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 __LINUX_BPF_H__
9 #define __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
,
116 BPF_MAP_TYPE_PROG_ARRAY
,
117 BPF_MAP_TYPE_PERF_EVENT_ARRAY
,
118 BPF_MAP_TYPE_PERCPU_HASH
,
119 BPF_MAP_TYPE_PERCPU_ARRAY
,
120 BPF_MAP_TYPE_STACK_TRACE
,
121 BPF_MAP_TYPE_CGROUP_ARRAY
,
122 BPF_MAP_TYPE_LRU_HASH
,
123 BPF_MAP_TYPE_LRU_PERCPU_HASH
,
124 BPF_MAP_TYPE_LPM_TRIE
,
125 BPF_MAP_TYPE_ARRAY_OF_MAPS
,
126 BPF_MAP_TYPE_HASH_OF_MAPS
,
128 BPF_MAP_TYPE_SOCKMAP
,
131 BPF_MAP_TYPE_SOCKHASH
,
132 BPF_MAP_TYPE_CGROUP_STORAGE
,
133 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY
,
134 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
,
137 BPF_MAP_TYPE_SK_STORAGE
,
138 BPF_MAP_TYPE_DEVMAP_HASH
,
141 /* Note that tracing related programs such as
142 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
143 * are not subject to a stable API since kernel internal data
144 * structures can change from release to release and may
145 * therefore break existing tracing BPF programs. Tracing BPF
146 * programs correspond to /a/ specific kernel which is to be
147 * analyzed, and not /a/ specific kernel /and/ all future ones.
150 BPF_PROG_TYPE_UNSPEC
,
151 BPF_PROG_TYPE_SOCKET_FILTER
,
152 BPF_PROG_TYPE_KPROBE
,
153 BPF_PROG_TYPE_SCHED_CLS
,
154 BPF_PROG_TYPE_SCHED_ACT
,
155 BPF_PROG_TYPE_TRACEPOINT
,
157 BPF_PROG_TYPE_PERF_EVENT
,
158 BPF_PROG_TYPE_CGROUP_SKB
,
159 BPF_PROG_TYPE_CGROUP_SOCK
,
160 BPF_PROG_TYPE_LWT_IN
,
161 BPF_PROG_TYPE_LWT_OUT
,
162 BPF_PROG_TYPE_LWT_XMIT
,
163 BPF_PROG_TYPE_SOCK_OPS
,
164 BPF_PROG_TYPE_SK_SKB
,
165 BPF_PROG_TYPE_CGROUP_DEVICE
,
166 BPF_PROG_TYPE_SK_MSG
,
167 BPF_PROG_TYPE_RAW_TRACEPOINT
,
168 BPF_PROG_TYPE_CGROUP_SOCK_ADDR
,
169 BPF_PROG_TYPE_LWT_SEG6LOCAL
,
170 BPF_PROG_TYPE_LIRC_MODE2
,
171 BPF_PROG_TYPE_SK_REUSEPORT
,
172 BPF_PROG_TYPE_FLOW_DISSECTOR
,
173 BPF_PROG_TYPE_CGROUP_SYSCTL
,
174 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE
,
175 BPF_PROG_TYPE_CGROUP_SOCKOPT
,
176 BPF_PROG_TYPE_TRACING
,
179 enum bpf_attach_type
{
180 BPF_CGROUP_INET_INGRESS
,
181 BPF_CGROUP_INET_EGRESS
,
182 BPF_CGROUP_INET_SOCK_CREATE
,
184 BPF_SK_SKB_STREAM_PARSER
,
185 BPF_SK_SKB_STREAM_VERDICT
,
188 BPF_CGROUP_INET4_BIND
,
189 BPF_CGROUP_INET6_BIND
,
190 BPF_CGROUP_INET4_CONNECT
,
191 BPF_CGROUP_INET6_CONNECT
,
192 BPF_CGROUP_INET4_POST_BIND
,
193 BPF_CGROUP_INET6_POST_BIND
,
194 BPF_CGROUP_UDP4_SENDMSG
,
195 BPF_CGROUP_UDP6_SENDMSG
,
199 BPF_CGROUP_UDP4_RECVMSG
,
200 BPF_CGROUP_UDP6_RECVMSG
,
201 BPF_CGROUP_GETSOCKOPT
,
202 BPF_CGROUP_SETSOCKOPT
,
206 __MAX_BPF_ATTACH_TYPE
209 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
211 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
213 * NONE(default): No further bpf programs allowed in the subtree.
215 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
216 * the program in this cgroup yields to sub-cgroup program.
218 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
219 * that cgroup program gets run in addition to the program in this cgroup.
221 * Only one program is allowed to be attached to a cgroup with
222 * NONE or BPF_F_ALLOW_OVERRIDE flag.
223 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
224 * release old program and attach the new one. Attach flags has to match.
226 * Multiple programs are allowed to be attached to a cgroup with
227 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
228 * (those that were attached first, run first)
229 * The programs of sub-cgroup are executed first, then programs of
230 * this cgroup and then programs of parent cgroup.
231 * When children program makes decision (like picking TCP CA or sock bind)
232 * parent program has a chance to override it.
234 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
235 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
237 * cgrp1 (MULTI progs A, B) ->
238 * cgrp2 (OVERRIDE prog C) ->
239 * cgrp3 (MULTI prog D) ->
240 * cgrp4 (OVERRIDE prog E) ->
241 * cgrp5 (NONE prog F)
242 * the event in cgrp5 triggers execution of F,D,A,B in that order.
243 * if prog F is detached, the execution is E,D,A,B
244 * if prog F and D are detached, the execution is E,A,B
245 * if prog F, E and D are detached, the execution is C,A,B
247 * All eligible programs are executed regardless of return code from
250 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
251 #define BPF_F_ALLOW_MULTI (1U << 1)
253 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
254 * verifier will perform strict alignment checking as if the kernel
255 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
256 * and NET_IP_ALIGN defined to 2.
258 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
260 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
261 * verifier will allow any alignment whatsoever. On platforms
262 * with strict alignment requirements for loads ands stores (such
263 * as sparc and mips) the verifier validates that all loads and
264 * stores provably follow this requirement. This flag turns that
265 * checking and enforcement off.
267 * It is mostly used for testing when we want to validate the
268 * context and memory access aspects of the verifier, but because
269 * of an unaligned access the alignment check would trigger before
270 * the one we are interested in.
272 #define BPF_F_ANY_ALIGNMENT (1U << 1)
274 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
275 * Verifier does sub-register def/use analysis and identifies instructions whose
276 * def only matters for low 32-bit, high 32-bit is never referenced later
277 * through implicit zero extension. Therefore verifier notifies JIT back-ends
278 * that it is safe to ignore clearing high 32-bit for these instructions. This
279 * saves some back-ends a lot of code-gen. However such optimization is not
280 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
281 * hence hasn't used verifier's analysis result. But, we really want to have a
282 * way to be able to verify the correctness of the described optimization on
283 * x86_64 on which testsuites are frequently exercised.
285 * So, this flag is introduced. Once it is set, verifier will randomize high
286 * 32-bit for those instructions who has been identified as safe to ignore them.
287 * Then, if verifier is not doing correct analysis, such randomization will
288 * regress tests to expose bugs.
290 #define BPF_F_TEST_RND_HI32 (1U << 2)
292 /* The verifier internal test flag. Behavior is undefined */
293 #define BPF_F_TEST_STATE_FREQ (1U << 3)
295 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
298 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
299 * insn[0].imm: map fd map fd
300 * insn[1].imm: 0 offset into value
303 * ldimm64 rewrite: address of map address of map[0]+offset
304 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
306 #define BPF_PSEUDO_MAP_FD 1
307 #define BPF_PSEUDO_MAP_VALUE 2
309 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
310 * offset to another bpf function
312 #define BPF_PSEUDO_CALL 1
314 /* flags for BPF_MAP_UPDATE_ELEM command */
315 #define BPF_ANY 0 /* create new element or update existing */
316 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
317 #define BPF_EXIST 2 /* update existing element */
318 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
320 /* flags for BPF_MAP_CREATE command */
321 #define BPF_F_NO_PREALLOC (1U << 0)
322 /* Instead of having one common LRU list in the
323 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
324 * which can scale and perform better.
325 * Note, the LRU nodes (including free nodes) cannot be moved
326 * across different LRU lists.
328 #define BPF_F_NO_COMMON_LRU (1U << 1)
329 /* Specify numa node during map creation */
330 #define BPF_F_NUMA_NODE (1U << 2)
332 #define BPF_OBJ_NAME_LEN 16U
334 /* Flags for accessing BPF object from syscall side. */
335 #define BPF_F_RDONLY (1U << 3)
336 #define BPF_F_WRONLY (1U << 4)
338 /* Flag for stack_map, store build_id+offset instead of pointer */
339 #define BPF_F_STACK_BUILD_ID (1U << 5)
341 /* Zero-initialize hash function seed. This should only be used for testing. */
342 #define BPF_F_ZERO_SEED (1U << 6)
344 /* Flags for accessing BPF object from program side. */
345 #define BPF_F_RDONLY_PROG (1U << 7)
346 #define BPF_F_WRONLY_PROG (1U << 8)
348 /* Clone map from listener for newly accepted socket */
349 #define BPF_F_CLONE (1U << 9)
351 /* Enable memory-mapping BPF map */
352 #define BPF_F_MMAPABLE (1U << 10)
354 /* flags for BPF_PROG_QUERY */
355 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
357 enum bpf_stack_build_id_status
{
358 /* user space need an empty entry to identify end of a trace */
359 BPF_STACK_BUILD_ID_EMPTY
= 0,
360 /* with valid build_id and offset */
361 BPF_STACK_BUILD_ID_VALID
= 1,
362 /* couldn't get build_id, fallback to ip */
363 BPF_STACK_BUILD_ID_IP
= 2,
366 #define BPF_BUILD_ID_SIZE 20
367 struct bpf_stack_build_id
{
369 unsigned char build_id
[BPF_BUILD_ID_SIZE
];
377 struct { /* anonymous struct used by BPF_MAP_CREATE command */
378 __u32 map_type
; /* one of enum bpf_map_type */
379 __u32 key_size
; /* size of key in bytes */
380 __u32 value_size
; /* size of value in bytes */
381 __u32 max_entries
; /* max number of entries in a map */
382 __u32 map_flags
; /* BPF_MAP_CREATE related
383 * flags defined above.
385 __u32 inner_map_fd
; /* fd pointing to the inner map */
386 __u32 numa_node
; /* numa node (effective only if
387 * BPF_F_NUMA_NODE is set).
389 char map_name
[BPF_OBJ_NAME_LEN
];
390 __u32 map_ifindex
; /* ifindex of netdev to create on */
391 __u32 btf_fd
; /* fd pointing to a BTF type data */
392 __u32 btf_key_type_id
; /* BTF type_id of the key */
393 __u32 btf_value_type_id
; /* BTF type_id of the value */
396 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
401 __aligned_u64 next_key
;
406 struct { /* anonymous struct used by BPF_PROG_LOAD command */
407 __u32 prog_type
; /* one of enum bpf_prog_type */
410 __aligned_u64 license
;
411 __u32 log_level
; /* verbosity level of verifier */
412 __u32 log_size
; /* size of user buffer */
413 __aligned_u64 log_buf
; /* user supplied buffer */
414 __u32 kern_version
; /* not used */
416 char prog_name
[BPF_OBJ_NAME_LEN
];
417 __u32 prog_ifindex
; /* ifindex of netdev to prep for */
418 /* For some prog types expected attach type must be known at
419 * load time to verify attach type specific parts of prog
420 * (context accesses, allowed helpers, etc).
422 __u32 expected_attach_type
;
423 __u32 prog_btf_fd
; /* fd pointing to BTF type data */
424 __u32 func_info_rec_size
; /* userspace bpf_func_info size */
425 __aligned_u64 func_info
; /* func info */
426 __u32 func_info_cnt
; /* number of bpf_func_info records */
427 __u32 line_info_rec_size
; /* userspace bpf_line_info size */
428 __aligned_u64 line_info
; /* line info */
429 __u32 line_info_cnt
; /* number of bpf_line_info records */
430 __u32 attach_btf_id
; /* in-kernel BTF type id to attach to */
431 __u32 attach_prog_fd
; /* 0 to attach to vmlinux */
434 struct { /* anonymous struct used by BPF_OBJ_* commands */
435 __aligned_u64 pathname
;
440 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
441 __u32 target_fd
; /* container object to attach to */
442 __u32 attach_bpf_fd
; /* eBPF program to attach */
447 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
450 __u32 data_size_in
; /* input: len of data_in */
451 __u32 data_size_out
; /* input/output: len of data_out
452 * returns ENOSPC if data_out
455 __aligned_u64 data_in
;
456 __aligned_u64 data_out
;
459 __u32 ctx_size_in
; /* input: len of ctx_in */
460 __u32 ctx_size_out
; /* input/output: len of ctx_out
461 * returns ENOSPC if ctx_out
464 __aligned_u64 ctx_in
;
465 __aligned_u64 ctx_out
;
468 struct { /* anonymous struct used by BPF_*_GET_*_ID */
479 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
485 struct { /* anonymous struct used by BPF_PROG_QUERY command */
486 __u32 target_fd
; /* container object to query */
490 __aligned_u64 prog_ids
;
499 struct { /* anonymous struct for BPF_BTF_LOAD */
501 __aligned_u64 btf_log_buf
;
508 __u32 pid
; /* input: pid */
509 __u32 fd
; /* input: fd */
510 __u32 flags
; /* input: flags */
511 __u32 buf_len
; /* input/output: buf len */
512 __aligned_u64 buf
; /* input/output:
513 * tp_name for tracepoint
515 * filename for uprobe
517 __u32 prog_id
; /* output: prod_id */
518 __u32 fd_type
; /* output: BPF_FD_TYPE_* */
519 __u64 probe_offset
; /* output: probe_offset */
520 __u64 probe_addr
; /* output: probe_addr */
522 } __attribute__((aligned(8)));
524 /* The description below is an attempt at providing documentation to eBPF
525 * developers about the multiple available eBPF helper functions. It can be
526 * parsed and used to produce a manual page. The workflow is the following,
527 * and requires the rst2man utility:
529 * $ ./scripts/bpf_helpers_doc.py \
530 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
531 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
532 * $ man /tmp/bpf-helpers.7
534 * Note that in order to produce this external documentation, some RST
535 * formatting is used in the descriptions to get "bold" and "italics" in
536 * manual pages. Also note that the few trailing white spaces are
537 * intentional, removing them would break paragraphs for rst2man.
539 * Start of BPF helper function descriptions:
541 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
543 * Perform a lookup in *map* for an entry associated to *key*.
545 * Map value associated to *key*, or **NULL** if no entry was
548 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
550 * Add or update the value of the entry associated to *key* in
551 * *map* with *value*. *flags* is one of:
554 * The entry for *key* must not exist in the map.
556 * The entry for *key* must already exist in the map.
558 * No condition on the existence of the entry for *key*.
560 * Flag value **BPF_NOEXIST** cannot be used for maps of types
561 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
562 * elements always exist), the helper would return an error.
564 * 0 on success, or a negative error in case of failure.
566 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
568 * Delete entry with *key* from *map*.
570 * 0 on success, or a negative error in case of failure.
572 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
574 * For tracing programs, safely attempt to read *size* bytes from
575 * kernel space address *unsafe_ptr* and store the data in *dst*.
577 * Generally, use bpf_probe_read_user() or bpf_probe_read_kernel()
580 * 0 on success, or a negative error in case of failure.
582 * u64 bpf_ktime_get_ns(void)
584 * Return the time elapsed since system boot, in nanoseconds.
588 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
590 * This helper is a "printk()-like" facility for debugging. It
591 * prints a message defined by format *fmt* (of size *fmt_size*)
592 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
593 * available. It can take up to three additional **u64**
594 * arguments (as an eBPF helpers, the total number of arguments is
597 * Each time the helper is called, it appends a line to the trace.
598 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
599 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
600 * The format of the trace is customizable, and the exact output
601 * one will get depends on the options set in
602 * *\/sys/kernel/debug/tracing/trace_options* (see also the
603 * *README* file under the same directory). However, it usually
604 * defaults to something like:
608 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
612 * * ``telnet`` is the name of the current task.
613 * * ``470`` is the PID of the current task.
614 * * ``001`` is the CPU number on which the task is
616 * * In ``.N..``, each character refers to a set of
617 * options (whether irqs are enabled, scheduling
618 * options, whether hard/softirqs are running, level of
619 * preempt_disabled respectively). **N** means that
620 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
622 * * ``419421.045894`` is a timestamp.
623 * * ``0x00000001`` is a fake value used by BPF for the
624 * instruction pointer register.
625 * * ``<formatted msg>`` is the message formatted with
628 * The conversion specifiers supported by *fmt* are similar, but
629 * more limited than for printk(). They are **%d**, **%i**,
630 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
631 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
632 * of field, padding with zeroes, etc.) is available, and the
633 * helper will return **-EINVAL** (but print nothing) if it
634 * encounters an unknown specifier.
636 * Also, note that **bpf_trace_printk**\ () is slow, and should
637 * only be used for debugging purposes. For this reason, a notice
638 * bloc (spanning several lines) is printed to kernel logs and
639 * states that the helper should not be used "for production use"
640 * the first time this helper is used (or more precisely, when
641 * **trace_printk**\ () buffers are allocated). For passing values
642 * to user space, perf events should be preferred.
644 * The number of bytes written to the buffer, or a negative error
645 * in case of failure.
647 * u32 bpf_get_prandom_u32(void)
649 * Get a pseudo-random number.
651 * From a security point of view, this helper uses its own
652 * pseudo-random internal state, and cannot be used to infer the
653 * seed of other random functions in the kernel. However, it is
654 * essential to note that the generator used by the helper is not
655 * cryptographically secure.
657 * A random 32-bit unsigned value.
659 * u32 bpf_get_smp_processor_id(void)
661 * Get the SMP (symmetric multiprocessing) processor id. Note that
662 * all programs run with preemption disabled, which means that the
663 * SMP processor id is stable during all the execution of the
666 * The SMP id of the processor running the program.
668 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
670 * Store *len* bytes from address *from* into the packet
671 * associated to *skb*, at *offset*. *flags* are a combination of
672 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
673 * checksum for the packet after storing the bytes) and
674 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
675 * **->swhash** and *skb*\ **->l4hash** to 0).
677 * A call to this helper is susceptible to change the underlying
678 * packet buffer. Therefore, at load time, all checks on pointers
679 * previously done by the verifier are invalidated and must be
680 * performed again, if the helper is used in combination with
681 * direct packet access.
683 * 0 on success, or a negative error in case of failure.
685 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
687 * Recompute the layer 3 (e.g. IP) checksum for the packet
688 * associated to *skb*. Computation is incremental, so the helper
689 * must know the former value of the header field that was
690 * modified (*from*), the new value of this field (*to*), and the
691 * number of bytes (2 or 4) for this field, stored in *size*.
692 * Alternatively, it is possible to store the difference between
693 * the previous and the new values of the header field in *to*, by
694 * setting *from* and *size* to 0. For both methods, *offset*
695 * indicates the location of the IP checksum within the packet.
697 * This helper works in combination with **bpf_csum_diff**\ (),
698 * which does not update the checksum in-place, but offers more
699 * flexibility and can handle sizes larger than 2 or 4 for the
700 * checksum to update.
702 * A call to this helper is susceptible to change the underlying
703 * packet buffer. Therefore, at load time, all checks on pointers
704 * previously done by the verifier are invalidated and must be
705 * performed again, if the helper is used in combination with
706 * direct packet access.
708 * 0 on success, or a negative error in case of failure.
710 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
712 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
713 * packet associated to *skb*. Computation is incremental, so the
714 * helper must know the former value of the header field that was
715 * modified (*from*), the new value of this field (*to*), and the
716 * number of bytes (2 or 4) for this field, stored on the lowest
717 * four bits of *flags*. Alternatively, it is possible to store
718 * the difference between the previous and the new values of the
719 * header field in *to*, by setting *from* and the four lowest
720 * bits of *flags* to 0. For both methods, *offset* indicates the
721 * location of the IP checksum within the packet. In addition to
722 * the size of the field, *flags* can be added (bitwise OR) actual
723 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
724 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
725 * for updates resulting in a null checksum the value is set to
726 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
727 * the checksum is to be computed against a pseudo-header.
729 * This helper works in combination with **bpf_csum_diff**\ (),
730 * which does not update the checksum in-place, but offers more
731 * flexibility and can handle sizes larger than 2 or 4 for the
732 * checksum to update.
734 * A call to this helper is susceptible to change the underlying
735 * packet buffer. Therefore, at load time, all checks on pointers
736 * previously done by the verifier are invalidated and must be
737 * performed again, if the helper is used in combination with
738 * direct packet access.
740 * 0 on success, or a negative error in case of failure.
742 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
744 * This special helper is used to trigger a "tail call", or in
745 * other words, to jump into another eBPF program. The same stack
746 * frame is used (but values on stack and in registers for the
747 * caller are not accessible to the callee). This mechanism allows
748 * for program chaining, either for raising the maximum number of
749 * available eBPF instructions, or to execute given programs in
750 * conditional blocks. For security reasons, there is an upper
751 * limit to the number of successive tail calls that can be
754 * Upon call of this helper, the program attempts to jump into a
755 * program referenced at index *index* in *prog_array_map*, a
756 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
757 * *ctx*, a pointer to the context.
759 * If the call succeeds, the kernel immediately runs the first
760 * instruction of the new program. This is not a function call,
761 * and it never returns to the previous program. If the call
762 * fails, then the helper has no effect, and the caller continues
763 * to run its subsequent instructions. A call can fail if the
764 * destination program for the jump does not exist (i.e. *index*
765 * is superior to the number of entries in *prog_array_map*), or
766 * if the maximum number of tail calls has been reached for this
767 * chain of programs. This limit is defined in the kernel by the
768 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
769 * which is currently set to 32.
771 * 0 on success, or a negative error in case of failure.
773 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
775 * Clone and redirect the packet associated to *skb* to another
776 * net device of index *ifindex*. Both ingress and egress
777 * interfaces can be used for redirection. The **BPF_F_INGRESS**
778 * value in *flags* is used to make the distinction (ingress path
779 * is selected if the flag is present, egress path otherwise).
780 * This is the only flag supported for now.
782 * In comparison with **bpf_redirect**\ () helper,
783 * **bpf_clone_redirect**\ () has the associated cost of
784 * duplicating the packet buffer, but this can be executed out of
785 * the eBPF program. Conversely, **bpf_redirect**\ () is more
786 * efficient, but it is handled through an action code where the
787 * redirection happens only after the eBPF program has returned.
789 * A call to this helper is susceptible to change the underlying
790 * packet buffer. Therefore, at load time, all checks on pointers
791 * previously done by the verifier are invalidated and must be
792 * performed again, if the helper is used in combination with
793 * direct packet access.
795 * 0 on success, or a negative error in case of failure.
797 * u64 bpf_get_current_pid_tgid(void)
799 * A 64-bit integer containing the current tgid and pid, and
801 * *current_task*\ **->tgid << 32 \|**
802 * *current_task*\ **->pid**.
804 * u64 bpf_get_current_uid_gid(void)
806 * A 64-bit integer containing the current GID and UID, and
807 * created as such: *current_gid* **<< 32 \|** *current_uid*.
809 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
811 * Copy the **comm** attribute of the current task into *buf* of
812 * *size_of_buf*. The **comm** attribute contains the name of
813 * the executable (excluding the path) for the current task. The
814 * *size_of_buf* must be strictly positive. On success, the
815 * helper makes sure that the *buf* is NUL-terminated. On failure,
816 * it is filled with zeroes.
818 * 0 on success, or a negative error in case of failure.
820 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
822 * Retrieve the classid for the current task, i.e. for the net_cls
823 * cgroup to which *skb* belongs.
825 * This helper can be used on TC egress path, but not on ingress.
827 * The net_cls cgroup provides an interface to tag network packets
828 * based on a user-provided identifier for all traffic coming from
829 * the tasks belonging to the related cgroup. See also the related
830 * kernel documentation, available from the Linux sources in file
831 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
833 * The Linux kernel has two versions for cgroups: there are
834 * cgroups v1 and cgroups v2. Both are available to users, who can
835 * use a mixture of them, but note that the net_cls cgroup is for
836 * cgroup v1 only. This makes it incompatible with BPF programs
837 * run on cgroups, which is a cgroup-v2-only feature (a socket can
838 * only hold data for one version of cgroups at a time).
840 * This helper is only available is the kernel was compiled with
841 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
842 * "**y**" or to "**m**".
844 * The classid, or 0 for the default unconfigured classid.
846 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
848 * Push a *vlan_tci* (VLAN tag control information) of protocol
849 * *vlan_proto* to the packet associated to *skb*, then update
850 * the checksum. Note that if *vlan_proto* is different from
851 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
852 * be **ETH_P_8021Q**.
854 * A call to this helper is susceptible to change the underlying
855 * packet buffer. Therefore, at load time, all checks on pointers
856 * previously done by the verifier are invalidated and must be
857 * performed again, if the helper is used in combination with
858 * direct packet access.
860 * 0 on success, or a negative error in case of failure.
862 * int bpf_skb_vlan_pop(struct sk_buff *skb)
864 * Pop a VLAN header from the packet associated to *skb*.
866 * A call to this helper is susceptible to change the underlying
867 * packet buffer. Therefore, at load time, all checks on pointers
868 * previously done by the verifier are invalidated and must be
869 * performed again, if the helper is used in combination with
870 * direct packet access.
872 * 0 on success, or a negative error in case of failure.
874 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
876 * Get tunnel metadata. This helper takes a pointer *key* to an
877 * empty **struct bpf_tunnel_key** of **size**, that will be
878 * filled with tunnel metadata for the packet associated to *skb*.
879 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
880 * indicates that the tunnel is based on IPv6 protocol instead of
883 * The **struct bpf_tunnel_key** is an object that generalizes the
884 * principal parameters used by various tunneling protocols into a
885 * single struct. This way, it can be used to easily make a
886 * decision based on the contents of the encapsulation header,
887 * "summarized" in this struct. In particular, it holds the IP
888 * address of the remote end (IPv4 or IPv6, depending on the case)
889 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
890 * this struct exposes the *key*\ **->tunnel_id**, which is
891 * generally mapped to a VNI (Virtual Network Identifier), making
892 * it programmable together with the **bpf_skb_set_tunnel_key**\
895 * Let's imagine that the following code is part of a program
896 * attached to the TC ingress interface, on one end of a GRE
897 * tunnel, and is supposed to filter out all messages coming from
898 * remote ends with IPv4 address other than 10.0.0.1:
903 * struct bpf_tunnel_key key = {};
905 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
907 * return TC_ACT_SHOT; // drop packet
909 * if (key.remote_ipv4 != 0x0a000001)
910 * return TC_ACT_SHOT; // drop packet
912 * return TC_ACT_OK; // accept packet
914 * This interface can also be used with all encapsulation devices
915 * that can operate in "collect metadata" mode: instead of having
916 * one network device per specific configuration, the "collect
917 * metadata" mode only requires a single device where the
918 * configuration can be extracted from this helper.
920 * This can be used together with various tunnels such as VXLan,
921 * Geneve, GRE or IP in IP (IPIP).
923 * 0 on success, or a negative error in case of failure.
925 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
927 * Populate tunnel metadata for packet associated to *skb.* The
928 * tunnel metadata is set to the contents of *key*, of *size*. The
929 * *flags* can be set to a combination of the following values:
931 * **BPF_F_TUNINFO_IPV6**
932 * Indicate that the tunnel is based on IPv6 protocol
934 * **BPF_F_ZERO_CSUM_TX**
935 * For IPv4 packets, add a flag to tunnel metadata
936 * indicating that checksum computation should be skipped
937 * and checksum set to zeroes.
938 * **BPF_F_DONT_FRAGMENT**
939 * Add a flag to tunnel metadata indicating that the
940 * packet should not be fragmented.
941 * **BPF_F_SEQ_NUMBER**
942 * Add a flag to tunnel metadata indicating that a
943 * sequence number should be added to tunnel header before
944 * sending the packet. This flag was added for GRE
945 * encapsulation, but might be used with other protocols
946 * as well in the future.
948 * Here is a typical usage on the transmit path:
952 * struct bpf_tunnel_key key;
954 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
955 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
957 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
958 * helper for additional information.
960 * 0 on success, or a negative error in case of failure.
962 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
964 * Read the value of a perf event counter. This helper relies on a
965 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
966 * the perf event counter is selected when *map* is updated with
967 * perf event file descriptors. The *map* is an array whose size
968 * is the number of available CPUs, and each cell contains a value
969 * relative to one CPU. The value to retrieve is indicated by
970 * *flags*, that contains the index of the CPU to look up, masked
971 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
972 * **BPF_F_CURRENT_CPU** to indicate that the value for the
973 * current CPU should be retrieved.
975 * Note that before Linux 4.13, only hardware perf event can be
978 * Also, be aware that the newer helper
979 * **bpf_perf_event_read_value**\ () is recommended over
980 * **bpf_perf_event_read**\ () in general. The latter has some ABI
981 * quirks where error and counter value are used as a return code
982 * (which is wrong to do since ranges may overlap). This issue is
983 * fixed with **bpf_perf_event_read_value**\ (), which at the same
984 * time provides more features over the **bpf_perf_event_read**\
985 * () interface. Please refer to the description of
986 * **bpf_perf_event_read_value**\ () for details.
988 * The value of the perf event counter read from the map, or a
989 * negative error code in case of failure.
991 * int bpf_redirect(u32 ifindex, u64 flags)
993 * Redirect the packet to another net device of index *ifindex*.
994 * This helper is somewhat similar to **bpf_clone_redirect**\
995 * (), except that the packet is not cloned, which provides
996 * increased performance.
998 * Except for XDP, both ingress and egress interfaces can be used
999 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1000 * to make the distinction (ingress path is selected if the flag
1001 * is present, egress path otherwise). Currently, XDP only
1002 * supports redirection to the egress interface, and accepts no
1005 * The same effect can be attained with the more generic
1006 * **bpf_redirect_map**\ (), which requires specific maps to be
1007 * used but offers better performance.
1009 * For XDP, the helper returns **XDP_REDIRECT** on success or
1010 * **XDP_ABORTED** on error. For other program types, the values
1011 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1014 * u32 bpf_get_route_realm(struct sk_buff *skb)
1016 * Retrieve the realm or the route, that is to say the
1017 * **tclassid** field of the destination for the *skb*. The
1018 * indentifier retrieved is a user-provided tag, similar to the
1019 * one used with the net_cls cgroup (see description for
1020 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1021 * held by a route (a destination entry), not by a task.
1023 * Retrieving this identifier works with the clsact TC egress hook
1024 * (see also **tc-bpf(8)**), or alternatively on conventional
1025 * classful egress qdiscs, but not on TC ingress path. In case of
1026 * clsact TC egress hook, this has the advantage that, internally,
1027 * the destination entry has not been dropped yet in the transmit
1028 * path. Therefore, the destination entry does not need to be
1029 * artificially held via **netif_keep_dst**\ () for a classful
1030 * qdisc until the *skb* is freed.
1032 * This helper is available only if the kernel was compiled with
1033 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1035 * The realm of the route for the packet associated to *skb*, or 0
1036 * if none was found.
1038 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1040 * Write raw *data* blob into a special BPF perf event held by
1041 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1042 * event must have the following attributes: **PERF_SAMPLE_RAW**
1043 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1044 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1046 * The *flags* are used to indicate the index in *map* for which
1047 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1048 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1049 * to indicate that the index of the current CPU core should be
1052 * The value to write, of *size*, is passed through eBPF stack and
1053 * pointed by *data*.
1055 * The context of the program *ctx* needs also be passed to the
1058 * On user space, a program willing to read the values needs to
1059 * call **perf_event_open**\ () on the perf event (either for
1060 * one or for all CPUs) and to store the file descriptor into the
1061 * *map*. This must be done before the eBPF program can send data
1062 * into it. An example is available in file
1063 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1064 * tree (the eBPF program counterpart is in
1065 * *samples/bpf/trace_output_kern.c*).
1067 * **bpf_perf_event_output**\ () achieves better performance
1068 * than **bpf_trace_printk**\ () for sharing data with user
1069 * space, and is much better suitable for streaming data from eBPF
1072 * Note that this helper is not restricted to tracing use cases
1073 * and can be used with programs attached to TC or XDP as well,
1074 * where it allows for passing data to user space listeners. Data
1077 * * Only custom structs,
1078 * * Only the packet payload, or
1079 * * A combination of both.
1081 * 0 on success, or a negative error in case of failure.
1083 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1085 * This helper was provided as an easy way to load data from a
1086 * packet. It can be used to load *len* bytes from *offset* from
1087 * the packet associated to *skb*, into the buffer pointed by
1090 * Since Linux 4.7, usage of this helper has mostly been replaced
1091 * by "direct packet access", enabling packet data to be
1092 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1093 * pointing respectively to the first byte of packet data and to
1094 * the byte after the last byte of packet data. However, it
1095 * remains useful if one wishes to read large quantities of data
1096 * at once from a packet into the eBPF stack.
1098 * 0 on success, or a negative error in case of failure.
1100 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1102 * Walk a user or a kernel stack and return its id. To achieve
1103 * this, the helper needs *ctx*, which is a pointer to the context
1104 * on which the tracing program is executed, and a pointer to a
1105 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1107 * The last argument, *flags*, holds the number of stack frames to
1108 * skip (from 0 to 255), masked with
1109 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1110 * a combination of the following flags:
1112 * **BPF_F_USER_STACK**
1113 * Collect a user space stack instead of a kernel stack.
1114 * **BPF_F_FAST_STACK_CMP**
1115 * Compare stacks by hash only.
1116 * **BPF_F_REUSE_STACKID**
1117 * If two different stacks hash into the same *stackid*,
1118 * discard the old one.
1120 * The stack id retrieved is a 32 bit long integer handle which
1121 * can be further combined with other data (including other stack
1122 * ids) and used as a key into maps. This can be useful for
1123 * generating a variety of graphs (such as flame graphs or off-cpu
1126 * For walking a stack, this helper is an improvement over
1127 * **bpf_probe_read**\ (), which can be used with unrolled loops
1128 * but is not efficient and consumes a lot of eBPF instructions.
1129 * Instead, **bpf_get_stackid**\ () can collect up to
1130 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1131 * this limit can be controlled with the **sysctl** program, and
1132 * that it should be manually increased in order to profile long
1133 * user stacks (such as stacks for Java programs). To do so, use:
1137 * # sysctl kernel.perf_event_max_stack=<new value>
1139 * The positive or null stack id on success, or a negative error
1140 * in case of failure.
1142 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1144 * Compute a checksum difference, from the raw buffer pointed by
1145 * *from*, of length *from_size* (that must be a multiple of 4),
1146 * towards the raw buffer pointed by *to*, of size *to_size*
1147 * (same remark). An optional *seed* can be added to the value
1148 * (this can be cascaded, the seed may come from a previous call
1151 * This is flexible enough to be used in several ways:
1153 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1154 * checksum, it can be used when pushing new data.
1155 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1156 * checksum, it can be used when removing data from a packet.
1157 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1158 * can be used to compute a diff. Note that *from_size* and
1159 * *to_size* do not need to be equal.
1161 * This helper can be used in combination with
1162 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1163 * which one can feed in the difference computed with
1164 * **bpf_csum_diff**\ ().
1166 * The checksum result, or a negative error code in case of
1169 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1171 * Retrieve tunnel options metadata for the packet associated to
1172 * *skb*, and store the raw tunnel option data to the buffer *opt*
1175 * This helper can be used with encapsulation devices that can
1176 * operate in "collect metadata" mode (please refer to the related
1177 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1178 * more details). A particular example where this can be used is
1179 * in combination with the Geneve encapsulation protocol, where it
1180 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1181 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1182 * the eBPF program. This allows for full customization of these
1185 * The size of the option data retrieved.
1187 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1189 * Set tunnel options metadata for the packet associated to *skb*
1190 * to the option data contained in the raw buffer *opt* of *size*.
1192 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1193 * helper for additional information.
1195 * 0 on success, or a negative error in case of failure.
1197 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1199 * Change the protocol of the *skb* to *proto*. Currently
1200 * supported are transition from IPv4 to IPv6, and from IPv6 to
1201 * IPv4. The helper takes care of the groundwork for the
1202 * transition, including resizing the socket buffer. The eBPF
1203 * program is expected to fill the new headers, if any, via
1204 * **skb_store_bytes**\ () and to recompute the checksums with
1205 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1206 * (). The main case for this helper is to perform NAT64
1207 * operations out of an eBPF program.
1209 * Internally, the GSO type is marked as dodgy so that headers are
1210 * checked and segments are recalculated by the GSO/GRO engine.
1211 * The size for GSO target is adapted as well.
1213 * All values for *flags* are reserved for future usage, and must
1216 * A call to this helper is susceptible to change the underlying
1217 * packet buffer. Therefore, at load time, all checks on pointers
1218 * previously done by the verifier are invalidated and must be
1219 * performed again, if the helper is used in combination with
1220 * direct packet access.
1222 * 0 on success, or a negative error in case of failure.
1224 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1226 * Change the packet type for the packet associated to *skb*. This
1227 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1228 * the eBPF program does not have a write access to *skb*\
1229 * **->pkt_type** beside this helper. Using a helper here allows
1230 * for graceful handling of errors.
1232 * The major use case is to change incoming *skb*s to
1233 * **PACKET_HOST** in a programmatic way instead of having to
1234 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1237 * Note that *type* only allows certain values. At this time, they
1242 * **PACKET_BROADCAST**
1243 * Send packet to all.
1244 * **PACKET_MULTICAST**
1245 * Send packet to group.
1246 * **PACKET_OTHERHOST**
1247 * Send packet to someone else.
1249 * 0 on success, or a negative error in case of failure.
1251 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1253 * Check whether *skb* is a descendant of the cgroup2 held by
1254 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1256 * The return value depends on the result of the test, and can be:
1258 * * 0, if the *skb* failed the cgroup2 descendant test.
1259 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1260 * * A negative error code, if an error occurred.
1262 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1264 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1265 * not set, in particular if the hash was cleared due to mangling,
1266 * recompute this hash. Later accesses to the hash can be done
1267 * directly with *skb*\ **->hash**.
1269 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1270 * prototype with **bpf_skb_change_proto**\ (), or calling
1271 * **bpf_skb_store_bytes**\ () with the
1272 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1273 * the hash and to trigger a new computation for the next call to
1274 * **bpf_get_hash_recalc**\ ().
1278 * u64 bpf_get_current_task(void)
1280 * A pointer to the current task struct.
1282 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1284 * Attempt in a safe way to write *len* bytes from the buffer
1285 * *src* to *dst* in memory. It only works for threads that are in
1286 * user context, and *dst* must be a valid user space address.
1288 * This helper should not be used to implement any kind of
1289 * security mechanism because of TOC-TOU attacks, but rather to
1290 * debug, divert, and manipulate execution of semi-cooperative
1293 * Keep in mind that this feature is meant for experiments, and it
1294 * has a risk of crashing the system and running programs.
1295 * Therefore, when an eBPF program using this helper is attached,
1296 * a warning including PID and process name is printed to kernel
1299 * 0 on success, or a negative error in case of failure.
1301 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1303 * Check whether the probe is being run is the context of a given
1304 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1305 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1307 * The return value depends on the result of the test, and can be:
1309 * * 0, if the *skb* task belongs to the cgroup2.
1310 * * 1, if the *skb* task does not belong to the cgroup2.
1311 * * A negative error code, if an error occurred.
1313 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1315 * Resize (trim or grow) the packet associated to *skb* to the
1316 * new *len*. The *flags* are reserved for future usage, and must
1319 * The basic idea is that the helper performs the needed work to
1320 * change the size of the packet, then the eBPF program rewrites
1321 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1322 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1323 * and others. This helper is a slow path utility intended for
1324 * replies with control messages. And because it is targeted for
1325 * slow path, the helper itself can afford to be slow: it
1326 * implicitly linearizes, unclones and drops offloads from the
1329 * A call to this helper is susceptible to change the underlying
1330 * packet buffer. Therefore, at load time, all checks on pointers
1331 * previously done by the verifier are invalidated and must be
1332 * performed again, if the helper is used in combination with
1333 * direct packet access.
1335 * 0 on success, or a negative error in case of failure.
1337 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1339 * Pull in non-linear data in case the *skb* is non-linear and not
1340 * all of *len* are part of the linear section. Make *len* bytes
1341 * from *skb* readable and writable. If a zero value is passed for
1342 * *len*, then the whole length of the *skb* is pulled.
1344 * This helper is only needed for reading and writing with direct
1347 * For direct packet access, testing that offsets to access
1348 * are within packet boundaries (test on *skb*\ **->data_end**) is
1349 * susceptible to fail if offsets are invalid, or if the requested
1350 * data is in non-linear parts of the *skb*. On failure the
1351 * program can just bail out, or in the case of a non-linear
1352 * buffer, use a helper to make the data available. The
1353 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1354 * the data. Another one consists in using **bpf_skb_pull_data**
1355 * to pull in once the non-linear parts, then retesting and
1356 * eventually access the data.
1358 * At the same time, this also makes sure the *skb* is uncloned,
1359 * which is a necessary condition for direct write. As this needs
1360 * to be an invariant for the write part only, the verifier
1361 * detects writes and adds a prologue that is calling
1362 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1363 * the very beginning in case it is indeed cloned.
1365 * A call to this helper is susceptible to change the underlying
1366 * packet buffer. Therefore, at load time, all checks on pointers
1367 * previously done by the verifier are invalidated and must be
1368 * performed again, if the helper is used in combination with
1369 * direct packet access.
1371 * 0 on success, or a negative error in case of failure.
1373 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1375 * Add the checksum *csum* into *skb*\ **->csum** in case the
1376 * driver has supplied a checksum for the entire packet into that
1377 * field. Return an error otherwise. This helper is intended to be
1378 * used in combination with **bpf_csum_diff**\ (), in particular
1379 * when the checksum needs to be updated after data has been
1380 * written into the packet through direct packet access.
1382 * The checksum on success, or a negative error code in case of
1385 * void bpf_set_hash_invalid(struct sk_buff *skb)
1387 * Invalidate the current *skb*\ **->hash**. It can be used after
1388 * mangling on headers through direct packet access, in order to
1389 * indicate that the hash is outdated and to trigger a
1390 * recalculation the next time the kernel tries to access this
1391 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1393 * int bpf_get_numa_node_id(void)
1395 * Return the id of the current NUMA node. The primary use case
1396 * for this helper is the selection of sockets for the local NUMA
1397 * node, when the program is attached to sockets using the
1398 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1399 * but the helper is also available to other eBPF program types,
1400 * similarly to **bpf_get_smp_processor_id**\ ().
1402 * The id of current NUMA node.
1404 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1406 * Grows headroom of packet associated to *skb* and adjusts the
1407 * offset of the MAC header accordingly, adding *len* bytes of
1408 * space. It automatically extends and reallocates memory as
1411 * This helper can be used on a layer 3 *skb* to push a MAC header
1412 * for redirection into a layer 2 device.
1414 * All values for *flags* are reserved for future usage, and must
1417 * A call to this helper is susceptible to change the underlying
1418 * packet buffer. Therefore, at load time, all checks on pointers
1419 * previously done by the verifier are invalidated and must be
1420 * performed again, if the helper is used in combination with
1421 * direct packet access.
1423 * 0 on success, or a negative error in case of failure.
1425 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1427 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1428 * it is possible to use a negative value for *delta*. This helper
1429 * can be used to prepare the packet for pushing or popping
1432 * A call to this helper is susceptible to change the underlying
1433 * packet buffer. Therefore, at load time, all checks on pointers
1434 * previously done by the verifier are invalidated and must be
1435 * performed again, if the helper is used in combination with
1436 * direct packet access.
1438 * 0 on success, or a negative error in case of failure.
1440 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1442 * Copy a NUL terminated string from an unsafe kernel address
1443 * *unsafe_ptr* to *dst*. See bpf_probe_read_kernel_str() for
1446 * Generally, use bpf_probe_read_user_str() or bpf_probe_read_kernel_str()
1449 * On success, the strictly positive length of the string,
1450 * including the trailing NUL character. On error, a negative
1453 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1455 * If the **struct sk_buff** pointed by *skb* has a known socket,
1456 * retrieve the cookie (generated by the kernel) of this socket.
1457 * If no cookie has been set yet, generate a new cookie. Once
1458 * generated, the socket cookie remains stable for the life of the
1459 * socket. This helper can be useful for monitoring per socket
1460 * networking traffic statistics as it provides a global socket
1461 * identifier that can be assumed unique.
1463 * A 8-byte long non-decreasing number on success, or 0 if the
1464 * socket field is missing inside *skb*.
1466 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1468 * Equivalent to bpf_get_socket_cookie() helper that accepts
1469 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1471 * A 8-byte long non-decreasing number.
1473 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1475 * Equivalent to bpf_get_socket_cookie() helper that accepts
1476 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1478 * A 8-byte long non-decreasing number.
1480 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1482 * The owner UID of the socket associated to *skb*. If the socket
1483 * is **NULL**, or if it is not a full socket (i.e. if it is a
1484 * time-wait or a request socket instead), **overflowuid** value
1485 * is returned (note that **overflowuid** might also be the actual
1486 * UID value for the socket).
1488 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1490 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1495 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1497 * Emulate a call to **setsockopt()** on the socket associated to
1498 * *bpf_socket*, which must be a full socket. The *level* at
1499 * which the option resides and the name *optname* of the option
1500 * must be specified, see **setsockopt(2)** for more information.
1501 * The option value of length *optlen* is pointed by *optval*.
1503 * This helper actually implements a subset of **setsockopt()**.
1504 * It supports the following *level*\ s:
1506 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1507 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1508 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1509 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1510 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1511 * **TCP_BPF_SNDCWND_CLAMP**.
1512 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1513 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1515 * 0 on success, or a negative error in case of failure.
1517 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1519 * Grow or shrink the room for data in the packet associated to
1520 * *skb* by *len_diff*, and according to the selected *mode*.
1522 * There are two supported modes at this time:
1524 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1525 * (room space is added or removed below the layer 2 header).
1527 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1528 * (room space is added or removed below the layer 3 header).
1530 * The following flags are supported at this time:
1532 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1533 * Adjusting mss in this way is not allowed for datagrams.
1535 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1536 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1537 * Any new space is reserved to hold a tunnel header.
1538 * Configure skb offsets and other fields accordingly.
1540 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1541 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1542 * Use with ENCAP_L3 flags to further specify the tunnel type.
1544 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1545 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1546 * type; *len* is the length of the inner MAC header.
1548 * A call to this helper is susceptible to change the underlying
1549 * packet buffer. Therefore, at load time, all checks on pointers
1550 * previously done by the verifier are invalidated and must be
1551 * performed again, if the helper is used in combination with
1552 * direct packet access.
1554 * 0 on success, or a negative error in case of failure.
1556 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1558 * Redirect the packet to the endpoint referenced by *map* at
1559 * index *key*. Depending on its type, this *map* can contain
1560 * references to net devices (for forwarding packets through other
1561 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1562 * but this is only implemented for native XDP (with driver
1563 * support) as of this writing).
1565 * The lower two bits of *flags* are used as the return code if
1566 * the map lookup fails. This is so that the return value can be
1567 * one of the XDP program return codes up to XDP_TX, as chosen by
1568 * the caller. Any higher bits in the *flags* argument must be
1571 * When used to redirect packets to net devices, this helper
1572 * provides a high performance increase over **bpf_redirect**\ ().
1573 * This is due to various implementation details of the underlying
1574 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1575 * () tries to send packet as a "bulk" to the device.
1577 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1579 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1581 * Redirect the packet to the socket referenced by *map* (of type
1582 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1583 * egress interfaces can be used for redirection. The
1584 * **BPF_F_INGRESS** value in *flags* is used to make the
1585 * distinction (ingress path is selected if the flag is present,
1586 * egress path otherwise). This is the only flag supported for now.
1588 * **SK_PASS** on success, or **SK_DROP** on error.
1590 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1592 * Add an entry to, or update a *map* referencing sockets. The
1593 * *skops* is used as a new value for the entry associated to
1594 * *key*. *flags* is one of:
1597 * The entry for *key* must not exist in the map.
1599 * The entry for *key* must already exist in the map.
1601 * No condition on the existence of the entry for *key*.
1603 * If the *map* has eBPF programs (parser and verdict), those will
1604 * be inherited by the socket being added. If the socket is
1605 * already attached to eBPF programs, this results in an error.
1607 * 0 on success, or a negative error in case of failure.
1609 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1611 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1612 * *delta* (which can be positive or negative). Note that this
1613 * operation modifies the address stored in *xdp_md*\ **->data**,
1614 * so the latter must be loaded only after the helper has been
1617 * The use of *xdp_md*\ **->data_meta** is optional and programs
1618 * are not required to use it. The rationale is that when the
1619 * packet is processed with XDP (e.g. as DoS filter), it is
1620 * possible to push further meta data along with it before passing
1621 * to the stack, and to give the guarantee that an ingress eBPF
1622 * program attached as a TC classifier on the same device can pick
1623 * this up for further post-processing. Since TC works with socket
1624 * buffers, it remains possible to set from XDP the **mark** or
1625 * **priority** pointers, or other pointers for the socket buffer.
1626 * Having this scratch space generic and programmable allows for
1627 * more flexibility as the user is free to store whatever meta
1630 * A call to this helper is susceptible to change the underlying
1631 * packet buffer. Therefore, at load time, all checks on pointers
1632 * previously done by the verifier are invalidated and must be
1633 * performed again, if the helper is used in combination with
1634 * direct packet access.
1636 * 0 on success, or a negative error in case of failure.
1638 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1640 * Read the value of a perf event counter, and store it into *buf*
1641 * of size *buf_size*. This helper relies on a *map* of type
1642 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1643 * counter is selected when *map* is updated with perf event file
1644 * descriptors. The *map* is an array whose size is the number of
1645 * available CPUs, and each cell contains a value relative to one
1646 * CPU. The value to retrieve is indicated by *flags*, that
1647 * contains the index of the CPU to look up, masked with
1648 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1649 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1650 * current CPU should be retrieved.
1652 * This helper behaves in a way close to
1653 * **bpf_perf_event_read**\ () helper, save that instead of
1654 * just returning the value observed, it fills the *buf*
1655 * structure. This allows for additional data to be retrieved: in
1656 * particular, the enabled and running times (in *buf*\
1657 * **->enabled** and *buf*\ **->running**, respectively) are
1658 * copied. In general, **bpf_perf_event_read_value**\ () is
1659 * recommended over **bpf_perf_event_read**\ (), which has some
1660 * ABI issues and provides fewer functionalities.
1662 * These values are interesting, because hardware PMU (Performance
1663 * Monitoring Unit) counters are limited resources. When there are
1664 * more PMU based perf events opened than available counters,
1665 * kernel will multiplex these events so each event gets certain
1666 * percentage (but not all) of the PMU time. In case that
1667 * multiplexing happens, the number of samples or counter value
1668 * will not reflect the case compared to when no multiplexing
1669 * occurs. This makes comparison between different runs difficult.
1670 * Typically, the counter value should be normalized before
1671 * comparing to other experiments. The usual normalization is done
1676 * normalized_counter = counter * t_enabled / t_running
1678 * Where t_enabled is the time enabled for event and t_running is
1679 * the time running for event since last normalization. The
1680 * enabled and running times are accumulated since the perf event
1681 * open. To achieve scaling factor between two invocations of an
1682 * eBPF program, users can can use CPU id as the key (which is
1683 * typical for perf array usage model) to remember the previous
1684 * value and do the calculation inside the eBPF program.
1686 * 0 on success, or a negative error in case of failure.
1688 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1690 * For en eBPF program attached to a perf event, retrieve the
1691 * value of the event counter associated to *ctx* and store it in
1692 * the structure pointed by *buf* and of size *buf_size*. Enabled
1693 * and running times are also stored in the structure (see
1694 * description of helper **bpf_perf_event_read_value**\ () for
1697 * 0 on success, or a negative error in case of failure.
1699 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1701 * Emulate a call to **getsockopt()** on the socket associated to
1702 * *bpf_socket*, which must be a full socket. The *level* at
1703 * which the option resides and the name *optname* of the option
1704 * must be specified, see **getsockopt(2)** for more information.
1705 * The retrieved value is stored in the structure pointed by
1706 * *opval* and of length *optlen*.
1708 * This helper actually implements a subset of **getsockopt()**.
1709 * It supports the following *level*\ s:
1711 * * **IPPROTO_TCP**, which supports *optname*
1712 * **TCP_CONGESTION**.
1713 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1714 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1716 * 0 on success, or a negative error in case of failure.
1718 * int bpf_override_return(struct pt_regs *regs, u64 rc)
1720 * Used for error injection, this helper uses kprobes to override
1721 * the return value of the probed function, and to set it to *rc*.
1722 * The first argument is the context *regs* on which the kprobe
1725 * This helper works by setting setting the PC (program counter)
1726 * to an override function which is run in place of the original
1727 * probed function. This means the probed function is not run at
1728 * all. The replacement function just returns with the required
1731 * This helper has security implications, and thus is subject to
1732 * restrictions. It is only available if the kernel was compiled
1733 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1734 * option, and in this case it only works on functions tagged with
1735 * **ALLOW_ERROR_INJECTION** in the kernel code.
1737 * Also, the helper is only available for the architectures having
1738 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1739 * x86 architecture is the only one to support this feature.
1743 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1745 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1746 * for the full TCP socket associated to *bpf_sock_ops* to
1749 * The primary use of this field is to determine if there should
1750 * be calls to eBPF programs of type
1751 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1752 * code. A program of the same type can change its value, per
1753 * connection and as necessary, when the connection is
1754 * established. This field is directly accessible for reading, but
1755 * this helper must be used for updates in order to return an
1756 * error if an eBPF program tries to set a callback that is not
1757 * supported in the current kernel.
1759 * *argval* is a flag array which can combine these flags:
1761 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1762 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1763 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1764 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1766 * Therefore, this function can be used to clear a callback flag by
1767 * setting the appropriate bit to zero. e.g. to disable the RTO
1770 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1771 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1773 * Here are some examples of where one could call such eBPF
1777 * * When a packet is retransmitted.
1778 * * When the connection terminates.
1779 * * When a packet is sent.
1780 * * When a packet is received.
1782 * Code **-EINVAL** if the socket is not a full TCP socket;
1783 * otherwise, a positive number containing the bits that could not
1784 * be set is returned (which comes down to 0 if all bits were set
1787 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1789 * This helper is used in programs implementing policies at the
1790 * socket level. If the message *msg* is allowed to pass (i.e. if
1791 * the verdict eBPF program returns **SK_PASS**), redirect it to
1792 * the socket referenced by *map* (of type
1793 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1794 * egress interfaces can be used for redirection. The
1795 * **BPF_F_INGRESS** value in *flags* is used to make the
1796 * distinction (ingress path is selected if the flag is present,
1797 * egress path otherwise). This is the only flag supported for now.
1799 * **SK_PASS** on success, or **SK_DROP** on error.
1801 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1803 * For socket policies, apply the verdict of the eBPF program to
1804 * the next *bytes* (number of bytes) of message *msg*.
1806 * For example, this helper can be used in the following cases:
1808 * * A single **sendmsg**\ () or **sendfile**\ () system call
1809 * contains multiple logical messages that the eBPF program is
1810 * supposed to read and for which it should apply a verdict.
1811 * * An eBPF program only cares to read the first *bytes* of a
1812 * *msg*. If the message has a large payload, then setting up
1813 * and calling the eBPF program repeatedly for all bytes, even
1814 * though the verdict is already known, would create unnecessary
1817 * When called from within an eBPF program, the helper sets a
1818 * counter internal to the BPF infrastructure, that is used to
1819 * apply the last verdict to the next *bytes*. If *bytes* is
1820 * smaller than the current data being processed from a
1821 * **sendmsg**\ () or **sendfile**\ () system call, the first
1822 * *bytes* will be sent and the eBPF program will be re-run with
1823 * the pointer for start of data pointing to byte number *bytes*
1824 * **+ 1**. If *bytes* is larger than the current data being
1825 * processed, then the eBPF verdict will be applied to multiple
1826 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1829 * Note that if a socket closes with the internal counter holding
1830 * a non-zero value, this is not a problem because data is not
1831 * being buffered for *bytes* and is sent as it is received.
1835 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1837 * For socket policies, prevent the execution of the verdict eBPF
1838 * program for message *msg* until *bytes* (byte number) have been
1841 * This can be used when one needs a specific number of bytes
1842 * before a verdict can be assigned, even if the data spans
1843 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1844 * case would be a user calling **sendmsg**\ () repeatedly with
1845 * 1-byte long message segments. Obviously, this is bad for
1846 * performance, but it is still valid. If the eBPF program needs
1847 * *bytes* bytes to validate a header, this helper can be used to
1848 * prevent the eBPF program to be called again until *bytes* have
1853 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1855 * For socket policies, pull in non-linear data from user space
1856 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1857 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1860 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1861 * *msg* it can only parse data that the (**data**, **data_end**)
1862 * pointers have already consumed. For **sendmsg**\ () hooks this
1863 * is likely the first scatterlist element. But for calls relying
1864 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1865 * be the range (**0**, **0**) because the data is shared with
1866 * user space and by default the objective is to avoid allowing
1867 * user space to modify data while (or after) eBPF verdict is
1868 * being decided. This helper can be used to pull in data and to
1869 * set the start and end pointer to given values. Data will be
1870 * copied if necessary (i.e. if data was not linear and if start
1871 * and end pointers do not point to the same chunk).
1873 * A call to this helper is susceptible to change the underlying
1874 * packet buffer. Therefore, at load time, all checks on pointers
1875 * previously done by the verifier are invalidated and must be
1876 * performed again, if the helper is used in combination with
1877 * direct packet access.
1879 * All values for *flags* are reserved for future usage, and must
1882 * 0 on success, or a negative error in case of failure.
1884 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1886 * Bind the socket associated to *ctx* to the address pointed by
1887 * *addr*, of length *addr_len*. This allows for making outgoing
1888 * connection from the desired IP address, which can be useful for
1889 * example when all processes inside a cgroup should use one
1890 * single IP address on a host that has multiple IP configured.
1892 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1893 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1894 * **AF_INET6**). Looking for a free port to bind to can be
1895 * expensive, therefore binding to port is not permitted by the
1896 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1897 * must be set to zero.
1899 * 0 on success, or a negative error in case of failure.
1901 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1903 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1904 * only possible to shrink the packet as of this writing,
1905 * therefore *delta* must be a negative integer.
1907 * A call to this helper is susceptible to change the underlying
1908 * packet buffer. Therefore, at load time, all checks on pointers
1909 * previously done by the verifier are invalidated and must be
1910 * performed again, if the helper is used in combination with
1911 * direct packet access.
1913 * 0 on success, or a negative error in case of failure.
1915 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1917 * Retrieve the XFRM state (IP transform framework, see also
1918 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1920 * The retrieved value is stored in the **struct bpf_xfrm_state**
1921 * pointed by *xfrm_state* and of length *size*.
1923 * All values for *flags* are reserved for future usage, and must
1926 * This helper is available only if the kernel was compiled with
1927 * **CONFIG_XFRM** configuration option.
1929 * 0 on success, or a negative error in case of failure.
1931 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
1933 * Return a user or a kernel stack in bpf program provided buffer.
1934 * To achieve this, the helper needs *ctx*, which is a pointer
1935 * to the context on which the tracing program is executed.
1936 * To store the stacktrace, the bpf program provides *buf* with
1937 * a nonnegative *size*.
1939 * The last argument, *flags*, holds the number of stack frames to
1940 * skip (from 0 to 255), masked with
1941 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1942 * the following flags:
1944 * **BPF_F_USER_STACK**
1945 * Collect a user space stack instead of a kernel stack.
1946 * **BPF_F_USER_BUILD_ID**
1947 * Collect buildid+offset instead of ips for user stack,
1948 * only valid if **BPF_F_USER_STACK** is also specified.
1950 * **bpf_get_stack**\ () can collect up to
1951 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1952 * to sufficient large buffer size. Note that
1953 * this limit can be controlled with the **sysctl** program, and
1954 * that it should be manually increased in order to profile long
1955 * user stacks (such as stacks for Java programs). To do so, use:
1959 * # sysctl kernel.perf_event_max_stack=<new value>
1961 * A non-negative value equal to or less than *size* on success,
1962 * or a negative error in case of failure.
1964 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
1966 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1967 * it provides an easy way to load *len* bytes from *offset*
1968 * from the packet associated to *skb*, into the buffer pointed
1969 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1970 * a fifth argument *start_header* exists in order to select a
1971 * base offset to start from. *start_header* can be one of:
1973 * **BPF_HDR_START_MAC**
1974 * Base offset to load data from is *skb*'s mac header.
1975 * **BPF_HDR_START_NET**
1976 * Base offset to load data from is *skb*'s network header.
1978 * In general, "direct packet access" is the preferred method to
1979 * access packet data, however, this helper is in particular useful
1980 * in socket filters where *skb*\ **->data** does not always point
1981 * to the start of the mac header and where "direct packet access"
1984 * 0 on success, or a negative error in case of failure.
1986 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1988 * Do FIB lookup in kernel tables using parameters in *params*.
1989 * If lookup is successful and result shows packet is to be
1990 * forwarded, the neighbor tables are searched for the nexthop.
1991 * If successful (ie., FIB lookup shows forwarding and nexthop
1992 * is resolved), the nexthop address is returned in ipv4_dst
1993 * or ipv6_dst based on family, smac is set to mac address of
1994 * egress device, dmac is set to nexthop mac address, rt_metric
1995 * is set to metric from route (IPv4/IPv6 only), and ifindex
1996 * is set to the device index of the nexthop from the FIB lookup.
1998 * *plen* argument is the size of the passed in struct.
1999 * *flags* argument can be a combination of one or more of the
2002 * **BPF_FIB_LOOKUP_DIRECT**
2003 * Do a direct table lookup vs full lookup using FIB
2005 * **BPF_FIB_LOOKUP_OUTPUT**
2006 * Perform lookup from an egress perspective (default is
2009 * *ctx* is either **struct xdp_md** for XDP programs or
2010 * **struct sk_buff** tc cls_act programs.
2012 * * < 0 if any input argument is invalid
2013 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2014 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2015 * packet is not forwarded or needs assist from full stack
2017 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2019 * Add an entry to, or update a sockhash *map* referencing sockets.
2020 * The *skops* is used as a new value for the entry associated to
2021 * *key*. *flags* is one of:
2024 * The entry for *key* must not exist in the map.
2026 * The entry for *key* must already exist in the map.
2028 * No condition on the existence of the entry for *key*.
2030 * If the *map* has eBPF programs (parser and verdict), those will
2031 * be inherited by the socket being added. If the socket is
2032 * already attached to eBPF programs, this results in an error.
2034 * 0 on success, or a negative error in case of failure.
2036 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2038 * This helper is used in programs implementing policies at the
2039 * socket level. If the message *msg* is allowed to pass (i.e. if
2040 * the verdict eBPF program returns **SK_PASS**), redirect it to
2041 * the socket referenced by *map* (of type
2042 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2043 * egress interfaces can be used for redirection. The
2044 * **BPF_F_INGRESS** value in *flags* is used to make the
2045 * distinction (ingress path is selected if the flag is present,
2046 * egress path otherwise). This is the only flag supported for now.
2048 * **SK_PASS** on success, or **SK_DROP** on error.
2050 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2052 * This helper is used in programs implementing policies at the
2053 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2054 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2055 * to the socket referenced by *map* (of type
2056 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2057 * egress interfaces can be used for redirection. The
2058 * **BPF_F_INGRESS** value in *flags* is used to make the
2059 * distinction (ingress path is selected if the flag is present,
2060 * egress otherwise). This is the only flag supported for now.
2062 * **SK_PASS** on success, or **SK_DROP** on error.
2064 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2066 * Encapsulate the packet associated to *skb* within a Layer 3
2067 * protocol header. This header is provided in the buffer at
2068 * address *hdr*, with *len* its size in bytes. *type* indicates
2069 * the protocol of the header and can be one of:
2071 * **BPF_LWT_ENCAP_SEG6**
2072 * IPv6 encapsulation with Segment Routing Header
2073 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2074 * the IPv6 header is computed by the kernel.
2075 * **BPF_LWT_ENCAP_SEG6_INLINE**
2076 * Only works if *skb* contains an IPv6 packet. Insert a
2077 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2079 * **BPF_LWT_ENCAP_IP**
2080 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2081 * must be IPv4 or IPv6, followed by zero or more
2082 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2083 * total bytes in all prepended headers. Please note that
2084 * if **skb_is_gso**\ (*skb*) is true, no more than two
2085 * headers can be prepended, and the inner header, if
2086 * present, should be either GRE or UDP/GUE.
2088 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2089 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2090 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2091 * **BPF_PROG_TYPE_LWT_XMIT**.
2093 * A call to this helper is susceptible to change the underlying
2094 * packet buffer. Therefore, at load time, all checks on pointers
2095 * previously done by the verifier are invalidated and must be
2096 * performed again, if the helper is used in combination with
2097 * direct packet access.
2099 * 0 on success, or a negative error in case of failure.
2101 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2103 * Store *len* bytes from address *from* into the packet
2104 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2105 * inside the outermost IPv6 Segment Routing Header can be
2106 * modified through this helper.
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_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2118 * Adjust the size allocated to TLVs in the outermost IPv6
2119 * Segment Routing Header contained in the packet associated to
2120 * *skb*, at position *offset* by *delta* bytes. Only offsets
2121 * after the segments are accepted. *delta* can be as well
2122 * positive (growing) as negative (shrinking).
2124 * A call to this helper is susceptible to change the underlying
2125 * packet buffer. Therefore, at load time, all checks on pointers
2126 * previously done by the verifier are invalidated and must be
2127 * performed again, if the helper is used in combination with
2128 * direct packet access.
2130 * 0 on success, or a negative error in case of failure.
2132 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2134 * Apply an IPv6 Segment Routing action of type *action* to the
2135 * packet associated to *skb*. Each action takes a parameter
2136 * contained at address *param*, and of length *param_len* bytes.
2137 * *action* can be one of:
2139 * **SEG6_LOCAL_ACTION_END_X**
2140 * End.X action: Endpoint with Layer-3 cross-connect.
2141 * Type of *param*: **struct in6_addr**.
2142 * **SEG6_LOCAL_ACTION_END_T**
2143 * End.T action: Endpoint with specific IPv6 table lookup.
2144 * Type of *param*: **int**.
2145 * **SEG6_LOCAL_ACTION_END_B6**
2146 * End.B6 action: Endpoint bound to an SRv6 policy.
2147 * Type of *param*: **struct ipv6_sr_hdr**.
2148 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2149 * End.B6.Encap action: Endpoint bound to an SRv6
2150 * encapsulation policy.
2151 * Type of *param*: **struct ipv6_sr_hdr**.
2153 * A call to this helper is susceptible to change the underlying
2154 * packet buffer. Therefore, at load time, all checks on pointers
2155 * previously done by the verifier are invalidated and must be
2156 * performed again, if the helper is used in combination with
2157 * direct packet access.
2159 * 0 on success, or a negative error in case of failure.
2161 * int bpf_rc_repeat(void *ctx)
2163 * This helper is used in programs implementing IR decoding, to
2164 * report a successfully decoded repeat key message. This delays
2165 * the generation of a key up event for previously generated
2168 * Some IR protocols like NEC have a special IR message for
2169 * repeating last button, for when a button is held down.
2171 * The *ctx* should point to the lirc sample as passed into
2174 * This helper is only available is the kernel was compiled with
2175 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2180 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2182 * This helper is used in programs implementing IR decoding, to
2183 * report a successfully decoded key press with *scancode*,
2184 * *toggle* value in the given *protocol*. The scancode will be
2185 * translated to a keycode using the rc keymap, and reported as
2186 * an input key down event. After a period a key up event is
2187 * generated. This period can be extended by calling either
2188 * **bpf_rc_keydown**\ () again with the same values, or calling
2189 * **bpf_rc_repeat**\ ().
2191 * Some protocols include a toggle bit, in case the button was
2192 * released and pressed again between consecutive scancodes.
2194 * The *ctx* should point to the lirc sample as passed into
2197 * The *protocol* is the decoded protocol number (see
2198 * **enum rc_proto** for some predefined values).
2200 * This helper is only available is the kernel was compiled with
2201 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2206 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2208 * Return the cgroup v2 id of the socket associated with the *skb*.
2209 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2210 * helper for cgroup v1 by providing a tag resp. identifier that
2211 * can be matched on or used for map lookups e.g. to implement
2212 * policy. The cgroup v2 id of a given path in the hierarchy is
2213 * exposed in user space through the f_handle API in order to get
2214 * to the same 64-bit id.
2216 * This helper can be used on TC egress path, but not on ingress,
2217 * and is available only if the kernel was compiled with the
2218 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2220 * The id is returned or 0 in case the id could not be retrieved.
2222 * u64 bpf_get_current_cgroup_id(void)
2224 * A 64-bit integer containing the current cgroup id based
2225 * on the cgroup within which the current task is running.
2227 * void *bpf_get_local_storage(void *map, u64 flags)
2229 * Get the pointer to the local storage area.
2230 * The type and the size of the local storage is defined
2231 * by the *map* argument.
2232 * The *flags* meaning is specific for each map type,
2233 * and has to be 0 for cgroup local storage.
2235 * Depending on the BPF program type, a local storage area
2236 * can be shared between multiple instances of the BPF program,
2237 * running simultaneously.
2239 * A user should care about the synchronization by himself.
2240 * For example, by using the **BPF_STX_XADD** instruction to alter
2243 * A pointer to the local storage area.
2245 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2247 * Select a **SO_REUSEPORT** socket from a
2248 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2249 * It checks the selected socket is matching the incoming
2250 * request in the socket buffer.
2252 * 0 on success, or a negative error in case of failure.
2254 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2256 * Return id of cgroup v2 that is ancestor of cgroup associated
2257 * with the *skb* at the *ancestor_level*. The root cgroup is at
2258 * *ancestor_level* zero and each step down the hierarchy
2259 * increments the level. If *ancestor_level* == level of cgroup
2260 * associated with *skb*, then return value will be same as that
2261 * of **bpf_skb_cgroup_id**\ ().
2263 * The helper is useful to implement policies based on cgroups
2264 * that are upper in hierarchy than immediate cgroup associated
2267 * The format of returned id and helper limitations are same as in
2268 * **bpf_skb_cgroup_id**\ ().
2270 * The id is returned or 0 in case the id could not be retrieved.
2272 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2274 * Look for TCP socket matching *tuple*, optionally in a child
2275 * network namespace *netns*. The return value must be checked,
2276 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2278 * The *ctx* should point to the context of the program, such as
2279 * the skb or socket (depending on the hook in use). This is used
2280 * to determine the base network namespace for the lookup.
2282 * *tuple_size* must be one of:
2284 * **sizeof**\ (*tuple*\ **->ipv4**)
2285 * Look for an IPv4 socket.
2286 * **sizeof**\ (*tuple*\ **->ipv6**)
2287 * Look for an IPv6 socket.
2289 * If the *netns* is a negative signed 32-bit integer, then the
2290 * socket lookup table in the netns associated with the *ctx* will
2291 * will be used. For the TC hooks, this is the netns of the device
2292 * in the skb. For socket hooks, this is the netns of the socket.
2293 * If *netns* is any other signed 32-bit value greater than or
2294 * equal to zero then it specifies the ID of the netns relative to
2295 * the netns associated with the *ctx*. *netns* values beyond the
2296 * range of 32-bit integers are reserved for future use.
2298 * All values for *flags* are reserved for future usage, and must
2301 * This helper is available only if the kernel was compiled with
2302 * **CONFIG_NET** configuration option.
2304 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2305 * For sockets with reuseport option, the **struct bpf_sock**
2306 * result is from *reuse*\ **->socks**\ [] using the hash of the
2309 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2311 * Look for UDP socket matching *tuple*, optionally in a child
2312 * network namespace *netns*. The return value must be checked,
2313 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2315 * The *ctx* should point to the context of the program, such as
2316 * the skb or socket (depending on the hook in use). This is used
2317 * to determine the base network namespace for the lookup.
2319 * *tuple_size* must be one of:
2321 * **sizeof**\ (*tuple*\ **->ipv4**)
2322 * Look for an IPv4 socket.
2323 * **sizeof**\ (*tuple*\ **->ipv6**)
2324 * Look for an IPv6 socket.
2326 * If the *netns* is a negative signed 32-bit integer, then the
2327 * socket lookup table in the netns associated with the *ctx* will
2328 * will be used. For the TC hooks, this is the netns of the device
2329 * in the skb. For socket hooks, this is the netns of the socket.
2330 * If *netns* is any other signed 32-bit value greater than or
2331 * equal to zero then it specifies the ID of the netns relative to
2332 * the netns associated with the *ctx*. *netns* values beyond the
2333 * range of 32-bit integers are reserved for future use.
2335 * All values for *flags* are reserved for future usage, and must
2338 * This helper is available only if the kernel was compiled with
2339 * **CONFIG_NET** configuration option.
2341 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2342 * For sockets with reuseport option, the **struct bpf_sock**
2343 * result is from *reuse*\ **->socks**\ [] using the hash of the
2346 * int bpf_sk_release(struct bpf_sock *sock)
2348 * Release the reference held by *sock*. *sock* must be a
2349 * non-**NULL** pointer that was returned from
2350 * **bpf_sk_lookup_xxx**\ ().
2352 * 0 on success, or a negative error in case of failure.
2354 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2356 * Push an element *value* in *map*. *flags* is one of:
2359 * If the queue/stack is full, the oldest element is
2360 * removed to make room for this.
2362 * 0 on success, or a negative error in case of failure.
2364 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2366 * Pop an element from *map*.
2368 * 0 on success, or a negative error in case of failure.
2370 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2372 * Get an element from *map* without removing it.
2374 * 0 on success, or a negative error in case of failure.
2376 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2378 * For socket policies, insert *len* bytes into *msg* at offset
2381 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2382 * *msg* it may want to insert metadata or options into the *msg*.
2383 * This can later be read and used by any of the lower layer BPF
2386 * This helper may fail if under memory pressure (a malloc
2387 * fails) in these cases BPF programs will get an appropriate
2388 * error and BPF programs will need to handle them.
2390 * 0 on success, or a negative error in case of failure.
2392 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2394 * Will remove *len* bytes from a *msg* starting at byte *start*.
2395 * This may result in **ENOMEM** errors under certain situations if
2396 * an allocation and copy are required due to a full ring buffer.
2397 * However, the helper will try to avoid doing the allocation
2398 * if possible. Other errors can occur if input parameters are
2399 * invalid either due to *start* byte not being valid part of *msg*
2400 * payload and/or *pop* value being to large.
2402 * 0 on success, or a negative error in case of failure.
2404 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2406 * This helper is used in programs implementing IR decoding, to
2407 * report a successfully decoded pointer movement.
2409 * The *ctx* should point to the lirc sample as passed into
2412 * This helper is only available is the kernel was compiled with
2413 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2418 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2420 * Acquire a spinlock represented by the pointer *lock*, which is
2421 * stored as part of a value of a map. Taking the lock allows to
2422 * safely update the rest of the fields in that value. The
2423 * spinlock can (and must) later be released with a call to
2424 * **bpf_spin_unlock**\ (\ *lock*\ ).
2426 * Spinlocks in BPF programs come with a number of restrictions
2429 * * **bpf_spin_lock** objects are only allowed inside maps of
2430 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2431 * list could be extended in the future).
2432 * * BTF description of the map is mandatory.
2433 * * The BPF program can take ONE lock at a time, since taking two
2434 * or more could cause dead locks.
2435 * * Only one **struct bpf_spin_lock** is allowed per map element.
2436 * * When the lock is taken, calls (either BPF to BPF or helpers)
2438 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2439 * allowed inside a spinlock-ed region.
2440 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2441 * the lock, on all execution paths, before it returns.
2442 * * The BPF program can access **struct bpf_spin_lock** only via
2443 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2444 * helpers. Loading or storing data into the **struct
2445 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2446 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2447 * of the map value must be a struct and have **struct
2448 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2449 * Nested lock inside another struct is not allowed.
2450 * * The **struct bpf_spin_lock** *lock* field in a map value must
2451 * be aligned on a multiple of 4 bytes in that value.
2452 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2453 * the **bpf_spin_lock** field to user space.
2454 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2455 * a BPF program, do not update the **bpf_spin_lock** field.
2456 * * **bpf_spin_lock** cannot be on the stack or inside a
2457 * networking packet (it can only be inside of a map values).
2458 * * **bpf_spin_lock** is available to root only.
2459 * * Tracing programs and socket filter programs cannot use
2460 * **bpf_spin_lock**\ () due to insufficient preemption checks
2461 * (but this may change in the future).
2462 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2466 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2468 * Release the *lock* previously locked by a call to
2469 * **bpf_spin_lock**\ (\ *lock*\ ).
2473 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2475 * This helper gets a **struct bpf_sock** pointer such
2476 * that all the fields in this **bpf_sock** can be accessed.
2478 * A **struct bpf_sock** pointer on success, or **NULL** in
2481 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2483 * This helper gets a **struct bpf_tcp_sock** pointer from a
2484 * **struct bpf_sock** pointer.
2486 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2489 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
2491 * Set ECN (Explicit Congestion Notification) field of IP header
2492 * to **CE** (Congestion Encountered) if current value is **ECT**
2493 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2496 * 1 if the **CE** flag is set (either by the current helper call
2497 * or because it was already present), 0 if it is not set.
2499 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2501 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2502 * **bpf_sk_release**\ () is unnecessary and not allowed.
2504 * A **struct bpf_sock** pointer on success, or **NULL** in
2507 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2509 * Look for TCP socket matching *tuple*, optionally in a child
2510 * network namespace *netns*. The return value must be checked,
2511 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2513 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2514 * that it also returns timewait or request sockets. Use
2515 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2518 * This helper is available only if the kernel was compiled with
2519 * **CONFIG_NET** configuration option.
2521 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2522 * For sockets with reuseport option, the **struct bpf_sock**
2523 * result is from *reuse*\ **->socks**\ [] using the hash of the
2526 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2528 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2529 * the listening socket in *sk*.
2531 * *iph* points to the start of the IPv4 or IPv6 header, while
2532 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2533 * **sizeof**\ (**struct ip6hdr**).
2535 * *th* points to the start of the TCP header, while *th_len*
2536 * contains **sizeof**\ (**struct tcphdr**).
2539 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2542 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2544 * Get name of sysctl in /proc/sys/ and copy it into provided by
2545 * program buffer *buf* of size *buf_len*.
2547 * The buffer is always NUL terminated, unless it's zero-sized.
2549 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2550 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2551 * only (e.g. "tcp_mem").
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 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2560 * Get current value of sysctl as it is presented in /proc/sys
2561 * (incl. newline, etc), and copy it as a string into provided
2562 * by program buffer *buf* of size *buf_len*.
2564 * The whole value is copied, no matter what file position user
2565 * space issued e.g. sys_read at.
2567 * The buffer is always NUL terminated, unless it's zero-sized.
2569 * Number of character copied (not including the trailing NUL).
2571 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2572 * truncated name in this case).
2574 * **-EINVAL** if current value was unavailable, e.g. because
2575 * sysctl is uninitialized and read returns -EIO for it.
2577 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2579 * Get new value being written by user space to sysctl (before
2580 * the actual write happens) and copy it as a string into
2581 * provided by program buffer *buf* of size *buf_len*.
2583 * User space may write new value at file position > 0.
2585 * The buffer is always NUL terminated, unless it's zero-sized.
2587 * Number of character copied (not including the trailing NUL).
2589 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2590 * truncated name in this case).
2592 * **-EINVAL** if sysctl is being read.
2594 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2596 * Override new value being written by user space to sysctl with
2597 * value provided by program in buffer *buf* of size *buf_len*.
2599 * *buf* should contain a string in same form as provided by user
2600 * space on sysctl write.
2602 * User space may write new value at file position > 0. To override
2603 * the whole sysctl value file position should be set to zero.
2607 * **-E2BIG** if the *buf_len* is too big.
2609 * **-EINVAL** if sysctl is being read.
2611 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2613 * Convert the initial part of the string from buffer *buf* of
2614 * size *buf_len* to a long integer according to the given base
2615 * and save the result in *res*.
2617 * The string may begin with an arbitrary amount of white space
2618 * (as determined by **isspace**\ (3)) followed by a single
2619 * optional '**-**' sign.
2621 * Five least significant bits of *flags* encode base, other bits
2622 * are currently unused.
2624 * Base must be either 8, 10, 16 or 0 to detect it automatically
2625 * similar to user space **strtol**\ (3).
2627 * Number of characters consumed on success. Must be positive but
2628 * no more than *buf_len*.
2630 * **-EINVAL** if no valid digits were found or unsupported base
2633 * **-ERANGE** if resulting value was out of range.
2635 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2637 * Convert the initial part of the string from buffer *buf* of
2638 * size *buf_len* to an unsigned long integer according to the
2639 * given base and save the result in *res*.
2641 * The string may begin with an arbitrary amount of white space
2642 * (as determined by **isspace**\ (3)).
2644 * Five least significant bits of *flags* encode base, other bits
2645 * are currently unused.
2647 * Base must be either 8, 10, 16 or 0 to detect it automatically
2648 * similar to user space **strtoul**\ (3).
2650 * Number of characters consumed on success. Must be positive but
2651 * no more than *buf_len*.
2653 * **-EINVAL** if no valid digits were found or unsupported base
2656 * **-ERANGE** if resulting value was out of range.
2658 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2660 * Get a bpf-local-storage from a *sk*.
2662 * Logically, it could be thought of getting the value from
2663 * a *map* with *sk* as the **key**. From this
2664 * perspective, the usage is not much different from
2665 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2666 * helper enforces the key must be a full socket and the map must
2667 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2669 * Underneath, the value is stored locally at *sk* instead of
2670 * the *map*. The *map* is used as the bpf-local-storage
2671 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2672 * searched against all bpf-local-storages residing at *sk*.
2674 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2675 * used such that a new bpf-local-storage will be
2676 * created if one does not exist. *value* can be used
2677 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2678 * the initial value of a bpf-local-storage. If *value* is
2679 * **NULL**, the new bpf-local-storage will be zero initialized.
2681 * A bpf-local-storage pointer is returned on success.
2683 * **NULL** if not found or there was an error in adding
2684 * a new bpf-local-storage.
2686 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2688 * Delete a bpf-local-storage from a *sk*.
2692 * **-ENOENT** if the bpf-local-storage cannot be found.
2694 * int bpf_send_signal(u32 sig)
2696 * Send signal *sig* to the current task.
2698 * 0 on success or successfully queued.
2700 * **-EBUSY** if work queue under nmi is full.
2702 * **-EINVAL** if *sig* is invalid.
2704 * **-EPERM** if no permission to send the *sig*.
2706 * **-EAGAIN** if bpf program can try again.
2708 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2710 * Try to issue a SYN cookie for the packet with corresponding
2711 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2713 * *iph* points to the start of the IPv4 or IPv6 header, while
2714 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2715 * **sizeof**\ (**struct ip6hdr**).
2717 * *th* points to the start of the TCP header, while *th_len*
2718 * contains the length of the TCP header.
2721 * On success, lower 32 bits hold the generated SYN cookie in
2722 * followed by 16 bits which hold the MSS value for that cookie,
2723 * and the top 16 bits are unused.
2725 * On failure, the returned value is one of the following:
2727 * **-EINVAL** SYN cookie cannot be issued due to error
2729 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2731 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2733 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2735 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2737 * Write raw *data* blob into a special BPF perf event held by
2738 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2739 * event must have the following attributes: **PERF_SAMPLE_RAW**
2740 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2741 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2743 * The *flags* are used to indicate the index in *map* for which
2744 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2745 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2746 * to indicate that the index of the current CPU core should be
2749 * The value to write, of *size*, is passed through eBPF stack and
2750 * pointed by *data*.
2752 * *ctx* is a pointer to in-kernel struct sk_buff.
2754 * This helper is similar to **bpf_perf_event_output**\ () but
2755 * restricted to raw_tracepoint bpf programs.
2757 * 0 on success, or a negative error in case of failure.
2759 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2761 * Safely attempt to read *size* bytes from user space address
2762 * *unsafe_ptr* and store the data in *dst*.
2764 * 0 on success, or a negative error in case of failure.
2766 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2768 * Safely attempt to read *size* bytes from kernel space address
2769 * *unsafe_ptr* and store the data in *dst*.
2771 * 0 on success, or a negative error in case of failure.
2773 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2775 * Copy a NUL terminated string from an unsafe user address
2776 * *unsafe_ptr* to *dst*. The *size* should include the
2777 * terminating NUL byte. In case the string length is smaller than
2778 * *size*, the target is not padded with further NUL bytes. If the
2779 * string length is larger than *size*, just *size*-1 bytes are
2780 * copied and the last byte is set to NUL.
2782 * On success, the length of the copied string is returned. This
2783 * makes this helper useful in tracing programs for reading
2784 * strings, and more importantly to get its length at runtime. See
2785 * the following snippet:
2789 * SEC("kprobe/sys_open")
2790 * void bpf_sys_open(struct pt_regs *ctx)
2792 * char buf[PATHLEN]; // PATHLEN is defined to 256
2793 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2796 * // Consume buf, for example push it to
2797 * // userspace via bpf_perf_event_output(); we
2798 * // can use res (the string length) as event
2799 * // size, after checking its boundaries.
2802 * In comparison, using **bpf_probe_read_user()** helper here
2803 * instead to read the string would require to estimate the length
2804 * at compile time, and would often result in copying more memory
2807 * Another useful use case is when parsing individual process
2808 * arguments or individual environment variables navigating
2809 * *current*\ **->mm->arg_start** and *current*\
2810 * **->mm->env_start**: using this helper and the return value,
2811 * one can quickly iterate at the right offset of the memory area.
2813 * On success, the strictly positive length of the string,
2814 * including the trailing NUL character. On error, a negative
2817 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2819 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
2820 * to *dst*. Same semantics as with bpf_probe_read_user_str() apply.
2822 * On success, the strictly positive length of the string, including
2823 * the trailing NUL character. On error, a negative value.
2825 #define __BPF_FUNC_MAPPER(FN) \
2827 FN(map_lookup_elem), \
2828 FN(map_update_elem), \
2829 FN(map_delete_elem), \
2833 FN(get_prandom_u32), \
2834 FN(get_smp_processor_id), \
2835 FN(skb_store_bytes), \
2836 FN(l3_csum_replace), \
2837 FN(l4_csum_replace), \
2839 FN(clone_redirect), \
2840 FN(get_current_pid_tgid), \
2841 FN(get_current_uid_gid), \
2842 FN(get_current_comm), \
2843 FN(get_cgroup_classid), \
2844 FN(skb_vlan_push), \
2846 FN(skb_get_tunnel_key), \
2847 FN(skb_set_tunnel_key), \
2848 FN(perf_event_read), \
2850 FN(get_route_realm), \
2851 FN(perf_event_output), \
2852 FN(skb_load_bytes), \
2855 FN(skb_get_tunnel_opt), \
2856 FN(skb_set_tunnel_opt), \
2857 FN(skb_change_proto), \
2858 FN(skb_change_type), \
2859 FN(skb_under_cgroup), \
2860 FN(get_hash_recalc), \
2861 FN(get_current_task), \
2862 FN(probe_write_user), \
2863 FN(current_task_under_cgroup), \
2864 FN(skb_change_tail), \
2865 FN(skb_pull_data), \
2867 FN(set_hash_invalid), \
2868 FN(get_numa_node_id), \
2869 FN(skb_change_head), \
2870 FN(xdp_adjust_head), \
2871 FN(probe_read_str), \
2872 FN(get_socket_cookie), \
2873 FN(get_socket_uid), \
2876 FN(skb_adjust_room), \
2878 FN(sk_redirect_map), \
2879 FN(sock_map_update), \
2880 FN(xdp_adjust_meta), \
2881 FN(perf_event_read_value), \
2882 FN(perf_prog_read_value), \
2884 FN(override_return), \
2885 FN(sock_ops_cb_flags_set), \
2886 FN(msg_redirect_map), \
2887 FN(msg_apply_bytes), \
2888 FN(msg_cork_bytes), \
2889 FN(msg_pull_data), \
2891 FN(xdp_adjust_tail), \
2892 FN(skb_get_xfrm_state), \
2894 FN(skb_load_bytes_relative), \
2896 FN(sock_hash_update), \
2897 FN(msg_redirect_hash), \
2898 FN(sk_redirect_hash), \
2899 FN(lwt_push_encap), \
2900 FN(lwt_seg6_store_bytes), \
2901 FN(lwt_seg6_adjust_srh), \
2902 FN(lwt_seg6_action), \
2905 FN(skb_cgroup_id), \
2906 FN(get_current_cgroup_id), \
2907 FN(get_local_storage), \
2908 FN(sk_select_reuseport), \
2909 FN(skb_ancestor_cgroup_id), \
2910 FN(sk_lookup_tcp), \
2911 FN(sk_lookup_udp), \
2913 FN(map_push_elem), \
2915 FN(map_peek_elem), \
2916 FN(msg_push_data), \
2918 FN(rc_pointer_rel), \
2923 FN(skb_ecn_set_ce), \
2924 FN(get_listener_sock), \
2925 FN(skc_lookup_tcp), \
2926 FN(tcp_check_syncookie), \
2927 FN(sysctl_get_name), \
2928 FN(sysctl_get_current_value), \
2929 FN(sysctl_get_new_value), \
2930 FN(sysctl_set_new_value), \
2933 FN(sk_storage_get), \
2934 FN(sk_storage_delete), \
2936 FN(tcp_gen_syncookie), \
2938 FN(probe_read_user), \
2939 FN(probe_read_kernel), \
2940 FN(probe_read_user_str), \
2941 FN(probe_read_kernel_str),
2943 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2944 * function eBPF program intends to call
2946 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2948 __BPF_FUNC_MAPPER(__BPF_ENUM_FN
)
2951 #undef __BPF_ENUM_FN
2953 /* All flags used by eBPF helper functions, placed here. */
2955 /* BPF_FUNC_skb_store_bytes flags. */
2956 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2957 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2959 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2960 * First 4 bits are for passing the header field size.
2962 #define BPF_F_HDR_FIELD_MASK 0xfULL
2964 /* BPF_FUNC_l4_csum_replace flags. */
2965 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2966 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2967 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2969 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2970 #define BPF_F_INGRESS (1ULL << 0)
2972 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2973 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2975 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2976 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2977 #define BPF_F_USER_STACK (1ULL << 8)
2978 /* flags used by BPF_FUNC_get_stackid only. */
2979 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2980 #define BPF_F_REUSE_STACKID (1ULL << 10)
2981 /* flags used by BPF_FUNC_get_stack only. */
2982 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2984 /* BPF_FUNC_skb_set_tunnel_key flags. */
2985 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2986 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2987 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2989 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2990 * BPF_FUNC_perf_event_read_value flags.
2992 #define BPF_F_INDEX_MASK 0xffffffffULL
2993 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2994 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2995 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2997 /* Current network namespace */
2998 #define BPF_F_CURRENT_NETNS (-1L)
3000 /* BPF_FUNC_skb_adjust_room flags. */
3001 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
3003 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
3004 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
3006 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
3007 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
3008 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
3009 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
3010 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
3011 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3012 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
3014 /* BPF_FUNC_sysctl_get_name flags. */
3015 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
3017 /* BPF_FUNC_sk_storage_get flags */
3018 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0)
3020 /* Mode for BPF_FUNC_skb_adjust_room helper. */
3021 enum bpf_adj_room_mode
{
3026 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3027 enum bpf_hdr_start_off
{
3032 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3033 enum bpf_lwt_encap_mode
{
3035 BPF_LWT_ENCAP_SEG6_INLINE
,
3039 #define __bpf_md_ptr(type, name) \
3043 } __attribute__((aligned(8)))
3045 /* user accessible mirror of in-kernel sk_buff.
3046 * new fields can only be added to the end of this structure
3052 __u32 queue_mapping
;
3058 __u32 ingress_ifindex
;
3068 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
3070 __u32 remote_ip4
; /* Stored in network byte order */
3071 __u32 local_ip4
; /* Stored in network byte order */
3072 __u32 remote_ip6
[4]; /* Stored in network byte order */
3073 __u32 local_ip6
[4]; /* Stored in network byte order */
3074 __u32 remote_port
; /* Stored in network byte order */
3075 __u32 local_port
; /* stored in host byte order */
3079 __bpf_md_ptr(struct bpf_flow_keys
*, flow_keys
);
3083 __bpf_md_ptr(struct bpf_sock
*, sk
);
3086 struct bpf_tunnel_key
{
3090 __u32 remote_ipv6
[4];
3094 __u16 tunnel_ext
; /* Padding, future use. */
3098 /* user accessible mirror of in-kernel xfrm_state.
3099 * new fields can only be added to the end of this structure
3101 struct bpf_xfrm_state
{
3103 __u32 spi
; /* Stored in network byte order */
3105 __u16 ext
; /* Padding, future use. */
3107 __u32 remote_ipv4
; /* Stored in network byte order */
3108 __u32 remote_ipv6
[4]; /* Stored in network byte order */
3112 /* Generic BPF return codes which all BPF program types may support.
3113 * The values are binary compatible with their TC_ACT_* counter-part to
3114 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3117 * XDP is handled seprately, see XDP_*.
3125 /* >127 are reserved for prog type specific return codes.
3127 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3128 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3129 * changed and should be routed based on its new L3 header.
3130 * (This is an L3 redirect, as opposed to L2 redirect
3131 * represented by BPF_REDIRECT above).
3133 BPF_LWT_REROUTE
= 128,
3143 /* IP address also allows 1 and 2 bytes access */
3146 __u32 src_port
; /* host byte order */
3147 __u32 dst_port
; /* network byte order */
3153 struct bpf_tcp_sock
{
3154 __u32 snd_cwnd
; /* Sending congestion window */
3155 __u32 srtt_us
; /* smoothed round trip time << 3 in usecs */
3157 __u32 snd_ssthresh
; /* Slow start size threshold */
3158 __u32 rcv_nxt
; /* What we want to receive next */
3159 __u32 snd_nxt
; /* Next sequence we send */
3160 __u32 snd_una
; /* First byte we want an ack for */
3161 __u32 mss_cache
; /* Cached effective mss, not including SACKS */
3162 __u32 ecn_flags
; /* ECN status bits. */
3163 __u32 rate_delivered
; /* saved rate sample: packets delivered */
3164 __u32 rate_interval_us
; /* saved rate sample: time elapsed */
3165 __u32 packets_out
; /* Packets which are "in flight" */
3166 __u32 retrans_out
; /* Retransmitted packets out */
3167 __u32 total_retrans
; /* Total retransmits for entire connection */
3168 __u32 segs_in
; /* RFC4898 tcpEStatsPerfSegsIn
3169 * total number of segments in.
3171 __u32 data_segs_in
; /* RFC4898 tcpEStatsPerfDataSegsIn
3172 * total number of data segments in.
3174 __u32 segs_out
; /* RFC4898 tcpEStatsPerfSegsOut
3175 * The total number of segments sent.
3177 __u32 data_segs_out
; /* RFC4898 tcpEStatsPerfDataSegsOut
3178 * total number of data segments sent.
3180 __u32 lost_out
; /* Lost packets */
3181 __u32 sacked_out
; /* SACK'd packets */
3182 __u64 bytes_received
; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3183 * sum(delta(rcv_nxt)), or how many bytes
3186 __u64 bytes_acked
; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3187 * sum(delta(snd_una)), or how many bytes
3190 __u32 dsack_dups
; /* RFC4898 tcpEStatsStackDSACKDups
3191 * total number of DSACK blocks received
3193 __u32 delivered
; /* Total data packets delivered incl. rexmits */
3194 __u32 delivered_ce
; /* Like the above but only ECE marked packets */
3195 __u32 icsk_retransmits
; /* Number of unrecovered [RTO] timeouts */
3198 struct bpf_sock_tuple
{
3215 struct bpf_xdp_sock
{
3219 #define XDP_PACKET_HEADROOM 256
3221 /* User return codes for XDP prog type.
3222 * A valid XDP program must return one of these defined values. All other
3223 * return codes are reserved for future use. Unknown return codes will
3224 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3234 /* user accessible metadata for XDP packet hook
3235 * new fields must be added to the end of this structure
3241 /* Below access go through struct xdp_rxq_info */
3242 __u32 ingress_ifindex
; /* rxq->dev->ifindex */
3243 __u32 rx_queue_index
; /* rxq->queue_index */
3251 /* user accessible metadata for SK_MSG packet hook, new fields must
3252 * be added to the end of this structure
3255 __bpf_md_ptr(void *, data
);
3256 __bpf_md_ptr(void *, data_end
);
3259 __u32 remote_ip4
; /* Stored in network byte order */
3260 __u32 local_ip4
; /* Stored in network byte order */
3261 __u32 remote_ip6
[4]; /* Stored in network byte order */
3262 __u32 local_ip6
[4]; /* Stored in network byte order */
3263 __u32 remote_port
; /* Stored in network byte order */
3264 __u32 local_port
; /* stored in host byte order */
3265 __u32 size
; /* Total size of sk_msg */
3268 struct sk_reuseport_md
{
3270 * Start of directly accessible data. It begins from
3271 * the tcp/udp header.
3273 __bpf_md_ptr(void *, data
);
3274 /* End of directly accessible data */
3275 __bpf_md_ptr(void *, data_end
);
3277 * Total length of packet (starting from the tcp/udp header).
3278 * Note that the directly accessible bytes (data_end - data)
3279 * could be less than this "len". Those bytes could be
3280 * indirectly read by a helper "bpf_skb_load_bytes()".
3284 * Eth protocol in the mac header (network byte order). e.g.
3285 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3288 __u32 ip_protocol
; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3289 __u32 bind_inany
; /* Is sock bound to an INANY address? */
3290 __u32 hash
; /* A hash of the packet 4 tuples */
3293 #define BPF_TAG_SIZE 8
3295 struct bpf_prog_info
{
3298 __u8 tag
[BPF_TAG_SIZE
];
3299 __u32 jited_prog_len
;
3300 __u32 xlated_prog_len
;
3301 __aligned_u64 jited_prog_insns
;
3302 __aligned_u64 xlated_prog_insns
;
3303 __u64 load_time
; /* ns since boottime */
3304 __u32 created_by_uid
;
3306 __aligned_u64 map_ids
;
3307 char name
[BPF_OBJ_NAME_LEN
];
3309 __u32 gpl_compatible
:1;
3310 __u32
:31; /* alignment pad */
3313 __u32 nr_jited_ksyms
;
3314 __u32 nr_jited_func_lens
;
3315 __aligned_u64 jited_ksyms
;
3316 __aligned_u64 jited_func_lens
;
3318 __u32 func_info_rec_size
;
3319 __aligned_u64 func_info
;
3322 __aligned_u64 line_info
;
3323 __aligned_u64 jited_line_info
;
3324 __u32 nr_jited_line_info
;
3325 __u32 line_info_rec_size
;
3326 __u32 jited_line_info_rec_size
;
3328 __aligned_u64 prog_tags
;
3331 } __attribute__((aligned(8)));
3333 struct bpf_map_info
{
3340 char name
[BPF_OBJ_NAME_LEN
];
3346 __u32 btf_key_type_id
;
3347 __u32 btf_value_type_id
;
3348 } __attribute__((aligned(8)));
3350 struct bpf_btf_info
{
3354 } __attribute__((aligned(8)));
3356 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3357 * by user and intended to be used by socket (e.g. to bind to, depends on
3358 * attach attach type).
3360 struct bpf_sock_addr
{
3361 __u32 user_family
; /* Allows 4-byte read, but no write. */
3362 __u32 user_ip4
; /* Allows 1,2,4-byte read and 4-byte write.
3363 * Stored in network byte order.
3365 __u32 user_ip6
[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3366 * Stored in network byte order.
3368 __u32 user_port
; /* Allows 4-byte read and write.
3369 * Stored in network byte order
3371 __u32 family
; /* Allows 4-byte read, but no write */
3372 __u32 type
; /* Allows 4-byte read, but no write */
3373 __u32 protocol
; /* Allows 4-byte read, but no write */
3374 __u32 msg_src_ip4
; /* Allows 1,2,4-byte read and 4-byte write.
3375 * Stored in network byte order.
3377 __u32 msg_src_ip6
[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3378 * Stored in network byte order.
3380 __bpf_md_ptr(struct bpf_sock
*, sk
);
3383 /* User bpf_sock_ops struct to access socket values and specify request ops
3384 * and their replies.
3385 * Some of this fields are in network (bigendian) byte order and may need
3386 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3387 * New fields can only be added at the end of this structure
3389 struct bpf_sock_ops
{
3392 __u32 args
[4]; /* Optionally passed to bpf program */
3393 __u32 reply
; /* Returned by bpf program */
3394 __u32 replylong
[4]; /* Optionally returned by bpf prog */
3397 __u32 remote_ip4
; /* Stored in network byte order */
3398 __u32 local_ip4
; /* Stored in network byte order */
3399 __u32 remote_ip6
[4]; /* Stored in network byte order */
3400 __u32 local_ip6
[4]; /* Stored in network byte order */
3401 __u32 remote_port
; /* Stored in network byte order */
3402 __u32 local_port
; /* stored in host byte order */
3403 __u32 is_fullsock
; /* Some TCP fields are only valid if
3404 * there is a full socket. If not, the
3405 * fields read as zero.
3408 __u32 srtt_us
; /* Averaged RTT << 3 in usecs */
3409 __u32 bpf_sock_ops_cb_flags
; /* flags defined in uapi/linux/tcp.h */
3418 __u32 rate_delivered
;
3419 __u32 rate_interval_us
;
3422 __u32 total_retrans
;
3426 __u32 data_segs_out
;
3430 __u64 bytes_received
;
3432 __bpf_md_ptr(struct bpf_sock
*, sk
);
3435 /* Definitions for bpf_sock_ops_cb_flags */
3436 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
3437 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
3438 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3439 #define BPF_SOCK_OPS_RTT_CB_FLAG (1<<3)
3440 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0xF /* Mask of all currently
3441 * supported cb flags
3444 /* List of known BPF sock_ops operators.
3445 * New entries can only be added at the end
3449 BPF_SOCK_OPS_TIMEOUT_INIT
, /* Should return SYN-RTO value to use or
3450 * -1 if default value should be used
3452 BPF_SOCK_OPS_RWND_INIT
, /* Should return initial advertized
3453 * window (in packets) or -1 if default
3454 * value should be used
3456 BPF_SOCK_OPS_TCP_CONNECT_CB
, /* Calls BPF program right before an
3457 * active connection is initialized
3459 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB
, /* Calls BPF program when an
3460 * active connection is
3463 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB
, /* Calls BPF program when a
3464 * passive connection is
3467 BPF_SOCK_OPS_NEEDS_ECN
, /* If connection's congestion control
3470 BPF_SOCK_OPS_BASE_RTT
, /* Get base RTT. The correct value is
3471 * based on the path and may be
3472 * dependent on the congestion control
3473 * algorithm. In general it indicates
3474 * a congestion threshold. RTTs above
3475 * this indicate congestion
3477 BPF_SOCK_OPS_RTO_CB
, /* Called when an RTO has triggered.
3478 * Arg1: value of icsk_retransmits
3479 * Arg2: value of icsk_rto
3480 * Arg3: whether RTO has expired
3482 BPF_SOCK_OPS_RETRANS_CB
, /* Called when skb is retransmitted.
3483 * Arg1: sequence number of 1st byte
3485 * Arg3: return value of
3486 * tcp_transmit_skb (0 => success)
3488 BPF_SOCK_OPS_STATE_CB
, /* Called when TCP changes state.
3492 BPF_SOCK_OPS_TCP_LISTEN_CB
, /* Called on listen(2), right after
3493 * socket transition to LISTEN state.
3495 BPF_SOCK_OPS_RTT_CB
, /* Called on every RTT.
3499 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3500 * changes between the TCP and BPF versions. Ideally this should never happen.
3501 * If it does, we need to add code to convert them before calling
3502 * the BPF sock_ops function.
3505 BPF_TCP_ESTABLISHED
= 1,
3515 BPF_TCP_CLOSING
, /* Now a valid state */
3516 BPF_TCP_NEW_SYN_RECV
,
3518 BPF_TCP_MAX_STATES
/* Leave at the end! */
3521 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3522 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3524 struct bpf_perf_event_value
{
3530 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3531 #define BPF_DEVCG_ACC_READ (1ULL << 1)
3532 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3534 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3535 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3537 struct bpf_cgroup_dev_ctx
{
3538 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3544 struct bpf_raw_tracepoint_args
{
3548 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3549 * OUTPUT: Do lookup from egress perspective; default is ingress
3551 #define BPF_FIB_LOOKUP_DIRECT (1U << 0)
3552 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
3555 BPF_FIB_LKUP_RET_SUCCESS
, /* lookup successful */
3556 BPF_FIB_LKUP_RET_BLACKHOLE
, /* dest is blackholed; can be dropped */
3557 BPF_FIB_LKUP_RET_UNREACHABLE
, /* dest is unreachable; can be dropped */
3558 BPF_FIB_LKUP_RET_PROHIBIT
, /* dest not allowed; can be dropped */
3559 BPF_FIB_LKUP_RET_NOT_FWDED
, /* packet is not forwarded */
3560 BPF_FIB_LKUP_RET_FWD_DISABLED
, /* fwding is not enabled on ingress */
3561 BPF_FIB_LKUP_RET_UNSUPP_LWT
, /* fwd requires encapsulation */
3562 BPF_FIB_LKUP_RET_NO_NEIGH
, /* no neighbor entry for nh */
3563 BPF_FIB_LKUP_RET_FRAG_NEEDED
, /* fragmentation required to fwd */
3566 struct bpf_fib_lookup
{
3567 /* input: network family for lookup (AF_INET, AF_INET6)
3568 * output: network family of egress nexthop
3572 /* set if lookup is to consider L4 data - e.g., FIB rules */
3577 /* total length of packet from network header - used for MTU check */
3580 /* input: L3 device index for lookup
3581 * output: device index from FIB lookup
3586 /* inputs to lookup */
3587 __u8 tos
; /* AF_INET */
3588 __be32 flowinfo
; /* AF_INET6, flow_label + priority */
3590 /* output: metric of fib result (IPv4/IPv6 only) */
3596 __u32 ipv6_src
[4]; /* in6_addr; network order */
3599 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3600 * network header. output: bpf_fib_lookup sets to gateway address
3601 * if FIB lookup returns gateway route
3605 __u32 ipv6_dst
[4]; /* in6_addr; network order */
3609 __be16 h_vlan_proto
;
3611 __u8 smac
[6]; /* ETH_ALEN */
3612 __u8 dmac
[6]; /* ETH_ALEN */
3615 enum bpf_task_fd_type
{
3616 BPF_FD_TYPE_RAW_TRACEPOINT
, /* tp name */
3617 BPF_FD_TYPE_TRACEPOINT
, /* tp name */
3618 BPF_FD_TYPE_KPROBE
, /* (symbol + offset) or addr */
3619 BPF_FD_TYPE_KRETPROBE
, /* (symbol + offset) or addr */
3620 BPF_FD_TYPE_UPROBE
, /* filename + offset */
3621 BPF_FD_TYPE_URETPROBE
, /* filename + offset */
3624 #define BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG (1U << 0)
3625 #define BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL (1U << 1)
3626 #define BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP (1U << 2)
3628 struct bpf_flow_keys
{
3631 __u16 addr_proto
; /* ETH_P_* of valid addrs */
3645 __u32 ipv6_src
[4]; /* in6_addr; network order */
3646 __u32 ipv6_dst
[4]; /* in6_addr; network order */
3653 struct bpf_func_info
{
3658 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3659 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3661 struct bpf_line_info
{
3663 __u32 file_name_off
;
3668 struct bpf_spin_lock
{
3673 __u32 write
; /* Sysctl is being read (= 0) or written (= 1).
3674 * Allows 1,2,4-byte read, but no write.
3676 __u32 file_pos
; /* Sysctl file position to read from, write to.
3677 * Allows 1,2,4-byte read an 4-byte write.
3681 struct bpf_sockopt
{
3682 __bpf_md_ptr(struct bpf_sock
*, sk
);
3683 __bpf_md_ptr(void *, optval
);
3684 __bpf_md_ptr(void *, optval_end
);
3692 #endif /* __LINUX_BPF_H__ */