags. */
2829 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2830 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2831 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2833 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2834 * BPF_FUNC_perf_event_read_value flags.
2836 #define BPF_F_INDEX_MASK 0xffffffffULL
2837 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2838 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2839 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2841 /* Current network namespace */
2842 #define BPF_F_CURRENT_NETNS (-1L)
2844 /* BPF_FUNC_skb_adjust_room flags. */
2845 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
2847 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
2848 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
2850 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
2851 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
2852 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
2853 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
2854 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
2855 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
2856 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
2858 /* BPF_FUNC_sysctl_get_name flags. */
2859 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
2861 /* BPF_FUNC_sk_storage_get flags */
2862 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0)
2864 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2865 enum bpf_adj_room_mode
{
2870 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2871 enum bpf_hdr_start_off
{
2876 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2877 enum bpf_lwt_encap_mode
{
2879 BPF_LWT_ENCAP_SEG6_INLINE
,
2883 #define __bpf_md_ptr(type, name) \
2887 } __attribute__((aligned(8)))
2889 /* user accessible mirror of in-kernel sk_buff.
2890 * new fields can only be added to the end of this structure
2896 __u32 queue_mapping
;
2902 __u32 ingress_ifindex
;
2912 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2914 __u32 remote_ip4
; /* Stored in network byte order */
2915 __u32 local_ip4
; /* Stored in network byte order */
2916 __u32 remote_ip6
[4]; /* Stored in network byte order */
2917 __u32 local_ip6
[4]; /* Stored in network byte order */
2918 __u32 remote_port
; /* Stored in network byte order */
2919 __u32 local_port
; /* stored in host byte order */
2923 __bpf_md_ptr(struct bpf_flow_keys
*, flow_keys
);
2927 __bpf_md_ptr(struct bpf_sock
*, sk
);
2930 struct bpf_tunnel_key
{
2934 __u32 remote_ipv6
[4];
2938 __u16 tunnel_ext
; /* Padding, future use. */
2942 /* user accessible mirror of in-kernel xfrm_state.
2943 * new fields can only be added to the end of this structure
2945 struct bpf_xfrm_state
{
2947 __u32 spi
; /* Stored in network byte order */
2949 __u16 ext
; /* Padding, future use. */
2951 __u32 remote_ipv4
; /* Stored in network byte order */
2952 __u32 remote_ipv6
[4]; /* Stored in network byte order */
2956 /* Generic BPF return codes which all BPF program types may support.
2957 * The values are binary compatible with their TC_ACT_* counter-part to
2958 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2961 * XDP is handled seprately, see XDP_*.
2969 /* >127 are reserved for prog type specific return codes.
2971 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
2972 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
2973 * changed and should be routed based on its new L3 header.
2974 * (This is an L3 redirect, as opposed to L2 redirect
2975 * represented by BPF_REDIRECT above).
2977 BPF_LWT_REROUTE
= 128,
2987 /* IP address also allows 1 and 2 bytes access */
2990 __u32 src_port
; /* host byte order */
2991 __u32 dst_port
; /* network byte order */
2997 struct bpf_tcp_sock
{
2998 __u32 snd_cwnd
; /* Sending congestion window */
2999 __u32 srtt_us
; /* smoothed round trip time << 3 in usecs */
3001 __u32 snd_ssthresh
; /* Slow start size threshold */
3002 __u32 rcv_nxt
; /* What we want to receive next */
3003 __u32 snd_nxt
; /* Next sequence we send */
3004 __u32 snd_una
; /* First byte we want an ack for */
3005 __u32 mss_cache
; /* Cached effective mss, not including SACKS */
3006 __u32 ecn_flags
; /* ECN status bits. */
3007 __u32 rate_delivered
; /* saved rate sample: packets delivered */
3008 __u32 rate_interval_us
; /* saved rate sample: time elapsed */
3009 __u32 packets_out
; /* Packets which are "in flight" */
3010 __u32 retrans_out
; /* Retransmitted packets out */
3011 __u32 total_retrans
; /* Total retransmits for entire connection */
3012 __u32 segs_in
; /* RFC4898 tcpEStatsPerfSegsIn
3013 * total number of segments in.
3015 __u32 data_segs_in
; /* RFC4898 tcpEStatsPerfDataSegsIn
3016 * total number of data segments in.
3018 __u32 segs_out
; /* RFC4898 tcpEStatsPerfSegsOut
3019 * The total number of segments sent.
3021 __u32 data_segs_out
; /* RFC4898 tcpEStatsPerfDataSegsOut
3022 * total number of data segments sent.
3024 __u32 lost_out
; /* Lost packets */
3025 __u32 sacked_out
; /* SACK'd packets */
3026 __u64 bytes_received
; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3027 * sum(delta(rcv_nxt)), or how many bytes
3030 __u64 bytes_acked
; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3031 * sum(delta(snd_una)), or how many bytes
3036 struct bpf_sock_tuple
{
3053 #define XDP_PACKET_HEADROOM 256
3055 /* User return codes for XDP prog type.
3056 * A valid XDP program must return one of these defined values. All other
3057 * return codes are reserved for future use. Unknown return codes will
3058 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3068 /* user accessible metadata for XDP packet hook
3069 * new fields must be added to the end of this structure
3075 /* Below access go through struct xdp_rxq_info */
3076 __u32 ingress_ifindex
; /* rxq->dev->ifindex */
3077 __u32 rx_queue_index
; /* rxq->queue_index */
3085 /* user accessible metadata for SK_MSG packet hook, new fields must
3086 * be added to the end of this structure
3089 __bpf_md_ptr(void *, data
);
3090 __bpf_md_ptr(void *, data_end
);
3093 __u32 remote_ip4
; /* Stored in network byte order */
3094 __u32 local_ip4
; /* Stored in network byte order */
3095 __u32 remote_ip6
[4]; /* Stored in network byte order */
3096 __u32 local_ip6
[4]; /* Stored in network byte order */
3097 __u32 remote_port
; /* Stored in network byte order */
3098 __u32 local_port
; /* stored in host byte order */
3099 __u32 size
; /* Total size of sk_msg */
3102 struct sk_reuseport_md
{
3104 * Start of directly accessible data. It begins from
3105 * the tcp/udp header.
3107 __bpf_md_ptr(void *, data
);
3108 /* End of directly accessible data */
3109 __bpf_md_ptr(void *, data_end
);
3111 * Total length of packet (starting from the tcp/udp header).
3112 * Note that the directly accessible bytes (data_end - data)
3113 * could be less than this "len". Those bytes could be
3114 * indirectly read by a helper "bpf_skb_load_bytes()".
3118 * Eth protocol in the mac header (network byte order). e.g.
3119 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3122 __u32 ip_protocol
; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3123 __u32 bind_inany
; /* Is sock bound to an INANY address? */
3124 __u32 hash
; /* A hash of the packet 4 tuples */
3127 #define BPF_TAG_SIZE 8
3129 struct bpf_prog_info
{
3132 __u8 tag
[BPF_TAG_SIZE
];
3133 __u32 jited_prog_len
;
3134 __u32 xlated_prog_len
;
3135 __aligned_u64 jited_prog_insns
;
3136 __aligned_u64 xlated_prog_insns
;
3137 __u64 load_time
; /* ns since boottime */
3138 __u32 created_by_uid
;
3140 __aligned_u64 map_ids
;
3141 char name
[BPF_OBJ_NAME_LEN
];
3143 __u32 gpl_compatible
:1;
3146 __u32 nr_jited_ksyms
;
3147 __u32 nr_jited_func_lens
;
3148 __aligned_u64 jited_ksyms
;
3149 __aligned_u64 jited_func_lens
;
3151 __u32 func_info_rec_size
;
3152 __aligned_u64 func_info
;
3155 __aligned_u64 line_info
;
3156 __aligned_u64 jited_line_info
;
3157 __u32 nr_jited_line_info
;
3158 __u32 line_info_rec_size
;
3159 __u32 jited_line_info_rec_size
;
3161 __aligned_u64 prog_tags
;
3164 } __attribute__((aligned(8)));
3166 struct bpf_map_info
{
3173 char name
[BPF_OBJ_NAME_LEN
];
3179 __u32 btf_key_type_id
;
3180 __u32 btf_value_type_id
;
3181 } __attribute__((aligned(8)));
3183 struct bpf_btf_info
{
3187 } __attribute__((aligned(8)));
3189 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3190 * by user and intended to be used by socket (e.g. to bind to, depends on
3191 * attach attach type).
3193 struct bpf_sock_addr
{
3194 __u32 user_family
; /* Allows 4-byte read, but no write. */
3195 __u32 user_ip4
; /* Allows 1,2,4-byte read and 4-byte write.
3196 * Stored in network byte order.
3198 __u32 user_ip6
[4]; /* Allows 1,2,4-byte read an 4-byte write.
3199 * Stored in network byte order.
3201 __u32 user_port
; /* Allows 4-byte read and write.
3202 * Stored in network byte order
3204 __u32 family
; /* Allows 4-byte read, but no write */
3205 __u32 type
; /* Allows 4-byte read, but no write */
3206 __u32 protocol
; /* Allows 4-byte read, but no write */
3207 __u32 msg_src_ip4
; /* Allows 1,2,4-byte read an 4-byte write.
3208 * Stored in network byte order.
3210 __u32 msg_src_ip6
[4]; /* Allows 1,2,4-byte read an 4-byte write.
3211 * Stored in network byte order.
3215 /* User bpf_sock_ops struct to access socket values and specify request ops
3216 * and their replies.
3217 * Some of this fields are in network (bigendian) byte order and may need
3218 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3219 * New fields can only be added at the end of this structure
3221 struct bpf_sock_ops
{
3224 __u32 args
[4]; /* Optionally passed to bpf program */
3225 __u32 reply
; /* Returned by bpf program */
3226 __u32 replylong
[4]; /* Optionally returned by bpf prog */
3229 __u32 remote_ip4
; /* Stored in network byte order */
3230 __u32 local_ip4
; /* Stored in network byte order */
3231 __u32 remote_ip6
[4]; /* Stored in network byte order */
3232 __u32 local_ip6
[4]; /* Stored in network byte order */
3233 __u32 remote_port
; /* Stored in network byte order */
3234 __u32 local_port
; /* stored in host byte order */
3235 __u32 is_fullsock
; /* Some TCP fields are only valid if
3236 * there is a full socket. If not, the
3237 * fields read as zero.
3240 __u32 srtt_us
; /* Averaged RTT << 3 in usecs */
3241 __u32 bpf_sock_ops_cb_flags
; /* flags defined in uapi/linux/tcp.h */
3250 __u32 rate_delivered
;
3251 __u32 rate_interval_us
;
3254 __u32 total_retrans
;
3258 __u32 data_segs_out
;
3262 __u64 bytes_received
;
3266 /* Definitions for bpf_sock_ops_cb_flags */
3267 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
3268 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
3269 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3270 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
3271 * supported cb flags
3274 /* List of known BPF sock_ops operators.
3275 * New entries can only be added at the end
3279 BPF_SOCK_OPS_TIMEOUT_INIT
, /* Should return SYN-RTO value to use or
3280 * -1 if default value should be used
3282 BPF_SOCK_OPS_RWND_INIT
, /* Should return initial advertized
3283 * window (in packets) or -1 if default
3284 * value should be used
3286 BPF_SOCK_OPS_TCP_CONNECT_CB
, /* Calls BPF program right before an
3287 * active connection is initialized
3289 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB
, /* Calls BPF program when an
3290 * active connection is
3293 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB
, /* Calls BPF program when a
3294 * passive connection is
3297 BPF_SOCK_OPS_NEEDS_ECN
, /* If connection's congestion control
3300 BPF_SOCK_OPS_BASE_RTT
, /* Get base RTT. The correct value is
3301 * based on the path and may be
3302 * dependent on the congestion control
3303 * algorithm. In general it indicates
3304 * a congestion threshold. RTTs above
3305 * this indicate congestion
3307 BPF_SOCK_OPS_RTO_CB
, /* Called when an RTO has triggered.
3308 * Arg1: value of icsk_retransmits
3309 * Arg2: value of icsk_rto
3310 * Arg3: whether RTO has expired
3312 BPF_SOCK_OPS_RETRANS_CB
, /* Called when skb is retransmitted.
3313 * Arg1: sequence number of 1st byte
3315 * Arg3: return value of
3316 * tcp_transmit_skb (0 => success)
3318 BPF_SOCK_OPS_STATE_CB
, /* Called when TCP changes state.
3322 BPF_SOCK_OPS_TCP_LISTEN_CB
, /* Called on listen(2), right after
3323 * socket transition to LISTEN state.
3327 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3328 * changes between the TCP and BPF versions. Ideally this should never happen.
3329 * If it does, we need to add code to convert them before calling
3330 * the BPF sock_ops function.
3333 BPF_TCP_ESTABLISHED
= 1,
3343 BPF_TCP_CLOSING
, /* Now a valid state */
3344 BPF_TCP_NEW_SYN_RECV
,
3346 BPF_TCP_MAX_STATES
/* Leave at the end! */
3349 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3350 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3352 struct bpf_perf_event_value
{
3358 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3359 #define BPF_DEVCG_ACC_READ (1ULL << 1)
3360 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3362 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3363 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3365 struct bpf_cgroup_dev_ctx
{
3366 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3372 struct bpf_raw_tracepoint_args
{
3376 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3377 * OUTPUT: Do lookup from egress perspective; default is ingress
3379 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
3380 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
3383 BPF_FIB_LKUP_RET_SUCCESS
, /* lookup successful */
3384 BPF_FIB_LKUP_RET_BLACKHOLE
, /* dest is blackholed; can be dropped */
3385 BPF_FIB_LKUP_RET_UNREACHABLE
, /* dest is unreachable; can be dropped */
3386 BPF_FIB_LKUP_RET_PROHIBIT
, /* dest not allowed; can be dropped */
3387 BPF_FIB_LKUP_RET_NOT_FWDED
, /* packet is not forwarded */
3388 BPF_FIB_LKUP_RET_FWD_DISABLED
, /* fwding is not enabled on ingress */
3389 BPF_FIB_LKUP_RET_UNSUPP_LWT
, /* fwd requires encapsulation */
3390 BPF_FIB_LKUP_RET_NO_NEIGH
, /* no neighbor entry for nh */
3391 BPF_FIB_LKUP_RET_FRAG_NEEDED
, /* fragmentation required to fwd */
3394 struct bpf_fib_lookup
{
3395 /* input: network family for lookup (AF_INET, AF_INET6)
3396 * output: network family of egress nexthop
3400 /* set if lookup is to consider L4 data - e.g., FIB rules */
3405 /* total length of packet from network header - used for MTU check */
3408 /* input: L3 device index for lookup
3409 * output: device index from FIB lookup
3414 /* inputs to lookup */
3415 __u8 tos
; /* AF_INET */
3416 __be32 flowinfo
; /* AF_INET6, flow_label + priority */
3418 /* output: metric of fib result (IPv4/IPv6 only) */
3424 __u32 ipv6_src
[4]; /* in6_addr; network order */
3427 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3428 * network header. output: bpf_fib_lookup sets to gateway address
3429 * if FIB lookup returns gateway route
3433 __u32 ipv6_dst
[4]; /* in6_addr; network order */
3437 __be16 h_vlan_proto
;
3439 __u8 smac
[6]; /* ETH_ALEN */
3440 __u8 dmac
[6]; /* ETH_ALEN */
3443 enum bpf_task_fd_type
{
3444 BPF_FD_TYPE_RAW_TRACEPOINT
, /* tp name */
3445 BPF_FD_TYPE_TRACEPOINT
, /* tp name */
3446 BPF_FD_TYPE_KPROBE
, /* (symbol + offset) or addr */
3447 BPF_FD_TYPE_KRETPROBE
, /* (symbol + offset) or addr */
3448 BPF_FD_TYPE_UPROBE
, /* filename + offset */
3449 BPF_FD_TYPE_URETPROBE
, /* filename + offset */
3452 struct bpf_flow_keys
{
3455 __u16 addr_proto
; /* ETH_P_* of valid addrs */
3469 __u32 ipv6_src
[4]; /* in6_addr; network order */
3470 __u32 ipv6_dst
[4]; /* in6_addr; network order */
3475 struct bpf_func_info
{
3480 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3481 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3483 struct bpf_line_info
{
3485 __u32 file_name_off
;
3490 struct bpf_spin_lock
{
3495 __u32 write
; /* Sysctl is being read (= 0) or written (= 1).
3496 * Allows 1,2,4-byte read, but no write.
3498 __u32 file_pos
; /* Sysctl file position to read from, write to.
3499 * Allows 1,2,4-byte read an 4-byte write.
3503 #endif /* _UAPI__LINUX_BPF_H__ */