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 BPF_CGROUP_UDP4_RECVMSG
,
196 BPF_CGROUP_UDP6_RECVMSG
,
197 __MAX_BPF_ATTACH_TYPE
200 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
202 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
204 * NONE(default): No further bpf programs allowed in the subtree.
206 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
207 * the program in this cgroup yields to sub-cgroup program.
209 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
210 * that cgroup program gets run in addition to the program in this cgroup.
212 * Only one program is allowed to be attached to a cgroup with
213 * NONE or BPF_F_ALLOW_OVERRIDE flag.
214 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
215 * release old program and attach the new one. Attach flags has to match.
217 * Multiple programs are allowed to be attached to a cgroup with
218 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
219 * (those that were attached first, run first)
220 * The programs of sub-cgroup are executed first, then programs of
221 * this cgroup and then programs of parent cgroup.
222 * When children program makes decision (like picking TCP CA or sock bind)
223 * parent program has a chance to override it.
225 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
226 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
228 * cgrp1 (MULTI progs A, B) ->
229 * cgrp2 (OVERRIDE prog C) ->
230 * cgrp3 (MULTI prog D) ->
231 * cgrp4 (OVERRIDE prog E) ->
232 * cgrp5 (NONE prog F)
233 * the event in cgrp5 triggers execution of F,D,A,B in that order.
234 * if prog F is detached, the execution is E,D,A,B
235 * if prog F and D are detached, the execution is E,A,B
236 * if prog F, E and D are detached, the execution is C,A,B
238 * All eligible programs are executed regardless of return code from
241 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
242 #define BPF_F_ALLOW_MULTI (1U << 1)
244 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
245 * verifier will perform strict alignment checking as if the kernel
246 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
247 * and NET_IP_ALIGN defined to 2.
249 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
251 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
252 * verifier will allow any alignment whatsoever. On platforms
253 * with strict alignment requirements for loads ands stores (such
254 * as sparc and mips) the verifier validates that all loads and
255 * stores provably follow this requirement. This flag turns that
256 * checking and enforcement off.
258 * It is mostly used for testing when we want to validate the
259 * context and memory access aspects of the verifier, but because
260 * of an unaligned access the alignment check would trigger before
261 * the one we are interested in.
263 #define BPF_F_ANY_ALIGNMENT (1U << 1)
265 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
268 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
269 * insn[0].imm: map fd map fd
270 * insn[1].imm: 0 offset into value
273 * ldimm64 rewrite: address of map address of map[0]+offset
274 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
276 #define BPF_PSEUDO_MAP_FD 1
277 #define BPF_PSEUDO_MAP_VALUE 2
279 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
280 * offset to another bpf function
282 #define BPF_PSEUDO_CALL 1
284 /* flags for BPF_MAP_UPDATE_ELEM command */
285 #define BPF_ANY 0 /* create new element or update existing */
286 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
287 #define BPF_EXIST 2 /* update existing element */
288 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
290 /* flags for BPF_MAP_CREATE command */
291 #define BPF_F_NO_PREALLOC (1U << 0)
292 /* Instead of having one common LRU list in the
293 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
294 * which can scale and perform better.
295 * Note, the LRU nodes (including free nodes) cannot be moved
296 * across different LRU lists.
298 #define BPF_F_NO_COMMON_LRU (1U << 1)
299 /* Specify numa node during map creation */
300 #define BPF_F_NUMA_NODE (1U << 2)
302 #define BPF_OBJ_NAME_LEN 16U
304 /* Flags for accessing BPF object from syscall side. */
305 #define BPF_F_RDONLY (1U << 3)
306 #define BPF_F_WRONLY (1U << 4)
308 /* Flag for stack_map, store build_id+offset instead of pointer */
309 #define BPF_F_STACK_BUILD_ID (1U << 5)
311 /* Zero-initialize hash function seed. This should only be used for testing. */
312 #define BPF_F_ZERO_SEED (1U << 6)
314 /* Flags for accessing BPF object from program side. */
315 #define BPF_F_RDONLY_PROG (1U << 7)
316 #define BPF_F_WRONLY_PROG (1U << 8)
318 /* flags for BPF_PROG_QUERY */
319 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
321 enum bpf_stack_build_id_status
{
322 /* user space need an empty entry to identify end of a trace */
323 BPF_STACK_BUILD_ID_EMPTY
= 0,
324 /* with valid build_id and offset */
325 BPF_STACK_BUILD_ID_VALID
= 1,
326 /* couldn't get build_id, fallback to ip */
327 BPF_STACK_BUILD_ID_IP
= 2,
330 #define BPF_BUILD_ID_SIZE 20
331 struct bpf_stack_build_id
{
333 unsigned char build_id
[BPF_BUILD_ID_SIZE
];
341 struct { /* anonymous struct used by BPF_MAP_CREATE command */
342 __u32 map_type
; /* one of enum bpf_map_type */
343 __u32 key_size
; /* size of key in bytes */
344 __u32 value_size
; /* size of value in bytes */
345 __u32 max_entries
; /* max number of entries in a map */
346 __u32 map_flags
; /* BPF_MAP_CREATE related
347 * flags defined above.
349 __u32 inner_map_fd
; /* fd pointing to the inner map */
350 __u32 numa_node
; /* numa node (effective only if
351 * BPF_F_NUMA_NODE is set).
353 char map_name
[BPF_OBJ_NAME_LEN
];
354 __u32 map_ifindex
; /* ifindex of netdev to create on */
355 __u32 btf_fd
; /* fd pointing to a BTF type data */
356 __u32 btf_key_type_id
; /* BTF type_id of the key */
357 __u32 btf_value_type_id
; /* BTF type_id of the value */
360 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
365 __aligned_u64 next_key
;
370 struct { /* anonymous struct used by BPF_PROG_LOAD command */
371 __u32 prog_type
; /* one of enum bpf_prog_type */
374 __aligned_u64 license
;
375 __u32 log_level
; /* verbosity level of verifier */
376 __u32 log_size
; /* size of user buffer */
377 __aligned_u64 log_buf
; /* user supplied buffer */
378 __u32 kern_version
; /* not used */
380 char prog_name
[BPF_OBJ_NAME_LEN
];
381 __u32 prog_ifindex
; /* ifindex of netdev to prep for */
382 /* For some prog types expected attach type must be known at
383 * load time to verify attach type specific parts of prog
384 * (context accesses, allowed helpers, etc).
386 __u32 expected_attach_type
;
387 __u32 prog_btf_fd
; /* fd pointing to BTF type data */
388 __u32 func_info_rec_size
; /* userspace bpf_func_info size */
389 __aligned_u64 func_info
; /* func info */
390 __u32 func_info_cnt
; /* number of bpf_func_info records */
391 __u32 line_info_rec_size
; /* userspace bpf_line_info size */
392 __aligned_u64 line_info
; /* line info */
393 __u32 line_info_cnt
; /* number of bpf_line_info records */
396 struct { /* anonymous struct used by BPF_OBJ_* commands */
397 __aligned_u64 pathname
;
402 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
403 __u32 target_fd
; /* container object to attach to */
404 __u32 attach_bpf_fd
; /* eBPF program to attach */
409 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
412 __u32 data_size_in
; /* input: len of data_in */
413 __u32 data_size_out
; /* input/output: len of data_out
414 * returns ENOSPC if data_out
417 __aligned_u64 data_in
;
418 __aligned_u64 data_out
;
421 __u32 ctx_size_in
; /* input: len of ctx_in */
422 __u32 ctx_size_out
; /* input/output: len of ctx_out
423 * returns ENOSPC if ctx_out
426 __aligned_u64 ctx_in
;
427 __aligned_u64 ctx_out
;
430 struct { /* anonymous struct used by BPF_*_GET_*_ID */
441 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
447 struct { /* anonymous struct used by BPF_PROG_QUERY command */
448 __u32 target_fd
; /* container object to query */
452 __aligned_u64 prog_ids
;
461 struct { /* anonymous struct for BPF_BTF_LOAD */
463 __aligned_u64 btf_log_buf
;
470 __u32 pid
; /* input: pid */
471 __u32 fd
; /* input: fd */
472 __u32 flags
; /* input: flags */
473 __u32 buf_len
; /* input/output: buf len */
474 __aligned_u64 buf
; /* input/output:
475 * tp_name for tracepoint
477 * filename for uprobe
479 __u32 prog_id
; /* output: prod_id */
480 __u32 fd_type
; /* output: BPF_FD_TYPE_* */
481 __u64 probe_offset
; /* output: probe_offset */
482 __u64 probe_addr
; /* output: probe_addr */
484 } __attribute__((aligned(8)));
486 /* The description below is an attempt at providing documentation to eBPF
487 * developers about the multiple available eBPF helper functions. It can be
488 * parsed and used to produce a manual page. The workflow is the following,
489 * and requires the rst2man utility:
491 * $ ./scripts/bpf_helpers_doc.py \
492 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
493 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
494 * $ man /tmp/bpf-helpers.7
496 * Note that in order to produce this external documentation, some RST
497 * formatting is used in the descriptions to get "bold" and "italics" in
498 * manual pages. Also note that the few trailing white spaces are
499 * intentional, removing them would break paragraphs for rst2man.
501 * Start of BPF helper function descriptions:
503 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
505 * Perform a lookup in *map* for an entry associated to *key*.
507 * Map value associated to *key*, or **NULL** if no entry was
510 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
512 * Add or update the value of the entry associated to *key* in
513 * *map* with *value*. *flags* is one of:
516 * The entry for *key* must not exist in the map.
518 * The entry for *key* must already exist in the map.
520 * No condition on the existence of the entry for *key*.
522 * Flag value **BPF_NOEXIST** cannot be used for maps of types
523 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
524 * elements always exist), the helper would return an error.
526 * 0 on success, or a negative error in case of failure.
528 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
530 * Delete entry with *key* from *map*.
532 * 0 on success, or a negative error in case of failure.
534 * int bpf_probe_read(void *dst, u32 size, const void *src)
536 * For tracing programs, safely attempt to read *size* bytes from
537 * address *src* and store the data in *dst*.
539 * 0 on success, or a negative error in case of failure.
541 * u64 bpf_ktime_get_ns(void)
543 * Return the time elapsed since system boot, in nanoseconds.
547 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
549 * This helper is a "printk()-like" facility for debugging. It
550 * prints a message defined by format *fmt* (of size *fmt_size*)
551 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
552 * available. It can take up to three additional **u64**
553 * arguments (as an eBPF helpers, the total number of arguments is
556 * Each time the helper is called, it appends a line to the trace.
557 * The format of the trace is customizable, and the exact output
558 * one will get depends on the options set in
559 * *\/sys/kernel/debug/tracing/trace_options* (see also the
560 * *README* file under the same directory). However, it usually
561 * defaults to something like:
565 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
569 * * ``telnet`` is the name of the current task.
570 * * ``470`` is the PID of the current task.
571 * * ``001`` is the CPU number on which the task is
573 * * In ``.N..``, each character refers to a set of
574 * options (whether irqs are enabled, scheduling
575 * options, whether hard/softirqs are running, level of
576 * preempt_disabled respectively). **N** means that
577 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
579 * * ``419421.045894`` is a timestamp.
580 * * ``0x00000001`` is a fake value used by BPF for the
581 * instruction pointer register.
582 * * ``<formatted msg>`` is the message formatted with
585 * The conversion specifiers supported by *fmt* are similar, but
586 * more limited than for printk(). They are **%d**, **%i**,
587 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
588 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
589 * of field, padding with zeroes, etc.) is available, and the
590 * helper will return **-EINVAL** (but print nothing) if it
591 * encounters an unknown specifier.
593 * Also, note that **bpf_trace_printk**\ () is slow, and should
594 * only be used for debugging purposes. For this reason, a notice
595 * bloc (spanning several lines) is printed to kernel logs and
596 * states that the helper should not be used "for production use"
597 * the first time this helper is used (or more precisely, when
598 * **trace_printk**\ () buffers are allocated). For passing values
599 * to user space, perf events should be preferred.
601 * The number of bytes written to the buffer, or a negative error
602 * in case of failure.
604 * u32 bpf_get_prandom_u32(void)
606 * Get a pseudo-random number.
608 * From a security point of view, this helper uses its own
609 * pseudo-random internal state, and cannot be used to infer the
610 * seed of other random functions in the kernel. However, it is
611 * essential to note that the generator used by the helper is not
612 * cryptographically secure.
614 * A random 32-bit unsigned value.
616 * u32 bpf_get_smp_processor_id(void)
618 * Get the SMP (symmetric multiprocessing) processor id. Note that
619 * all programs run with preemption disabled, which means that the
620 * SMP processor id is stable during all the execution of the
623 * The SMP id of the processor running the program.
625 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
627 * Store *len* bytes from address *from* into the packet
628 * associated to *skb*, at *offset*. *flags* are a combination of
629 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
630 * checksum for the packet after storing the bytes) and
631 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
632 * **->swhash** and *skb*\ **->l4hash** to 0).
634 * A call to this helper is susceptible to change the underlying
635 * packet buffer. Therefore, at load time, all checks on pointers
636 * previously done by the verifier are invalidated and must be
637 * performed again, if the helper is used in combination with
638 * direct packet access.
640 * 0 on success, or a negative error in case of failure.
642 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
644 * Recompute the layer 3 (e.g. IP) checksum for the packet
645 * associated to *skb*. Computation is incremental, so the helper
646 * must know the former value of the header field that was
647 * modified (*from*), the new value of this field (*to*), and the
648 * number of bytes (2 or 4) for this field, stored in *size*.
649 * Alternatively, it is possible to store the difference between
650 * the previous and the new values of the header field in *to*, by
651 * setting *from* and *size* to 0. For both methods, *offset*
652 * indicates the location of the IP checksum within the packet.
654 * This helper works in combination with **bpf_csum_diff**\ (),
655 * which does not update the checksum in-place, but offers more
656 * flexibility and can handle sizes larger than 2 or 4 for the
657 * checksum to update.
659 * A call to this helper is susceptible to change the underlying
660 * packet buffer. Therefore, at load time, all checks on pointers
661 * previously done by the verifier are invalidated and must be
662 * performed again, if the helper is used in combination with
663 * direct packet access.
665 * 0 on success, or a negative error in case of failure.
667 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
669 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
670 * packet associated to *skb*. Computation is incremental, so the
671 * helper must know the former value of the header field that was
672 * modified (*from*), the new value of this field (*to*), and the
673 * number of bytes (2 or 4) for this field, stored on the lowest
674 * four bits of *flags*. Alternatively, it is possible to store
675 * the difference between the previous and the new values of the
676 * header field in *to*, by setting *from* and the four lowest
677 * bits of *flags* to 0. For both methods, *offset* indicates the
678 * location of the IP checksum within the packet. In addition to
679 * the size of the field, *flags* can be added (bitwise OR) actual
680 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
681 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
682 * for updates resulting in a null checksum the value is set to
683 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
684 * the checksum is to be computed against a pseudo-header.
686 * This helper works in combination with **bpf_csum_diff**\ (),
687 * which does not update the checksum in-place, but offers more
688 * flexibility and can handle sizes larger than 2 or 4 for the
689 * checksum to update.
691 * A call to this helper is susceptible to change the underlying
692 * packet buffer. Therefore, at load time, all checks on pointers
693 * previously done by the verifier are invalidated and must be
694 * performed again, if the helper is used in combination with
695 * direct packet access.
697 * 0 on success, or a negative error in case of failure.
699 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
701 * This special helper is used to trigger a "tail call", or in
702 * other words, to jump into another eBPF program. The same stack
703 * frame is used (but values on stack and in registers for the
704 * caller are not accessible to the callee). This mechanism allows
705 * for program chaining, either for raising the maximum number of
706 * available eBPF instructions, or to execute given programs in
707 * conditional blocks. For security reasons, there is an upper
708 * limit to the number of successive tail calls that can be
711 * Upon call of this helper, the program attempts to jump into a
712 * program referenced at index *index* in *prog_array_map*, a
713 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
714 * *ctx*, a pointer to the context.
716 * If the call succeeds, the kernel immediately runs the first
717 * instruction of the new program. This is not a function call,
718 * and it never returns to the previous program. If the call
719 * fails, then the helper has no effect, and the caller continues
720 * to run its subsequent instructions. A call can fail if the
721 * destination program for the jump does not exist (i.e. *index*
722 * is superior to the number of entries in *prog_array_map*), or
723 * if the maximum number of tail calls has been reached for this
724 * chain of programs. This limit is defined in the kernel by the
725 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
726 * which is currently set to 32.
728 * 0 on success, or a negative error in case of failure.
730 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
732 * Clone and redirect the packet associated to *skb* to another
733 * net device of index *ifindex*. Both ingress and egress
734 * interfaces can be used for redirection. The **BPF_F_INGRESS**
735 * value in *flags* is used to make the distinction (ingress path
736 * is selected if the flag is present, egress path otherwise).
737 * This is the only flag supported for now.
739 * In comparison with **bpf_redirect**\ () helper,
740 * **bpf_clone_redirect**\ () has the associated cost of
741 * duplicating the packet buffer, but this can be executed out of
742 * the eBPF program. Conversely, **bpf_redirect**\ () is more
743 * efficient, but it is handled through an action code where the
744 * redirection happens only after the eBPF program has returned.
746 * A call to this helper is susceptible to change the underlying
747 * packet buffer. Therefore, at load time, all checks on pointers
748 * previously done by the verifier are invalidated and must be
749 * performed again, if the helper is used in combination with
750 * direct packet access.
752 * 0 on success, or a negative error in case of failure.
754 * u64 bpf_get_current_pid_tgid(void)
756 * A 64-bit integer containing the current tgid and pid, and
758 * *current_task*\ **->tgid << 32 \|**
759 * *current_task*\ **->pid**.
761 * u64 bpf_get_current_uid_gid(void)
763 * A 64-bit integer containing the current GID and UID, and
764 * created as such: *current_gid* **<< 32 \|** *current_uid*.
766 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
768 * Copy the **comm** attribute of the current task into *buf* of
769 * *size_of_buf*. The **comm** attribute contains the name of
770 * the executable (excluding the path) for the current task. The
771 * *size_of_buf* must be strictly positive. On success, the
772 * helper makes sure that the *buf* is NUL-terminated. On failure,
773 * it is filled with zeroes.
775 * 0 on success, or a negative error in case of failure.
777 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
779 * Retrieve the classid for the current task, i.e. for the net_cls
780 * cgroup to which *skb* belongs.
782 * This helper can be used on TC egress path, but not on ingress.
784 * The net_cls cgroup provides an interface to tag network packets
785 * based on a user-provided identifier for all traffic coming from
786 * the tasks belonging to the related cgroup. See also the related
787 * kernel documentation, available from the Linux sources in file
788 * *Documentation/cgroup-v1/net_cls.txt*.
790 * The Linux kernel has two versions for cgroups: there are
791 * cgroups v1 and cgroups v2. Both are available to users, who can
792 * use a mixture of them, but note that the net_cls cgroup is for
793 * cgroup v1 only. This makes it incompatible with BPF programs
794 * run on cgroups, which is a cgroup-v2-only feature (a socket can
795 * only hold data for one version of cgroups at a time).
797 * This helper is only available is the kernel was compiled with
798 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
799 * "**y**" or to "**m**".
801 * The classid, or 0 for the default unconfigured classid.
803 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
805 * Push a *vlan_tci* (VLAN tag control information) of protocol
806 * *vlan_proto* to the packet associated to *skb*, then update
807 * the checksum. Note that if *vlan_proto* is different from
808 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
809 * be **ETH_P_8021Q**.
811 * A call to this helper is susceptible to change the underlying
812 * packet buffer. Therefore, at load time, all checks on pointers
813 * previously done by the verifier are invalidated and must be
814 * performed again, if the helper is used in combination with
815 * direct packet access.
817 * 0 on success, or a negative error in case of failure.
819 * int bpf_skb_vlan_pop(struct sk_buff *skb)
821 * Pop a VLAN header from the packet associated to *skb*.
823 * A call to this helper is susceptible to change the underlying
824 * packet buffer. Therefore, at load time, all checks on pointers
825 * previously done by the verifier are invalidated and must be
826 * performed again, if the helper is used in combination with
827 * direct packet access.
829 * 0 on success, or a negative error in case of failure.
831 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
833 * Get tunnel metadata. This helper takes a pointer *key* to an
834 * empty **struct bpf_tunnel_key** of **size**, that will be
835 * filled with tunnel metadata for the packet associated to *skb*.
836 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
837 * indicates that the tunnel is based on IPv6 protocol instead of
840 * The **struct bpf_tunnel_key** is an object that generalizes the
841 * principal parameters used by various tunneling protocols into a
842 * single struct. This way, it can be used to easily make a
843 * decision based on the contents of the encapsulation header,
844 * "summarized" in this struct. In particular, it holds the IP
845 * address of the remote end (IPv4 or IPv6, depending on the case)
846 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
847 * this struct exposes the *key*\ **->tunnel_id**, which is
848 * generally mapped to a VNI (Virtual Network Identifier), making
849 * it programmable together with the **bpf_skb_set_tunnel_key**\
852 * Let's imagine that the following code is part of a program
853 * attached to the TC ingress interface, on one end of a GRE
854 * tunnel, and is supposed to filter out all messages coming from
855 * remote ends with IPv4 address other than 10.0.0.1:
860 * struct bpf_tunnel_key key = {};
862 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
864 * return TC_ACT_SHOT; // drop packet
866 * if (key.remote_ipv4 != 0x0a000001)
867 * return TC_ACT_SHOT; // drop packet
869 * return TC_ACT_OK; // accept packet
871 * This interface can also be used with all encapsulation devices
872 * that can operate in "collect metadata" mode: instead of having
873 * one network device per specific configuration, the "collect
874 * metadata" mode only requires a single device where the
875 * configuration can be extracted from this helper.
877 * This can be used together with various tunnels such as VXLan,
878 * Geneve, GRE or IP in IP (IPIP).
880 * 0 on success, or a negative error in case of failure.
882 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
884 * Populate tunnel metadata for packet associated to *skb.* The
885 * tunnel metadata is set to the contents of *key*, of *size*. The
886 * *flags* can be set to a combination of the following values:
888 * **BPF_F_TUNINFO_IPV6**
889 * Indicate that the tunnel is based on IPv6 protocol
891 * **BPF_F_ZERO_CSUM_TX**
892 * For IPv4 packets, add a flag to tunnel metadata
893 * indicating that checksum computation should be skipped
894 * and checksum set to zeroes.
895 * **BPF_F_DONT_FRAGMENT**
896 * Add a flag to tunnel metadata indicating that the
897 * packet should not be fragmented.
898 * **BPF_F_SEQ_NUMBER**
899 * Add a flag to tunnel metadata indicating that a
900 * sequence number should be added to tunnel header before
901 * sending the packet. This flag was added for GRE
902 * encapsulation, but might be used with other protocols
903 * as well in the future.
905 * Here is a typical usage on the transmit path:
909 * struct bpf_tunnel_key key;
911 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
912 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
914 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
915 * helper for additional information.
917 * 0 on success, or a negative error in case of failure.
919 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
921 * Read the value of a perf event counter. This helper relies on a
922 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
923 * the perf event counter is selected when *map* is updated with
924 * perf event file descriptors. The *map* is an array whose size
925 * is the number of available CPUs, and each cell contains a value
926 * relative to one CPU. The value to retrieve is indicated by
927 * *flags*, that contains the index of the CPU to look up, masked
928 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
929 * **BPF_F_CURRENT_CPU** to indicate that the value for the
930 * current CPU should be retrieved.
932 * Note that before Linux 4.13, only hardware perf event can be
935 * Also, be aware that the newer helper
936 * **bpf_perf_event_read_value**\ () is recommended over
937 * **bpf_perf_event_read**\ () in general. The latter has some ABI
938 * quirks where error and counter value are used as a return code
939 * (which is wrong to do since ranges may overlap). This issue is
940 * fixed with **bpf_perf_event_read_value**\ (), which at the same
941 * time provides more features over the **bpf_perf_event_read**\
942 * () interface. Please refer to the description of
943 * **bpf_perf_event_read_value**\ () for details.
945 * The value of the perf event counter read from the map, or a
946 * negative error code in case of failure.
948 * int bpf_redirect(u32 ifindex, u64 flags)
950 * Redirect the packet to another net device of index *ifindex*.
951 * This helper is somewhat similar to **bpf_clone_redirect**\
952 * (), except that the packet is not cloned, which provides
953 * increased performance.
955 * Except for XDP, both ingress and egress interfaces can be used
956 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
957 * to make the distinction (ingress path is selected if the flag
958 * is present, egress path otherwise). Currently, XDP only
959 * supports redirection to the egress interface, and accepts no
962 * The same effect can be attained with the more generic
963 * **bpf_redirect_map**\ (), which requires specific maps to be
964 * used but offers better performance.
966 * For XDP, the helper returns **XDP_REDIRECT** on success or
967 * **XDP_ABORTED** on error. For other program types, the values
968 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
971 * u32 bpf_get_route_realm(struct sk_buff *skb)
973 * Retrieve the realm or the route, that is to say the
974 * **tclassid** field of the destination for the *skb*. The
975 * indentifier retrieved is a user-provided tag, similar to the
976 * one used with the net_cls cgroup (see description for
977 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
978 * held by a route (a destination entry), not by a task.
980 * Retrieving this identifier works with the clsact TC egress hook
981 * (see also **tc-bpf(8)**), or alternatively on conventional
982 * classful egress qdiscs, but not on TC ingress path. In case of
983 * clsact TC egress hook, this has the advantage that, internally,
984 * the destination entry has not been dropped yet in the transmit
985 * path. Therefore, the destination entry does not need to be
986 * artificially held via **netif_keep_dst**\ () for a classful
987 * qdisc until the *skb* is freed.
989 * This helper is available only if the kernel was compiled with
990 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
992 * The realm of the route for the packet associated to *skb*, or 0
995 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
997 * Write raw *data* blob into a special BPF perf event held by
998 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
999 * event must have the following attributes: **PERF_SAMPLE_RAW**
1000 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1001 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1003 * The *flags* are used to indicate the index in *map* for which
1004 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1005 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1006 * to indicate that the index of the current CPU core should be
1009 * The value to write, of *size*, is passed through eBPF stack and
1010 * pointed by *data*.
1012 * The context of the program *ctx* needs also be passed to the
1015 * On user space, a program willing to read the values needs to
1016 * call **perf_event_open**\ () on the perf event (either for
1017 * one or for all CPUs) and to store the file descriptor into the
1018 * *map*. This must be done before the eBPF program can send data
1019 * into it. An example is available in file
1020 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1021 * tree (the eBPF program counterpart is in
1022 * *samples/bpf/trace_output_kern.c*).
1024 * **bpf_perf_event_output**\ () achieves better performance
1025 * than **bpf_trace_printk**\ () for sharing data with user
1026 * space, and is much better suitable for streaming data from eBPF
1029 * Note that this helper is not restricted to tracing use cases
1030 * and can be used with programs attached to TC or XDP as well,
1031 * where it allows for passing data to user space listeners. Data
1034 * * Only custom structs,
1035 * * Only the packet payload, or
1036 * * A combination of both.
1038 * 0 on success, or a negative error in case of failure.
1040 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1042 * This helper was provided as an easy way to load data from a
1043 * packet. It can be used to load *len* bytes from *offset* from
1044 * the packet associated to *skb*, into the buffer pointed by
1047 * Since Linux 4.7, usage of this helper has mostly been replaced
1048 * by "direct packet access", enabling packet data to be
1049 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1050 * pointing respectively to the first byte of packet data and to
1051 * the byte after the last byte of packet data. However, it
1052 * remains useful if one wishes to read large quantities of data
1053 * at once from a packet into the eBPF stack.
1055 * 0 on success, or a negative error in case of failure.
1057 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1059 * Walk a user or a kernel stack and return its id. To achieve
1060 * this, the helper needs *ctx*, which is a pointer to the context
1061 * on which the tracing program is executed, and a pointer to a
1062 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1064 * The last argument, *flags*, holds the number of stack frames to
1065 * skip (from 0 to 255), masked with
1066 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1067 * a combination of the following flags:
1069 * **BPF_F_USER_STACK**
1070 * Collect a user space stack instead of a kernel stack.
1071 * **BPF_F_FAST_STACK_CMP**
1072 * Compare stacks by hash only.
1073 * **BPF_F_REUSE_STACKID**
1074 * If two different stacks hash into the same *stackid*,
1075 * discard the old one.
1077 * The stack id retrieved is a 32 bit long integer handle which
1078 * can be further combined with other data (including other stack
1079 * ids) and used as a key into maps. This can be useful for
1080 * generating a variety of graphs (such as flame graphs or off-cpu
1083 * For walking a stack, this helper is an improvement over
1084 * **bpf_probe_read**\ (), which can be used with unrolled loops
1085 * but is not efficient and consumes a lot of eBPF instructions.
1086 * Instead, **bpf_get_stackid**\ () can collect up to
1087 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1088 * this limit can be controlled with the **sysctl** program, and
1089 * that it should be manually increased in order to profile long
1090 * user stacks (such as stacks for Java programs). To do so, use:
1094 * # sysctl kernel.perf_event_max_stack=<new value>
1096 * The positive or null stack id on success, or a negative error
1097 * in case of failure.
1099 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1101 * Compute a checksum difference, from the raw buffer pointed by
1102 * *from*, of length *from_size* (that must be a multiple of 4),
1103 * towards the raw buffer pointed by *to*, of size *to_size*
1104 * (same remark). An optional *seed* can be added to the value
1105 * (this can be cascaded, the seed may come from a previous call
1108 * This is flexible enough to be used in several ways:
1110 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1111 * checksum, it can be used when pushing new data.
1112 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1113 * checksum, it can be used when removing data from a packet.
1114 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1115 * can be used to compute a diff. Note that *from_size* and
1116 * *to_size* do not need to be equal.
1118 * This helper can be used in combination with
1119 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1120 * which one can feed in the difference computed with
1121 * **bpf_csum_diff**\ ().
1123 * The checksum result, or a negative error code in case of
1126 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1128 * Retrieve tunnel options metadata for the packet associated to
1129 * *skb*, and store the raw tunnel option data to the buffer *opt*
1132 * This helper can be used with encapsulation devices that can
1133 * operate in "collect metadata" mode (please refer to the related
1134 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1135 * more details). A particular example where this can be used is
1136 * in combination with the Geneve encapsulation protocol, where it
1137 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1138 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1139 * the eBPF program. This allows for full customization of these
1142 * The size of the option data retrieved.
1144 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1146 * Set tunnel options metadata for the packet associated to *skb*
1147 * to the option data contained in the raw buffer *opt* of *size*.
1149 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1150 * helper for additional information.
1152 * 0 on success, or a negative error in case of failure.
1154 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1156 * Change the protocol of the *skb* to *proto*. Currently
1157 * supported are transition from IPv4 to IPv6, and from IPv6 to
1158 * IPv4. The helper takes care of the groundwork for the
1159 * transition, including resizing the socket buffer. The eBPF
1160 * program is expected to fill the new headers, if any, via
1161 * **skb_store_bytes**\ () and to recompute the checksums with
1162 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1163 * (). The main case for this helper is to perform NAT64
1164 * operations out of an eBPF program.
1166 * Internally, the GSO type is marked as dodgy so that headers are
1167 * checked and segments are recalculated by the GSO/GRO engine.
1168 * The size for GSO target is adapted as well.
1170 * All values for *flags* are reserved for future usage, and must
1173 * A call to this helper is susceptible to change the underlying
1174 * packet buffer. Therefore, at load time, all checks on pointers
1175 * previously done by the verifier are invalidated and must be
1176 * performed again, if the helper is used in combination with
1177 * direct packet access.
1179 * 0 on success, or a negative error in case of failure.
1181 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1183 * Change the packet type for the packet associated to *skb*. This
1184 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1185 * the eBPF program does not have a write access to *skb*\
1186 * **->pkt_type** beside this helper. Using a helper here allows
1187 * for graceful handling of errors.
1189 * The major use case is to change incoming *skb*s to
1190 * **PACKET_HOST** in a programmatic way instead of having to
1191 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1194 * Note that *type* only allows certain values. At this time, they
1199 * **PACKET_BROADCAST**
1200 * Send packet to all.
1201 * **PACKET_MULTICAST**
1202 * Send packet to group.
1203 * **PACKET_OTHERHOST**
1204 * Send packet to someone else.
1206 * 0 on success, or a negative error in case of failure.
1208 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1210 * Check whether *skb* is a descendant of the cgroup2 held by
1211 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1213 * The return value depends on the result of the test, and can be:
1215 * * 0, if the *skb* failed the cgroup2 descendant test.
1216 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1217 * * A negative error code, if an error occurred.
1219 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1221 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1222 * not set, in particular if the hash was cleared due to mangling,
1223 * recompute this hash. Later accesses to the hash can be done
1224 * directly with *skb*\ **->hash**.
1226 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1227 * prototype with **bpf_skb_change_proto**\ (), or calling
1228 * **bpf_skb_store_bytes**\ () with the
1229 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1230 * the hash and to trigger a new computation for the next call to
1231 * **bpf_get_hash_recalc**\ ().
1235 * u64 bpf_get_current_task(void)
1237 * A pointer to the current task struct.
1239 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1241 * Attempt in a safe way to write *len* bytes from the buffer
1242 * *src* to *dst* in memory. It only works for threads that are in
1243 * user context, and *dst* must be a valid user space address.
1245 * This helper should not be used to implement any kind of
1246 * security mechanism because of TOC-TOU attacks, but rather to
1247 * debug, divert, and manipulate execution of semi-cooperative
1250 * Keep in mind that this feature is meant for experiments, and it
1251 * has a risk of crashing the system and running programs.
1252 * Therefore, when an eBPF program using this helper is attached,
1253 * a warning including PID and process name is printed to kernel
1256 * 0 on success, or a negative error in case of failure.
1258 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1260 * Check whether the probe is being run is the context of a given
1261 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1262 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1264 * The return value depends on the result of the test, and can be:
1266 * * 0, if the *skb* task belongs to the cgroup2.
1267 * * 1, if the *skb* task does not belong to the cgroup2.
1268 * * A negative error code, if an error occurred.
1270 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1272 * Resize (trim or grow) the packet associated to *skb* to the
1273 * new *len*. The *flags* are reserved for future usage, and must
1276 * The basic idea is that the helper performs the needed work to
1277 * change the size of the packet, then the eBPF program rewrites
1278 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1279 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1280 * and others. This helper is a slow path utility intended for
1281 * replies with control messages. And because it is targeted for
1282 * slow path, the helper itself can afford to be slow: it
1283 * implicitly linearizes, unclones and drops offloads from the
1286 * A call to this helper is susceptible to change the underlying
1287 * packet buffer. Therefore, at load time, all checks on pointers
1288 * previously done by the verifier are invalidated and must be
1289 * performed again, if the helper is used in combination with
1290 * direct packet access.
1292 * 0 on success, or a negative error in case of failure.
1294 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1296 * Pull in non-linear data in case the *skb* is non-linear and not
1297 * all of *len* are part of the linear section. Make *len* bytes
1298 * from *skb* readable and writable. If a zero value is passed for
1299 * *len*, then the whole length of the *skb* is pulled.
1301 * This helper is only needed for reading and writing with direct
1304 * For direct packet access, testing that offsets to access
1305 * are within packet boundaries (test on *skb*\ **->data_end**) is
1306 * susceptible to fail if offsets are invalid, or if the requested
1307 * data is in non-linear parts of the *skb*. On failure the
1308 * program can just bail out, or in the case of a non-linear
1309 * buffer, use a helper to make the data available. The
1310 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1311 * the data. Another one consists in using **bpf_skb_pull_data**
1312 * to pull in once the non-linear parts, then retesting and
1313 * eventually access the data.
1315 * At the same time, this also makes sure the *skb* is uncloned,
1316 * which is a necessary condition for direct write. As this needs
1317 * to be an invariant for the write part only, the verifier
1318 * detects writes and adds a prologue that is calling
1319 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1320 * the very beginning in case it is indeed cloned.
1322 * A call to this helper is susceptible to change the underlying
1323 * packet buffer. Therefore, at load time, all checks on pointers
1324 * previously done by the verifier are invalidated and must be
1325 * performed again, if the helper is used in combination with
1326 * direct packet access.
1328 * 0 on success, or a negative error in case of failure.
1330 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1332 * Add the checksum *csum* into *skb*\ **->csum** in case the
1333 * driver has supplied a checksum for the entire packet into that
1334 * field. Return an error otherwise. This helper is intended to be
1335 * used in combination with **bpf_csum_diff**\ (), in particular
1336 * when the checksum needs to be updated after data has been
1337 * written into the packet through direct packet access.
1339 * The checksum on success, or a negative error code in case of
1342 * void bpf_set_hash_invalid(struct sk_buff *skb)
1344 * Invalidate the current *skb*\ **->hash**. It can be used after
1345 * mangling on headers through direct packet access, in order to
1346 * indicate that the hash is outdated and to trigger a
1347 * recalculation the next time the kernel tries to access this
1348 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1350 * int bpf_get_numa_node_id(void)
1352 * Return the id of the current NUMA node. The primary use case
1353 * for this helper is the selection of sockets for the local NUMA
1354 * node, when the program is attached to sockets using the
1355 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1356 * but the helper is also available to other eBPF program types,
1357 * similarly to **bpf_get_smp_processor_id**\ ().
1359 * The id of current NUMA node.
1361 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1363 * Grows headroom of packet associated to *skb* and adjusts the
1364 * offset of the MAC header accordingly, adding *len* bytes of
1365 * space. It automatically extends and reallocates memory as
1368 * This helper can be used on a layer 3 *skb* to push a MAC header
1369 * for redirection into a layer 2 device.
1371 * All values for *flags* are reserved for future usage, and must
1374 * A call to this helper is susceptible to change the underlying
1375 * packet buffer. Therefore, at load time, all checks on pointers
1376 * previously done by the verifier are invalidated and must be
1377 * performed again, if the helper is used in combination with
1378 * direct packet access.
1380 * 0 on success, or a negative error in case of failure.
1382 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1384 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1385 * it is possible to use a negative value for *delta*. This helper
1386 * can be used to prepare the packet for pushing or popping
1389 * A call to this helper is susceptible to change the underlying
1390 * packet buffer. Therefore, at load time, all checks on pointers
1391 * previously done by the verifier are invalidated and must be
1392 * performed again, if the helper is used in combination with
1393 * direct packet access.
1395 * 0 on success, or a negative error in case of failure.
1397 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1399 * Copy a NUL terminated string from an unsafe address
1400 * *unsafe_ptr* to *dst*. The *size* should include the
1401 * terminating NUL byte. In case the string length is smaller than
1402 * *size*, the target is not padded with further NUL bytes. If the
1403 * string length is larger than *size*, just *size*-1 bytes are
1404 * copied and the last byte is set to NUL.
1406 * On success, the length of the copied string is returned. This
1407 * makes this helper useful in tracing programs for reading
1408 * strings, and more importantly to get its length at runtime. See
1409 * the following snippet:
1413 * SEC("kprobe/sys_open")
1414 * void bpf_sys_open(struct pt_regs *ctx)
1416 * char buf[PATHLEN]; // PATHLEN is defined to 256
1417 * int res = bpf_probe_read_str(buf, sizeof(buf),
1420 * // Consume buf, for example push it to
1421 * // userspace via bpf_perf_event_output(); we
1422 * // can use res (the string length) as event
1423 * // size, after checking its boundaries.
1426 * In comparison, using **bpf_probe_read()** helper here instead
1427 * to read the string would require to estimate the length at
1428 * compile time, and would often result in copying more memory
1431 * Another useful use case is when parsing individual process
1432 * arguments or individual environment variables navigating
1433 * *current*\ **->mm->arg_start** and *current*\
1434 * **->mm->env_start**: using this helper and the return value,
1435 * one can quickly iterate at the right offset of the memory area.
1437 * On success, the strictly positive length of the string,
1438 * including the trailing NUL character. On error, a negative
1441 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1443 * If the **struct sk_buff** pointed by *skb* has a known socket,
1444 * retrieve the cookie (generated by the kernel) of this socket.
1445 * If no cookie has been set yet, generate a new cookie. Once
1446 * generated, the socket cookie remains stable for the life of the
1447 * socket. This helper can be useful for monitoring per socket
1448 * networking traffic statistics as it provides a unique socket
1449 * identifier per namespace.
1451 * A 8-byte long non-decreasing number on success, or 0 if the
1452 * socket field is missing inside *skb*.
1454 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1456 * Equivalent to bpf_get_socket_cookie() helper that accepts
1457 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1459 * A 8-byte long non-decreasing number.
1461 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1463 * Equivalent to bpf_get_socket_cookie() helper that accepts
1464 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1466 * A 8-byte long non-decreasing number.
1468 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1470 * The owner UID of the socket associated to *skb*. If the socket
1471 * is **NULL**, or if it is not a full socket (i.e. if it is a
1472 * time-wait or a request socket instead), **overflowuid** value
1473 * is returned (note that **overflowuid** might also be the actual
1474 * UID value for the socket).
1476 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1478 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1483 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1485 * Emulate a call to **setsockopt()** on the socket associated to
1486 * *bpf_socket*, which must be a full socket. The *level* at
1487 * which the option resides and the name *optname* of the option
1488 * must be specified, see **setsockopt(2)** for more information.
1489 * The option value of length *optlen* is pointed by *optval*.
1491 * This helper actually implements a subset of **setsockopt()**.
1492 * It supports the following *level*\ s:
1494 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1495 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1496 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1497 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1498 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1499 * **TCP_BPF_SNDCWND_CLAMP**.
1500 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1501 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1503 * 0 on success, or a negative error in case of failure.
1505 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1507 * Grow or shrink the room for data in the packet associated to
1508 * *skb* by *len_diff*, and according to the selected *mode*.
1510 * There are two supported modes at this time:
1512 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1513 * (room space is added or removed below the layer 2 header).
1515 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1516 * (room space is added or removed below the layer 3 header).
1518 * The following flags are supported at this time:
1520 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1521 * Adjusting mss in this way is not allowed for datagrams.
1523 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1524 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1525 * Any new space is reserved to hold a tunnel header.
1526 * Configure skb offsets and other fields accordingly.
1528 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1529 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1530 * Use with ENCAP_L3 flags to further specify the tunnel type.
1532 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1533 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1534 * type; *len* is the length of the inner MAC header.
1536 * A call to this helper is susceptible to change the underlying
1537 * packet buffer. Therefore, at load time, all checks on pointers
1538 * previously done by the verifier are invalidated and must be
1539 * performed again, if the helper is used in combination with
1540 * direct packet access.
1542 * 0 on success, or a negative error in case of failure.
1544 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1546 * Redirect the packet to the endpoint referenced by *map* at
1547 * index *key*. Depending on its type, this *map* can contain
1548 * references to net devices (for forwarding packets through other
1549 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1550 * but this is only implemented for native XDP (with driver
1551 * support) as of this writing).
1553 * All values for *flags* are reserved for future usage, and must
1556 * When used to redirect packets to net devices, this helper
1557 * provides a high performance increase over **bpf_redirect**\ ().
1558 * This is due to various implementation details of the underlying
1559 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1560 * () tries to send packet as a "bulk" to the device.
1562 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1564 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1566 * Redirect the packet to the socket referenced by *map* (of type
1567 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1568 * egress interfaces can be used for redirection. The
1569 * **BPF_F_INGRESS** value in *flags* is used to make the
1570 * distinction (ingress path is selected if the flag is present,
1571 * egress path otherwise). This is the only flag supported for now.
1573 * **SK_PASS** on success, or **SK_DROP** on error.
1575 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1577 * Add an entry to, or update a *map* referencing sockets. The
1578 * *skops* is used as a new value for the entry associated to
1579 * *key*. *flags* is one of:
1582 * The entry for *key* must not exist in the map.
1584 * The entry for *key* must already exist in the map.
1586 * No condition on the existence of the entry for *key*.
1588 * If the *map* has eBPF programs (parser and verdict), those will
1589 * be inherited by the socket being added. If the socket is
1590 * already attached to eBPF programs, this results in an error.
1592 * 0 on success, or a negative error in case of failure.
1594 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1596 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1597 * *delta* (which can be positive or negative). Note that this
1598 * operation modifies the address stored in *xdp_md*\ **->data**,
1599 * so the latter must be loaded only after the helper has been
1602 * The use of *xdp_md*\ **->data_meta** is optional and programs
1603 * are not required to use it. The rationale is that when the
1604 * packet is processed with XDP (e.g. as DoS filter), it is
1605 * possible to push further meta data along with it before passing
1606 * to the stack, and to give the guarantee that an ingress eBPF
1607 * program attached as a TC classifier on the same device can pick
1608 * this up for further post-processing. Since TC works with socket
1609 * buffers, it remains possible to set from XDP the **mark** or
1610 * **priority** pointers, or other pointers for the socket buffer.
1611 * Having this scratch space generic and programmable allows for
1612 * more flexibility as the user is free to store whatever meta
1615 * A call to this helper is susceptible to change the underlying
1616 * packet buffer. Therefore, at load time, all checks on pointers
1617 * previously done by the verifier are invalidated and must be
1618 * performed again, if the helper is used in combination with
1619 * direct packet access.
1621 * 0 on success, or a negative error in case of failure.
1623 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1625 * Read the value of a perf event counter, and store it into *buf*
1626 * of size *buf_size*. This helper relies on a *map* of type
1627 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1628 * counter is selected when *map* is updated with perf event file
1629 * descriptors. The *map* is an array whose size is the number of
1630 * available CPUs, and each cell contains a value relative to one
1631 * CPU. The value to retrieve is indicated by *flags*, that
1632 * contains the index of the CPU to look up, masked with
1633 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1634 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1635 * current CPU should be retrieved.
1637 * This helper behaves in a way close to
1638 * **bpf_perf_event_read**\ () helper, save that instead of
1639 * just returning the value observed, it fills the *buf*
1640 * structure. This allows for additional data to be retrieved: in
1641 * particular, the enabled and running times (in *buf*\
1642 * **->enabled** and *buf*\ **->running**, respectively) are
1643 * copied. In general, **bpf_perf_event_read_value**\ () is
1644 * recommended over **bpf_perf_event_read**\ (), which has some
1645 * ABI issues and provides fewer functionalities.
1647 * These values are interesting, because hardware PMU (Performance
1648 * Monitoring Unit) counters are limited resources. When there are
1649 * more PMU based perf events opened than available counters,
1650 * kernel will multiplex these events so each event gets certain
1651 * percentage (but not all) of the PMU time. In case that
1652 * multiplexing happens, the number of samples or counter value
1653 * will not reflect the case compared to when no multiplexing
1654 * occurs. This makes comparison between different runs difficult.
1655 * Typically, the counter value should be normalized before
1656 * comparing to other experiments. The usual normalization is done
1661 * normalized_counter = counter * t_enabled / t_running
1663 * Where t_enabled is the time enabled for event and t_running is
1664 * the time running for event since last normalization. The
1665 * enabled and running times are accumulated since the perf event
1666 * open. To achieve scaling factor between two invocations of an
1667 * eBPF program, users can can use CPU id as the key (which is
1668 * typical for perf array usage model) to remember the previous
1669 * value and do the calculation inside the eBPF program.
1671 * 0 on success, or a negative error in case of failure.
1673 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1675 * For en eBPF program attached to a perf event, retrieve the
1676 * value of the event counter associated to *ctx* and store it in
1677 * the structure pointed by *buf* and of size *buf_size*. Enabled
1678 * and running times are also stored in the structure (see
1679 * description of helper **bpf_perf_event_read_value**\ () for
1682 * 0 on success, or a negative error in case of failure.
1684 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1686 * Emulate a call to **getsockopt()** on the socket associated to
1687 * *bpf_socket*, which must be a full socket. The *level* at
1688 * which the option resides and the name *optname* of the option
1689 * must be specified, see **getsockopt(2)** for more information.
1690 * The retrieved value is stored in the structure pointed by
1691 * *opval* and of length *optlen*.
1693 * This helper actually implements a subset of **getsockopt()**.
1694 * It supports the following *level*\ s:
1696 * * **IPPROTO_TCP**, which supports *optname*
1697 * **TCP_CONGESTION**.
1698 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1699 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1701 * 0 on success, or a negative error in case of failure.
1703 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1705 * Used for error injection, this helper uses kprobes to override
1706 * the return value of the probed function, and to set it to *rc*.
1707 * The first argument is the context *regs* on which the kprobe
1710 * This helper works by setting setting the PC (program counter)
1711 * to an override function which is run in place of the original
1712 * probed function. This means the probed function is not run at
1713 * all. The replacement function just returns with the required
1716 * This helper has security implications, and thus is subject to
1717 * restrictions. It is only available if the kernel was compiled
1718 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1719 * option, and in this case it only works on functions tagged with
1720 * **ALLOW_ERROR_INJECTION** in the kernel code.
1722 * Also, the helper is only available for the architectures having
1723 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1724 * x86 architecture is the only one to support this feature.
1728 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1730 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1731 * for the full TCP socket associated to *bpf_sock_ops* to
1734 * The primary use of this field is to determine if there should
1735 * be calls to eBPF programs of type
1736 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1737 * code. A program of the same type can change its value, per
1738 * connection and as necessary, when the connection is
1739 * established. This field is directly accessible for reading, but
1740 * this helper must be used for updates in order to return an
1741 * error if an eBPF program tries to set a callback that is not
1742 * supported in the current kernel.
1744 * *argval* is a flag array which can combine these flags:
1746 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1747 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1748 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1750 * Therefore, this function can be used to clear a callback flag by
1751 * setting the appropriate bit to zero. e.g. to disable the RTO
1754 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1755 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1757 * Here are some examples of where one could call such eBPF
1761 * * When a packet is retransmitted.
1762 * * When the connection terminates.
1763 * * When a packet is sent.
1764 * * When a packet is received.
1766 * Code **-EINVAL** if the socket is not a full TCP socket;
1767 * otherwise, a positive number containing the bits that could not
1768 * be set is returned (which comes down to 0 if all bits were set
1771 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1773 * This helper is used in programs implementing policies at the
1774 * socket level. If the message *msg* is allowed to pass (i.e. if
1775 * the verdict eBPF program returns **SK_PASS**), redirect it to
1776 * the socket referenced by *map* (of type
1777 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1778 * egress interfaces can be used for redirection. The
1779 * **BPF_F_INGRESS** value in *flags* is used to make the
1780 * distinction (ingress path is selected if the flag is present,
1781 * egress path otherwise). This is the only flag supported for now.
1783 * **SK_PASS** on success, or **SK_DROP** on error.
1785 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1787 * For socket policies, apply the verdict of the eBPF program to
1788 * the next *bytes* (number of bytes) of message *msg*.
1790 * For example, this helper can be used in the following cases:
1792 * * A single **sendmsg**\ () or **sendfile**\ () system call
1793 * contains multiple logical messages that the eBPF program is
1794 * supposed to read and for which it should apply a verdict.
1795 * * An eBPF program only cares to read the first *bytes* of a
1796 * *msg*. If the message has a large payload, then setting up
1797 * and calling the eBPF program repeatedly for all bytes, even
1798 * though the verdict is already known, would create unnecessary
1801 * When called from within an eBPF program, the helper sets a
1802 * counter internal to the BPF infrastructure, that is used to
1803 * apply the last verdict to the next *bytes*. If *bytes* is
1804 * smaller than the current data being processed from a
1805 * **sendmsg**\ () or **sendfile**\ () system call, the first
1806 * *bytes* will be sent and the eBPF program will be re-run with
1807 * the pointer for start of data pointing to byte number *bytes*
1808 * **+ 1**. If *bytes* is larger than the current data being
1809 * processed, then the eBPF verdict will be applied to multiple
1810 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1813 * Note that if a socket closes with the internal counter holding
1814 * a non-zero value, this is not a problem because data is not
1815 * being buffered for *bytes* and is sent as it is received.
1819 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1821 * For socket policies, prevent the execution of the verdict eBPF
1822 * program for message *msg* until *bytes* (byte number) have been
1825 * This can be used when one needs a specific number of bytes
1826 * before a verdict can be assigned, even if the data spans
1827 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1828 * case would be a user calling **sendmsg**\ () repeatedly with
1829 * 1-byte long message segments. Obviously, this is bad for
1830 * performance, but it is still valid. If the eBPF program needs
1831 * *bytes* bytes to validate a header, this helper can be used to
1832 * prevent the eBPF program to be called again until *bytes* have
1837 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1839 * For socket policies, pull in non-linear data from user space
1840 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1841 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1844 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1845 * *msg* it can only parse data that the (**data**, **data_end**)
1846 * pointers have already consumed. For **sendmsg**\ () hooks this
1847 * is likely the first scatterlist element. But for calls relying
1848 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1849 * be the range (**0**, **0**) because the data is shared with
1850 * user space and by default the objective is to avoid allowing
1851 * user space to modify data while (or after) eBPF verdict is
1852 * being decided. This helper can be used to pull in data and to
1853 * set the start and end pointer to given values. Data will be
1854 * copied if necessary (i.e. if data was not linear and if start
1855 * and end pointers do not point to the same chunk).
1857 * A call to this helper is susceptible to change the underlying
1858 * packet buffer. Therefore, at load time, all checks on pointers
1859 * previously done by the verifier are invalidated and must be
1860 * performed again, if the helper is used in combination with
1861 * direct packet access.
1863 * All values for *flags* are reserved for future usage, and must
1866 * 0 on success, or a negative error in case of failure.
1868 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1870 * Bind the socket associated to *ctx* to the address pointed by
1871 * *addr*, of length *addr_len*. This allows for making outgoing
1872 * connection from the desired IP address, which can be useful for
1873 * example when all processes inside a cgroup should use one
1874 * single IP address on a host that has multiple IP configured.
1876 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1877 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1878 * **AF_INET6**). Looking for a free port to bind to can be
1879 * expensive, therefore binding to port is not permitted by the
1880 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1881 * must be set to zero.
1883 * 0 on success, or a negative error in case of failure.
1885 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1887 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1888 * only possible to shrink the packet as of this writing,
1889 * therefore *delta* must be a negative integer.
1891 * A call to this helper is susceptible to change the underlying
1892 * packet buffer. Therefore, at load time, all checks on pointers
1893 * previously done by the verifier are invalidated and must be
1894 * performed again, if the helper is used in combination with
1895 * direct packet access.
1897 * 0 on success, or a negative error in case of failure.
1899 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1901 * Retrieve the XFRM state (IP transform framework, see also
1902 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1904 * The retrieved value is stored in the **struct bpf_xfrm_state**
1905 * pointed by *xfrm_state* and of length *size*.
1907 * All values for *flags* are reserved for future usage, and must
1910 * This helper is available only if the kernel was compiled with
1911 * **CONFIG_XFRM** configuration option.
1913 * 0 on success, or a negative error in case of failure.
1915 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1917 * Return a user or a kernel stack in bpf program provided buffer.
1918 * To achieve this, the helper needs *ctx*, which is a pointer
1919 * to the context on which the tracing program is executed.
1920 * To store the stacktrace, the bpf program provides *buf* with
1921 * a nonnegative *size*.
1923 * The last argument, *flags*, holds the number of stack frames to
1924 * skip (from 0 to 255), masked with
1925 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1926 * the following flags:
1928 * **BPF_F_USER_STACK**
1929 * Collect a user space stack instead of a kernel stack.
1930 * **BPF_F_USER_BUILD_ID**
1931 * Collect buildid+offset instead of ips for user stack,
1932 * only valid if **BPF_F_USER_STACK** is also specified.
1934 * **bpf_get_stack**\ () can collect up to
1935 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1936 * to sufficient large buffer size. Note that
1937 * this limit can be controlled with the **sysctl** program, and
1938 * that it should be manually increased in order to profile long
1939 * user stacks (such as stacks for Java programs). To do so, use:
1943 * # sysctl kernel.perf_event_max_stack=<new value>
1945 * A non-negative value equal to or less than *size* on success,
1946 * or a negative error in case of failure.
1948 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1950 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1951 * it provides an easy way to load *len* bytes from *offset*
1952 * from the packet associated to *skb*, into the buffer pointed
1953 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1954 * a fifth argument *start_header* exists in order to select a
1955 * base offset to start from. *start_header* can be one of:
1957 * **BPF_HDR_START_MAC**
1958 * Base offset to load data from is *skb*'s mac header.
1959 * **BPF_HDR_START_NET**
1960 * Base offset to load data from is *skb*'s network header.
1962 * In general, "direct packet access" is the preferred method to
1963 * access packet data, however, this helper is in particular useful
1964 * in socket filters where *skb*\ **->data** does not always point
1965 * to the start of the mac header and where "direct packet access"
1968 * 0 on success, or a negative error in case of failure.
1970 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1972 * Do FIB lookup in kernel tables using parameters in *params*.
1973 * If lookup is successful and result shows packet is to be
1974 * forwarded, the neighbor tables are searched for the nexthop.
1975 * If successful (ie., FIB lookup shows forwarding and nexthop
1976 * is resolved), the nexthop address is returned in ipv4_dst
1977 * or ipv6_dst based on family, smac is set to mac address of
1978 * egress device, dmac is set to nexthop mac address, rt_metric
1979 * is set to metric from route (IPv4/IPv6 only), and ifindex
1980 * is set to the device index of the nexthop from the FIB lookup.
1982 * *plen* argument is the size of the passed in struct.
1983 * *flags* argument can be a combination of one or more of the
1986 * **BPF_FIB_LOOKUP_DIRECT**
1987 * Do a direct table lookup vs full lookup using FIB
1989 * **BPF_FIB_LOOKUP_OUTPUT**
1990 * Perform lookup from an egress perspective (default is
1993 * *ctx* is either **struct xdp_md** for XDP programs or
1994 * **struct sk_buff** tc cls_act programs.
1996 * * < 0 if any input argument is invalid
1997 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1998 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1999 * packet is not forwarded or needs assist from full stack
2001 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
2003 * Add an entry to, or update a sockhash *map* referencing sockets.
2004 * The *skops* is used as a new value for the entry associated to
2005 * *key*. *flags* is one of:
2008 * The entry for *key* must not exist in the map.
2010 * The entry for *key* must already exist in the map.
2012 * No condition on the existence of the entry for *key*.
2014 * If the *map* has eBPF programs (parser and verdict), those will
2015 * be inherited by the socket being added. If the socket is
2016 * already attached to eBPF programs, this results in an error.
2018 * 0 on success, or a negative error in case of failure.
2020 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2022 * This helper is used in programs implementing policies at the
2023 * socket level. If the message *msg* is allowed to pass (i.e. if
2024 * the verdict eBPF program returns **SK_PASS**), redirect it to
2025 * the socket referenced by *map* (of type
2026 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2027 * egress interfaces can be used for redirection. The
2028 * **BPF_F_INGRESS** value in *flags* is used to make the
2029 * distinction (ingress path is selected if the flag is present,
2030 * egress path otherwise). This is the only flag supported for now.
2032 * **SK_PASS** on success, or **SK_DROP** on error.
2034 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2036 * This helper is used in programs implementing policies at the
2037 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2038 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2039 * to the socket referenced by *map* (of type
2040 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2041 * egress interfaces can be used for redirection. The
2042 * **BPF_F_INGRESS** value in *flags* is used to make the
2043 * distinction (ingress path is selected if the flag is present,
2044 * egress otherwise). This is the only flag supported for now.
2046 * **SK_PASS** on success, or **SK_DROP** on error.
2048 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2050 * Encapsulate the packet associated to *skb* within a Layer 3
2051 * protocol header. This header is provided in the buffer at
2052 * address *hdr*, with *len* its size in bytes. *type* indicates
2053 * the protocol of the header and can be one of:
2055 * **BPF_LWT_ENCAP_SEG6**
2056 * IPv6 encapsulation with Segment Routing Header
2057 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2058 * the IPv6 header is computed by the kernel.
2059 * **BPF_LWT_ENCAP_SEG6_INLINE**
2060 * Only works if *skb* contains an IPv6 packet. Insert a
2061 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2063 * **BPF_LWT_ENCAP_IP**
2064 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2065 * must be IPv4 or IPv6, followed by zero or more
2066 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2067 * total bytes in all prepended headers. Please note that
2068 * if **skb_is_gso**\ (*skb*) is true, no more than two
2069 * headers can be prepended, and the inner header, if
2070 * present, should be either GRE or UDP/GUE.
2072 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2073 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2074 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2075 * **BPF_PROG_TYPE_LWT_XMIT**.
2077 * A call to this helper is susceptible to change the underlying
2078 * packet buffer. Therefore, at load time, all checks on pointers
2079 * previously done by the verifier are invalidated and must be
2080 * performed again, if the helper is used in combination with
2081 * direct packet access.
2083 * 0 on success, or a negative error in case of failure.
2085 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2087 * Store *len* bytes from address *from* into the packet
2088 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2089 * inside the outermost IPv6 Segment Routing Header can be
2090 * modified through this helper.
2092 * A call to this helper is susceptible to change the underlying
2093 * packet buffer. Therefore, at load time, all checks on pointers
2094 * previously done by the verifier are invalidated and must be
2095 * performed again, if the helper is used in combination with
2096 * direct packet access.
2098 * 0 on success, or a negative error in case of failure.
2100 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2102 * Adjust the size allocated to TLVs in the outermost IPv6
2103 * Segment Routing Header contained in the packet associated to
2104 * *skb*, at position *offset* by *delta* bytes. Only offsets
2105 * after the segments are accepted. *delta* can be as well
2106 * positive (growing) as negative (shrinking).
2108 * A call to this helper is susceptible to change the underlying
2109 * packet buffer. Therefore, at load time, all checks on pointers
2110 * previously done by the verifier are invalidated and must be
2111 * performed again, if the helper is used in combination with
2112 * direct packet access.
2114 * 0 on success, or a negative error in case of failure.
2116 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2118 * Apply an IPv6 Segment Routing action of type *action* to the
2119 * packet associated to *skb*. Each action takes a parameter
2120 * contained at address *param*, and of length *param_len* bytes.
2121 * *action* can be one of:
2123 * **SEG6_LOCAL_ACTION_END_X**
2124 * End.X action: Endpoint with Layer-3 cross-connect.
2125 * Type of *param*: **struct in6_addr**.
2126 * **SEG6_LOCAL_ACTION_END_T**
2127 * End.T action: Endpoint with specific IPv6 table lookup.
2128 * Type of *param*: **int**.
2129 * **SEG6_LOCAL_ACTION_END_B6**
2130 * End.B6 action: Endpoint bound to an SRv6 policy.
2131 * Type of *param*: **struct ipv6_sr_hdr**.
2132 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2133 * End.B6.Encap action: Endpoint bound to an SRv6
2134 * encapsulation policy.
2135 * Type of *param*: **struct ipv6_sr_hdr**.
2137 * A call to this helper is susceptible to change the underlying
2138 * packet buffer. Therefore, at load time, all checks on pointers
2139 * previously done by the verifier are invalidated and must be
2140 * performed again, if the helper is used in combination with
2141 * direct packet access.
2143 * 0 on success, or a negative error in case of failure.
2145 * int bpf_rc_repeat(void *ctx)
2147 * This helper is used in programs implementing IR decoding, to
2148 * report a successfully decoded repeat key message. This delays
2149 * the generation of a key up event for previously generated
2152 * Some IR protocols like NEC have a special IR message for
2153 * repeating last button, for when a button is held down.
2155 * The *ctx* should point to the lirc sample as passed into
2158 * This helper is only available is the kernel was compiled with
2159 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2164 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2166 * This helper is used in programs implementing IR decoding, to
2167 * report a successfully decoded key press with *scancode*,
2168 * *toggle* value in the given *protocol*. The scancode will be
2169 * translated to a keycode using the rc keymap, and reported as
2170 * an input key down event. After a period a key up event is
2171 * generated. This period can be extended by calling either
2172 * **bpf_rc_keydown**\ () again with the same values, or calling
2173 * **bpf_rc_repeat**\ ().
2175 * Some protocols include a toggle bit, in case the button was
2176 * released and pressed again between consecutive scancodes.
2178 * The *ctx* should point to the lirc sample as passed into
2181 * The *protocol* is the decoded protocol number (see
2182 * **enum rc_proto** for some predefined values).
2184 * This helper is only available is the kernel was compiled with
2185 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2190 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2192 * Return the cgroup v2 id of the socket associated with the *skb*.
2193 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2194 * helper for cgroup v1 by providing a tag resp. identifier that
2195 * can be matched on or used for map lookups e.g. to implement
2196 * policy. The cgroup v2 id of a given path in the hierarchy is
2197 * exposed in user space through the f_handle API in order to get
2198 * to the same 64-bit id.
2200 * This helper can be used on TC egress path, but not on ingress,
2201 * and is available only if the kernel was compiled with the
2202 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2204 * The id is returned or 0 in case the id could not be retrieved.
2206 * u64 bpf_get_current_cgroup_id(void)
2208 * A 64-bit integer containing the current cgroup id based
2209 * on the cgroup within which the current task is running.
2211 * void *bpf_get_local_storage(void *map, u64 flags)
2213 * Get the pointer to the local storage area.
2214 * The type and the size of the local storage is defined
2215 * by the *map* argument.
2216 * The *flags* meaning is specific for each map type,
2217 * and has to be 0 for cgroup local storage.
2219 * Depending on the BPF program type, a local storage area
2220 * can be shared between multiple instances of the BPF program,
2221 * running simultaneously.
2223 * A user should care about the synchronization by himself.
2224 * For example, by using the **BPF_STX_XADD** instruction to alter
2227 * A pointer to the local storage area.
2229 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2231 * Select a **SO_REUSEPORT** socket from a
2232 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2233 * It checks the selected socket is matching the incoming
2234 * request in the socket buffer.
2236 * 0 on success, or a negative error in case of failure.
2238 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2240 * Return id of cgroup v2 that is ancestor of cgroup associated
2241 * with the *skb* at the *ancestor_level*. The root cgroup is at
2242 * *ancestor_level* zero and each step down the hierarchy
2243 * increments the level. If *ancestor_level* == level of cgroup
2244 * associated with *skb*, then return value will be same as that
2245 * of **bpf_skb_cgroup_id**\ ().
2247 * The helper is useful to implement policies based on cgroups
2248 * that are upper in hierarchy than immediate cgroup associated
2251 * The format of returned id and helper limitations are same as in
2252 * **bpf_skb_cgroup_id**\ ().
2254 * The id is returned or 0 in case the id could not be retrieved.
2256 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2258 * Look for TCP socket matching *tuple*, optionally in a child
2259 * network namespace *netns*. The return value must be checked,
2260 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2262 * The *ctx* should point to the context of the program, such as
2263 * the skb or socket (depending on the hook in use). This is used
2264 * to determine the base network namespace for the lookup.
2266 * *tuple_size* must be one of:
2268 * **sizeof**\ (*tuple*\ **->ipv4**)
2269 * Look for an IPv4 socket.
2270 * **sizeof**\ (*tuple*\ **->ipv6**)
2271 * Look for an IPv6 socket.
2273 * If the *netns* is a negative signed 32-bit integer, then the
2274 * socket lookup table in the netns associated with the *ctx* will
2275 * will be used. For the TC hooks, this is the netns of the device
2276 * in the skb. For socket hooks, this is the netns of the socket.
2277 * If *netns* is any other signed 32-bit value greater than or
2278 * equal to zero then it specifies the ID of the netns relative to
2279 * the netns associated with the *ctx*. *netns* values beyond the
2280 * range of 32-bit integers are reserved for future use.
2282 * All values for *flags* are reserved for future usage, and must
2285 * This helper is available only if the kernel was compiled with
2286 * **CONFIG_NET** configuration option.
2288 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2289 * For sockets with reuseport option, the **struct bpf_sock**
2290 * result is from *reuse*\ **->socks**\ [] using the hash of the
2293 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2295 * Look for UDP socket matching *tuple*, optionally in a child
2296 * network namespace *netns*. The return value must be checked,
2297 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2299 * The *ctx* should point to the context of the program, such as
2300 * the skb or socket (depending on the hook in use). This is used
2301 * to determine the base network namespace for the lookup.
2303 * *tuple_size* must be one of:
2305 * **sizeof**\ (*tuple*\ **->ipv4**)
2306 * Look for an IPv4 socket.
2307 * **sizeof**\ (*tuple*\ **->ipv6**)
2308 * Look for an IPv6 socket.
2310 * If the *netns* is a negative signed 32-bit integer, then the
2311 * socket lookup table in the netns associated with the *ctx* will
2312 * will be used. For the TC hooks, this is the netns of the device
2313 * in the skb. For socket hooks, this is the netns of the socket.
2314 * If *netns* is any other signed 32-bit value greater than or
2315 * equal to zero then it specifies the ID of the netns relative to
2316 * the netns associated with the *ctx*. *netns* values beyond the
2317 * range of 32-bit integers are reserved for future use.
2319 * All values for *flags* are reserved for future usage, and must
2322 * This helper is available only if the kernel was compiled with
2323 * **CONFIG_NET** configuration option.
2325 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2326 * For sockets with reuseport option, the **struct bpf_sock**
2327 * result is from *reuse*\ **->socks**\ [] using the hash of the
2330 * int bpf_sk_release(struct bpf_sock *sock)
2332 * Release the reference held by *sock*. *sock* must be a
2333 * non-**NULL** pointer that was returned from
2334 * **bpf_sk_lookup_xxx**\ ().
2336 * 0 on success, or a negative error in case of failure.
2338 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2340 * Push an element *value* in *map*. *flags* is one of:
2343 * If the queue/stack is full, the oldest element is
2344 * removed to make room for this.
2346 * 0 on success, or a negative error in case of failure.
2348 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2350 * Pop an element from *map*.
2352 * 0 on success, or a negative error in case of failure.
2354 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2356 * Get an element from *map* without removing it.
2358 * 0 on success, or a negative error in case of failure.
2360 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2362 * For socket policies, insert *len* bytes into *msg* at offset
2365 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2366 * *msg* it may want to insert metadata or options into the *msg*.
2367 * This can later be read and used by any of the lower layer BPF
2370 * This helper may fail if under memory pressure (a malloc
2371 * fails) in these cases BPF programs will get an appropriate
2372 * error and BPF programs will need to handle them.
2374 * 0 on success, or a negative error in case of failure.
2376 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2378 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2379 * This may result in **ENOMEM** errors under certain situations if
2380 * an allocation and copy are required due to a full ring buffer.
2381 * However, the helper will try to avoid doing the allocation
2382 * if possible. Other errors can occur if input parameters are
2383 * invalid either due to *start* byte not being valid part of *msg*
2384 * payload and/or *pop* value being to large.
2386 * 0 on success, or a negative error in case of failure.
2388 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2390 * This helper is used in programs implementing IR decoding, to
2391 * report a successfully decoded pointer movement.
2393 * The *ctx* should point to the lirc sample as passed into
2396 * This helper is only available is the kernel was compiled with
2397 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2402 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2404 * Acquire a spinlock represented by the pointer *lock*, which is
2405 * stored as part of a value of a map. Taking the lock allows to
2406 * safely update the rest of the fields in that value. The
2407 * spinlock can (and must) later be released with a call to
2408 * **bpf_spin_unlock**\ (\ *lock*\ ).
2410 * Spinlocks in BPF programs come with a number of restrictions
2413 * * **bpf_spin_lock** objects are only allowed inside maps of
2414 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2415 * list could be extended in the future).
2416 * * BTF description of the map is mandatory.
2417 * * The BPF program can take ONE lock at a time, since taking two
2418 * or more could cause dead locks.
2419 * * Only one **struct bpf_spin_lock** is allowed per map element.
2420 * * When the lock is taken, calls (either BPF to BPF or helpers)
2422 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2423 * allowed inside a spinlock-ed region.
2424 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2425 * the lock, on all execution paths, before it returns.
2426 * * The BPF program can access **struct bpf_spin_lock** only via
2427 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2428 * helpers. Loading or storing data into the **struct
2429 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2430 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2431 * of the map value must be a struct and have **struct
2432 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2433 * Nested lock inside another struct is not allowed.
2434 * * The **struct bpf_spin_lock** *lock* field in a map value must
2435 * be aligned on a multiple of 4 bytes in that value.
2436 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2437 * the **bpf_spin_lock** field to user space.
2438 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2439 * a BPF program, do not update the **bpf_spin_lock** field.
2440 * * **bpf_spin_lock** cannot be on the stack or inside a
2441 * networking packet (it can only be inside of a map values).
2442 * * **bpf_spin_lock** is available to root only.
2443 * * Tracing programs and socket filter programs cannot use
2444 * **bpf_spin_lock**\ () due to insufficient preemption checks
2445 * (but this may change in the future).
2446 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2450 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2452 * Release the *lock* previously locked by a call to
2453 * **bpf_spin_lock**\ (\ *lock*\ ).
2457 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2459 * This helper gets a **struct bpf_sock** pointer such
2460 * that all the fields in this **bpf_sock** can be accessed.
2462 * A **struct bpf_sock** pointer on success, or **NULL** in
2465 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2467 * This helper gets a **struct bpf_tcp_sock** pointer from a
2468 * **struct bpf_sock** pointer.
2470 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2473 * int bpf_skb_ecn_set_ce(struct sk_buf *skb)
2475 * Set ECN (Explicit Congestion Notification) field of IP header
2476 * to **CE** (Congestion Encountered) if current value is **ECT**
2477 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2480 * 1 if the **CE** flag is set (either by the current helper call
2481 * or because it was already present), 0 if it is not set.
2483 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2485 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2486 * **bpf_sk_release**\ () is unnecessary and not allowed.
2488 * A **struct bpf_sock** pointer on success, or **NULL** in
2491 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2493 * Look for TCP socket matching *tuple*, optionally in a child
2494 * network namespace *netns*. The return value must be checked,
2495 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2497 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2498 * that it also returns timewait or request sockets. Use
2499 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2502 * This helper is available only if the kernel was compiled with
2503 * **CONFIG_NET** configuration option.
2505 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2506 * For sockets with reuseport option, the **struct bpf_sock**
2507 * result is from *reuse*\ **->socks**\ [] using the hash of the
2510 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2512 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2513 * the listening socket in *sk*.
2515 * *iph* points to the start of the IPv4 or IPv6 header, while
2516 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2517 * **sizeof**\ (**struct ip6hdr**).
2519 * *th* points to the start of the TCP header, while *th_len*
2520 * contains **sizeof**\ (**struct tcphdr**).
2523 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2526 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2528 * Get name of sysctl in /proc/sys/ and copy it into provided by
2529 * program buffer *buf* of size *buf_len*.
2531 * The buffer is always NUL terminated, unless it's zero-sized.
2533 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2534 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2535 * only (e.g. "tcp_mem").
2537 * Number of character copied (not including the trailing NUL).
2539 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2540 * truncated name in this case).
2542 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2544 * Get current value of sysctl as it is presented in /proc/sys
2545 * (incl. newline, etc), and copy it as a string into provided
2546 * by program buffer *buf* of size *buf_len*.
2548 * The whole value is copied, no matter what file position user
2549 * space issued e.g. sys_read at.
2551 * The buffer is always NUL terminated, unless it's zero-sized.
2553 * Number of character copied (not including the trailing NUL).
2555 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2556 * truncated name in this case).
2558 * **-EINVAL** if current value was unavailable, e.g. because
2559 * sysctl is uninitialized and read returns -EIO for it.
2561 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2563 * Get new value being written by user space to sysctl (before
2564 * the actual write happens) and copy it as a string into
2565 * provided by program buffer *buf* of size *buf_len*.
2567 * User space may write new value at file position > 0.
2569 * The buffer is always NUL terminated, unless it's zero-sized.
2571 * Number of character copied (not including the trailing NUL).
2573 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2574 * truncated name in this case).
2576 * **-EINVAL** if sysctl is being read.
2578 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2580 * Override new value being written by user space to sysctl with
2581 * value provided by program in buffer *buf* of size *buf_len*.
2583 * *buf* should contain a string in same form as provided by user
2584 * space on sysctl write.
2586 * User space may write new value at file position > 0. To override
2587 * the whole sysctl value file position should be set to zero.
2591 * **-E2BIG** if the *buf_len* is too big.
2593 * **-EINVAL** if sysctl is being read.
2595 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2597 * Convert the initial part of the string from buffer *buf* of
2598 * size *buf_len* to a long integer according to the given base
2599 * and save the result in *res*.
2601 * The string may begin with an arbitrary amount of white space
2602 * (as determined by **isspace**\ (3)) followed by a single
2603 * optional '**-**' sign.
2605 * Five least significant bits of *flags* encode base, other bits
2606 * are currently unused.
2608 * Base must be either 8, 10, 16 or 0 to detect it automatically
2609 * similar to user space **strtol**\ (3).
2611 * Number of characters consumed on success. Must be positive but
2612 * no more than *buf_len*.
2614 * **-EINVAL** if no valid digits were found or unsupported base
2617 * **-ERANGE** if resulting value was out of range.
2619 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2621 * Convert the initial part of the string from buffer *buf* of
2622 * size *buf_len* to an unsigned long integer according to the
2623 * given base and save the result in *res*.
2625 * The string may begin with an arbitrary amount of white space
2626 * (as determined by **isspace**\ (3)).
2628 * Five least significant bits of *flags* encode base, other bits
2629 * are currently unused.
2631 * Base must be either 8, 10, 16 or 0 to detect it automatically
2632 * similar to user space **strtoul**\ (3).
2634 * Number of characters consumed on success. Must be positive but
2635 * no more than *buf_len*.
2637 * **-EINVAL** if no valid digits were found or unsupported base
2640 * **-ERANGE** if resulting value was out of range.
2642 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2644 * Get a bpf-local-storage from a *sk*.
2646 * Logically, it could be thought of getting the value from
2647 * a *map* with *sk* as the **key**. From this
2648 * perspective, the usage is not much different from
2649 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2650 * helper enforces the key must be a full socket and the map must
2651 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2653 * Underneath, the value is stored locally at *sk* instead of
2654 * the *map*. The *map* is used as the bpf-local-storage
2655 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2656 * searched against all bpf-local-storages residing at *sk*.
2658 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2659 * used such that a new bpf-local-storage will be
2660 * created if one does not exist. *value* can be used
2661 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2662 * the initial value of a bpf-local-storage. If *value* is
2663 * **NULL**, the new bpf-local-storage will be zero initialized.
2665 * A bpf-local-storage pointer is returned on success.
2667 * **NULL** if not found or there was an error in adding
2668 * a new bpf-local-storage.
2670 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2672 * Delete a bpf-local-storage from a *sk*.
2676 * **-ENOENT** if the bpf-local-storage cannot be found.
2678 #define __BPF_FUNC_MAPPER(FN) \
2680 FN(map_lookup_elem), \
2681 FN(map_update_elem), \
2682 FN(map_delete_elem), \
2686 FN(get_prandom_u32), \
2687 FN(get_smp_processor_id), \
2688 FN(skb_store_bytes), \
2689 FN(l3_csum_replace), \
2690 FN(l4_csum_replace), \
2692 FN(clone_redirect), \
2693 FN(get_current_pid_tgid), \
2694 FN(get_current_uid_gid), \
2695 FN(get_current_comm), \
2696 FN(get_cgroup_classid), \
2697 FN(skb_vlan_push), \
2699 FN(skb_get_tunnel_key), \
2700 FN(skb_set_tunnel_key), \
2701 FN(perf_event_read), \
2703 FN(get_route_realm), \
2704 FN(perf_event_output), \
2705 FN(skb_load_bytes), \
2708 FN(skb_get_tunnel_opt), \
2709 FN(skb_set_tunnel_opt), \
2710 FN(skb_change_proto), \
2711 FN(skb_change_type), \
2712 FN(skb_under_cgroup), \
2713 FN(get_hash_recalc), \
2714 FN(get_current_task), \
2715 FN(probe_write_user), \
2716 FN(current_task_under_cgroup), \
2717 FN(skb_change_tail), \
2718 FN(skb_pull_data), \
2720 FN(set_hash_invalid), \
2721 FN(get_numa_node_id), \
2722 FN(skb_change_head), \
2723 FN(xdp_adjust_head), \
2724 FN(probe_read_str), \
2725 FN(get_socket_cookie), \
2726 FN(get_socket_uid), \
2729 FN(skb_adjust_room), \
2731 FN(sk_redirect_map), \
2732 FN(sock_map_update), \
2733 FN(xdp_adjust_meta), \
2734 FN(perf_event_read_value), \
2735 FN(perf_prog_read_value), \
2737 FN(override_return), \
2738 FN(sock_ops_cb_flags_set), \
2739 FN(msg_redirect_map), \
2740 FN(msg_apply_bytes), \
2741 FN(msg_cork_bytes), \
2742 FN(msg_pull_data), \
2744 FN(xdp_adjust_tail), \
2745 FN(skb_get_xfrm_state), \
2747 FN(skb_load_bytes_relative), \
2749 FN(sock_hash_update), \
2750 FN(msg_redirect_hash), \
2751 FN(sk_redirect_hash), \
2752 FN(lwt_push_encap), \
2753 FN(lwt_seg6_store_bytes), \
2754 FN(lwt_seg6_adjust_srh), \
2755 FN(lwt_seg6_action), \
2758 FN(skb_cgroup_id), \
2759 FN(get_current_cgroup_id), \
2760 FN(get_local_storage), \
2761 FN(sk_select_reuseport), \
2762 FN(skb_ancestor_cgroup_id), \
2763 FN(sk_lookup_tcp), \
2764 FN(sk_lookup_udp), \
2766 FN(map_push_elem), \
2768 FN(map_peek_elem), \
2769 FN(msg_push_data), \
2771 FN(rc_pointer_rel), \
2776 FN(skb_ecn_set_ce), \
2777 FN(get_listener_sock), \
2778 FN(skc_lookup_tcp), \
2779 FN(tcp_check_syncookie), \
2780 FN(sysctl_get_name), \
2781 FN(sysctl_get_current_value), \
2782 FN(sysctl_get_new_value), \
2783 FN(sysctl_set_new_value), \
2786 FN(sk_storage_get), \
2787 FN(sk_storage_delete),
2789 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2790 * function eBPF program intends to call
2792 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2794 __BPF_FUNC_MAPPER(__BPF_ENUM_FN
)
2797 #undef __BPF_ENUM_FN
2799 /* All flags used by eBPF helper functions, placed here. */
2801 /* BPF_FUNC_skb_store_bytes flags. */
2802 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2803 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2805 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2806 * First 4 bits are for passing the header field size.
2808 #define BPF_F_HDR_FIELD_MASK 0xfULL
2810 /* BPF_FUNC_l4_csum_replace flags. */
2811 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2812 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2813 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2815 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2816 #define BPF_F_INGRESS (1ULL << 0)
2818 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2819 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2821 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2822 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2823 #define BPF_F_USER_STACK (1ULL << 8)
2824 /* flags used by BPF_FUNC_get_stackid only. */
2825 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2826 #define BPF_F_REUSE_STACKID (1ULL << 10)
2827 /* flags used by BPF_FUNC_get_stack only. */
2828 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2830 /* BPF_FUNC_skb_set_tunnel_key flags. */
2831 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2832 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2833 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2835 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2836 * BPF_FUNC_perf_event_read_value flags.
2838 #define BPF_F_INDEX_MASK 0xffffffffULL
2839 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2840 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2841 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2843 /* Current network namespace */
2844 #define BPF_F_CURRENT_NETNS (-1L)
2846 /* BPF_FUNC_skb_adjust_room flags. */
2847 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
2849 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
2850 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
2852 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
2853 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
2854 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
2855 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
2856 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
2857 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
2858 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
2860 /* BPF_FUNC_sysctl_get_name flags. */
2861 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
2863 /* BPF_FUNC_sk_storage_get flags */
2864 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0)
2866 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2867 enum bpf_adj_room_mode
{
2872 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2873 enum bpf_hdr_start_off
{
2878 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2879 enum bpf_lwt_encap_mode
{
2881 BPF_LWT_ENCAP_SEG6_INLINE
,
2885 #define __bpf_md_ptr(type, name) \
2889 } __attribute__((aligned(8)))
2891 /* user accessible mirror of in-kernel sk_buff.
2892 * new fields can only be added to the end of this structure
2898 __u32 queue_mapping
;
2904 __u32 ingress_ifindex
;
2914 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2916 __u32 remote_ip4
; /* Stored in network byte order */
2917 __u32 local_ip4
; /* Stored in network byte order */
2918 __u32 remote_ip6
[4]; /* Stored in network byte order */
2919 __u32 local_ip6
[4]; /* Stored in network byte order */
2920 __u32 remote_port
; /* Stored in network byte order */
2921 __u32 local_port
; /* stored in host byte order */
2925 __bpf_md_ptr(struct bpf_flow_keys
*, flow_keys
);
2929 __bpf_md_ptr(struct bpf_sock
*, sk
);
2932 struct bpf_tunnel_key
{
2936 __u32 remote_ipv6
[4];
2940 __u16 tunnel_ext
; /* Padding, future use. */
2944 /* user accessible mirror of in-kernel xfrm_state.
2945 * new fields can only be added to the end of this structure
2947 struct bpf_xfrm_state
{
2949 __u32 spi
; /* Stored in network byte order */
2951 __u16 ext
; /* Padding, future use. */
2953 __u32 remote_ipv4
; /* Stored in network byte order */
2954 __u32 remote_ipv6
[4]; /* Stored in network byte order */
2958 /* Generic BPF return codes which all BPF program types may support.
2959 * The values are binary compatible with their TC_ACT_* counter-part to
2960 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2963 * XDP is handled seprately, see XDP_*.
2971 /* >127 are reserved for prog type specific return codes.
2973 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
2974 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
2975 * changed and should be routed based on its new L3 header.
2976 * (This is an L3 redirect, as opposed to L2 redirect
2977 * represented by BPF_REDIRECT above).
2979 BPF_LWT_REROUTE
= 128,
2989 /* IP address also allows 1 and 2 bytes access */
2992 __u32 src_port
; /* host byte order */
2993 __u32 dst_port
; /* network byte order */
2999 struct bpf_tcp_sock
{
3000 __u32 snd_cwnd
; /* Sending congestion window */
3001 __u32 srtt_us
; /* smoothed round trip time << 3 in usecs */
3003 __u32 snd_ssthresh
; /* Slow start size threshold */
3004 __u32 rcv_nxt
; /* What we want to receive next */
3005 __u32 snd_nxt
; /* Next sequence we send */
3006 __u32 snd_una
; /* First byte we want an ack for */
3007 __u32 mss_cache
; /* Cached effective mss, not including SACKS */
3008 __u32 ecn_flags
; /* ECN status bits. */
3009 __u32 rate_delivered
; /* saved rate sample: packets delivered */
3010 __u32 rate_interval_us
; /* saved rate sample: time elapsed */
3011 __u32 packets_out
; /* Packets which are "in flight" */
3012 __u32 retrans_out
; /* Retransmitted packets out */
3013 __u32 total_retrans
; /* Total retransmits for entire connection */
3014 __u32 segs_in
; /* RFC4898 tcpEStatsPerfSegsIn
3015 * total number of segments in.
3017 __u32 data_segs_in
; /* RFC4898 tcpEStatsPerfDataSegsIn
3018 * total number of data segments in.
3020 __u32 segs_out
; /* RFC4898 tcpEStatsPerfSegsOut
3021 * The total number of segments sent.
3023 __u32 data_segs_out
; /* RFC4898 tcpEStatsPerfDataSegsOut
3024 * total number of data segments sent.
3026 __u32 lost_out
; /* Lost packets */
3027 __u32 sacked_out
; /* SACK'd packets */
3028 __u64 bytes_received
; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3029 * sum(delta(rcv_nxt)), or how many bytes
3032 __u64 bytes_acked
; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3033 * sum(delta(snd_una)), or how many bytes
3038 struct bpf_sock_tuple
{
3055 #define XDP_PACKET_HEADROOM 256
3057 /* User return codes for XDP prog type.
3058 * A valid XDP program must return one of these defined values. All other
3059 * return codes are reserved for future use. Unknown return codes will
3060 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3070 /* user accessible metadata for XDP packet hook
3071 * new fields must be added to the end of this structure
3077 /* Below access go through struct xdp_rxq_info */
3078 __u32 ingress_ifindex
; /* rxq->dev->ifindex */
3079 __u32 rx_queue_index
; /* rxq->queue_index */
3087 /* user accessible metadata for SK_MSG packet hook, new fields must
3088 * be added to the end of this structure
3091 __bpf_md_ptr(void *, data
);
3092 __bpf_md_ptr(void *, data_end
);
3095 __u32 remote_ip4
; /* Stored in network byte order */
3096 __u32 local_ip4
; /* Stored in network byte order */
3097 __u32 remote_ip6
[4]; /* Stored in network byte order */
3098 __u32 local_ip6
[4]; /* Stored in network byte order */
3099 __u32 remote_port
; /* Stored in network byte order */
3100 __u32 local_port
; /* stored in host byte order */
3101 __u32 size
; /* Total size of sk_msg */
3104 struct sk_reuseport_md
{
3106 * Start of directly accessible data. It begins from
3107 * the tcp/udp header.
3109 __bpf_md_ptr(void *, data
);
3110 /* End of directly accessible data */
3111 __bpf_md_ptr(void *, data_end
);
3113 * Total length of packet (starting from the tcp/udp header).
3114 * Note that the directly accessible bytes (data_end - data)
3115 * could be less than this "len". Those bytes could be
3116 * indirectly read by a helper "bpf_skb_load_bytes()".
3120 * Eth protocol in the mac header (network byte order). e.g.
3121 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3124 __u32 ip_protocol
; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3125 __u32 bind_inany
; /* Is sock bound to an INANY address? */
3126 __u32 hash
; /* A hash of the packet 4 tuples */
3129 #define BPF_TAG_SIZE 8
3131 struct bpf_prog_info
{
3134 __u8 tag
[BPF_TAG_SIZE
];
3135 __u32 jited_prog_len
;
3136 __u32 xlated_prog_len
;
3137 __aligned_u64 jited_prog_insns
;
3138 __aligned_u64 xlated_prog_insns
;
3139 __u64 load_time
; /* ns since boottime */
3140 __u32 created_by_uid
;
3142 __aligned_u64 map_ids
;
3143 char name
[BPF_OBJ_NAME_LEN
];
3145 __u32 gpl_compatible
:1;
3148 __u32 nr_jited_ksyms
;
3149 __u32 nr_jited_func_lens
;
3150 __aligned_u64 jited_ksyms
;
3151 __aligned_u64 jited_func_lens
;
3153 __u32 func_info_rec_size
;
3154 __aligned_u64 func_info
;
3157 __aligned_u64 line_info
;
3158 __aligned_u64 jited_line_info
;
3159 __u32 nr_jited_line_info
;
3160 __u32 line_info_rec_size
;
3161 __u32 jited_line_info_rec_size
;
3163 __aligned_u64 prog_tags
;
3166 } __attribute__((aligned(8)));
3168 struct bpf_map_info
{
3175 char name
[BPF_OBJ_NAME_LEN
];
3181 __u32 btf_key_type_id
;
3182 __u32 btf_value_type_id
;
3183 } __attribute__((aligned(8)));
3185 struct bpf_btf_info
{
3189 } __attribute__((aligned(8)));
3191 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3192 * by user and intended to be used by socket (e.g. to bind to, depends on
3193 * attach attach type).
3195 struct bpf_sock_addr
{
3196 __u32 user_family
; /* Allows 4-byte read, but no write. */
3197 __u32 user_ip4
; /* Allows 1,2,4-byte read and 4-byte write.
3198 * Stored in network byte order.
3200 __u32 user_ip6
[4]; /* Allows 1,2,4-byte read an 4-byte write.
3201 * Stored in network byte order.
3203 __u32 user_port
; /* Allows 4-byte read and write.
3204 * Stored in network byte order
3206 __u32 family
; /* Allows 4-byte read, but no write */
3207 __u32 type
; /* Allows 4-byte read, but no write */
3208 __u32 protocol
; /* Allows 4-byte read, but no write */
3209 __u32 msg_src_ip4
; /* Allows 1,2,4-byte read an 4-byte write.
3210 * Stored in network byte order.
3212 __u32 msg_src_ip6
[4]; /* Allows 1,2,4-byte read an 4-byte write.
3213 * Stored in network byte order.
3217 /* User bpf_sock_ops struct to access socket values and specify request ops
3218 * and their replies.
3219 * Some of this fields are in network (bigendian) byte order and may need
3220 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3221 * New fields can only be added at the end of this structure
3223 struct bpf_sock_ops
{
3226 __u32 args
[4]; /* Optionally passed to bpf program */
3227 __u32 reply
; /* Returned by bpf program */
3228 __u32 replylong
[4]; /* Optionally returned by bpf prog */
3231 __u32 remote_ip4
; /* Stored in network byte order */
3232 __u32 local_ip4
; /* Stored in network byte order */
3233 __u32 remote_ip6
[4]; /* Stored in network byte order */
3234 __u32 local_ip6
[4]; /* Stored in network byte order */
3235 __u32 remote_port
; /* Stored in network byte order */
3236 __u32 local_port
; /* stored in host byte order */
3237 __u32 is_fullsock
; /* Some TCP fields are only valid if
3238 * there is a full socket. If not, the
3239 * fields read as zero.
3242 __u32 srtt_us
; /* Averaged RTT << 3 in usecs */
3243 __u32 bpf_sock_ops_cb_flags
; /* flags defined in uapi/linux/tcp.h */
3252 __u32 rate_delivered
;
3253 __u32 rate_interval_us
;
3256 __u32 total_retrans
;
3260 __u32 data_segs_out
;
3264 __u64 bytes_received
;
3268 /* Definitions for bpf_sock_ops_cb_flags */
3269 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
3270 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
3271 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3272 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
3273 * supported cb flags
3276 /* List of known BPF sock_ops operators.
3277 * New entries can only be added at the end
3281 BPF_SOCK_OPS_TIMEOUT_INIT
, /* Should return SYN-RTO value to use or
3282 * -1 if default value should be used
3284 BPF_SOCK_OPS_RWND_INIT
, /* Should return initial advertized
3285 * window (in packets) or -1 if default
3286 * value should be used
3288 BPF_SOCK_OPS_TCP_CONNECT_CB
, /* Calls BPF program right before an
3289 * active connection is initialized
3291 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB
, /* Calls BPF program when an
3292 * active connection is
3295 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB
, /* Calls BPF program when a
3296 * passive connection is
3299 BPF_SOCK_OPS_NEEDS_ECN
, /* If connection's congestion control
3302 BPF_SOCK_OPS_BASE_RTT
, /* Get base RTT. The correct value is
3303 * based on the path and may be
3304 * dependent on the congestion control
3305 * algorithm. In general it indicates
3306 * a congestion threshold. RTTs above
3307 * this indicate congestion
3309 BPF_SOCK_OPS_RTO_CB
, /* Called when an RTO has triggered.
3310 * Arg1: value of icsk_retransmits
3311 * Arg2: value of icsk_rto
3312 * Arg3: whether RTO has expired
3314 BPF_SOCK_OPS_RETRANS_CB
, /* Called when skb is retransmitted.
3315 * Arg1: sequence number of 1st byte
3317 * Arg3: return value of
3318 * tcp_transmit_skb (0 => success)
3320 BPF_SOCK_OPS_STATE_CB
, /* Called when TCP changes state.
3324 BPF_SOCK_OPS_TCP_LISTEN_CB
, /* Called on listen(2), right after
3325 * socket transition to LISTEN state.
3329 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3330 * changes between the TCP and BPF versions. Ideally this should never happen.
3331 * If it does, we need to add code to convert them before calling
3332 * the BPF sock_ops function.
3335 BPF_TCP_ESTABLISHED
= 1,
3345 BPF_TCP_CLOSING
, /* Now a valid state */
3346 BPF_TCP_NEW_SYN_RECV
,
3348 BPF_TCP_MAX_STATES
/* Leave at the end! */
3351 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3352 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3354 struct bpf_perf_event_value
{
3360 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3361 #define BPF_DEVCG_ACC_READ (1ULL << 1)
3362 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3364 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3365 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3367 struct bpf_cgroup_dev_ctx
{
3368 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3374 struct bpf_raw_tracepoint_args
{
3378 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3379 * OUTPUT: Do lookup from egress perspective; default is ingress
3381 #define BPF_FIB_LOOKUP_DIRECT (1U << 0)
3382 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
3385 BPF_FIB_LKUP_RET_SUCCESS
, /* lookup successful */
3386 BPF_FIB_LKUP_RET_BLACKHOLE
, /* dest is blackholed; can be dropped */
3387 BPF_FIB_LKUP_RET_UNREACHABLE
, /* dest is unreachable; can be dropped */
3388 BPF_FIB_LKUP_RET_PROHIBIT
, /* dest not allowed; can be dropped */
3389 BPF_FIB_LKUP_RET_NOT_FWDED
, /* packet is not forwarded */
3390 BPF_FIB_LKUP_RET_FWD_DISABLED
, /* fwding is not enabled on ingress */
3391 BPF_FIB_LKUP_RET_UNSUPP_LWT
, /* fwd requires encapsulation */
3392 BPF_FIB_LKUP_RET_NO_NEIGH
, /* no neighbor entry for nh */
3393 BPF_FIB_LKUP_RET_FRAG_NEEDED
, /* fragmentation required to fwd */
3396 struct bpf_fib_lookup
{
3397 /* input: network family for lookup (AF_INET, AF_INET6)
3398 * output: network family of egress nexthop
3402 /* set if lookup is to consider L4 data - e.g., FIB rules */
3407 /* total length of packet from network header - used for MTU check */
3410 /* input: L3 device index for lookup
3411 * output: device index from FIB lookup
3416 /* inputs to lookup */
3417 __u8 tos
; /* AF_INET */
3418 __be32 flowinfo
; /* AF_INET6, flow_label + priority */
3420 /* output: metric of fib result (IPv4/IPv6 only) */
3426 __u32 ipv6_src
[4]; /* in6_addr; network order */
3429 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3430 * network header. output: bpf_fib_lookup sets to gateway address
3431 * if FIB lookup returns gateway route
3435 __u32 ipv6_dst
[4]; /* in6_addr; network order */
3439 __be16 h_vlan_proto
;
3441 __u8 smac
[6]; /* ETH_ALEN */
3442 __u8 dmac
[6]; /* ETH_ALEN */
3445 enum bpf_task_fd_type
{
3446 BPF_FD_TYPE_RAW_TRACEPOINT
, /* tp name */
3447 BPF_FD_TYPE_TRACEPOINT
, /* tp name */
3448 BPF_FD_TYPE_KPROBE
, /* (symbol + offset) or addr */
3449 BPF_FD_TYPE_KRETPROBE
, /* (symbol + offset) or addr */
3450 BPF_FD_TYPE_UPROBE
, /* filename + offset */
3451 BPF_FD_TYPE_URETPROBE
, /* filename + offset */
3454 struct bpf_flow_keys
{
3457 __u16 addr_proto
; /* ETH_P_* of valid addrs */
3471 __u32 ipv6_src
[4]; /* in6_addr; network order */
3472 __u32 ipv6_dst
[4]; /* in6_addr; network order */
3477 struct bpf_func_info
{
3482 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3483 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3485 struct bpf_line_info
{
3487 __u32 file_name_off
;
3492 struct bpf_spin_lock
{
3497 __u32 write
; /* Sysctl is being read (= 0) or written (= 1).
3498 * Allows 1,2,4-byte read, but no write.
3500 __u32 file_pos
; /* Sysctl file position to read from, write to.
3501 * Allows 1,2,4-byte read an 4-byte write.
3505 #endif /* _UAPI__LINUX_BPF_H__ */