1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_XADD 0xc0 /* exclusive add */
25 #define BPF_MOV 0xb0 /* mov reg to reg */
26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28 /* change endianness of a register */
29 #define BPF_END 0xd0 /* flags for endianness conversion: */
30 #define BPF_TO_LE 0x00 /* convert to little-endian */
31 #define BPF_TO_BE 0x08 /* convert to big-endian */
32 #define BPF_FROM_LE BPF_TO_LE
33 #define BPF_FROM_BE BPF_TO_BE
36 #define BPF_JNE 0x50 /* jump != */
37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
43 #define BPF_CALL 0x80 /* function call */
44 #define BPF_EXIT 0x90 /* function return */
46 /* Register numbers */
62 /* BPF has 10 general purpose 64-bit registers and stack frame. */
63 #define MAX_BPF_REG __MAX_BPF_REG
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74 struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data[]; /* 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,
110 BPF_MAP_LOOKUP_BATCH,
111 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
112 BPF_MAP_UPDATE_BATCH,
113 BPF_MAP_DELETE_BATCH,
120 BPF_MAP_TYPE_PROG_ARRAY,
121 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
122 BPF_MAP_TYPE_PERCPU_HASH,
123 BPF_MAP_TYPE_PERCPU_ARRAY,
124 BPF_MAP_TYPE_STACK_TRACE,
125 BPF_MAP_TYPE_CGROUP_ARRAY,
126 BPF_MAP_TYPE_LRU_HASH,
127 BPF_MAP_TYPE_LRU_PERCPU_HASH,
128 BPF_MAP_TYPE_LPM_TRIE,
129 BPF_MAP_TYPE_ARRAY_OF_MAPS,
130 BPF_MAP_TYPE_HASH_OF_MAPS,
132 BPF_MAP_TYPE_SOCKMAP,
135 BPF_MAP_TYPE_SOCKHASH,
136 BPF_MAP_TYPE_CGROUP_STORAGE,
137 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
138 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
141 BPF_MAP_TYPE_SK_STORAGE,
142 BPF_MAP_TYPE_DEVMAP_HASH,
143 BPF_MAP_TYPE_STRUCT_OPS,
146 /* Note that tracing related programs such as
147 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
148 * are not subject to a stable API since kernel internal data
149 * structures can change from release to release and may
150 * therefore break existing tracing BPF programs. Tracing BPF
151 * programs correspond to /a/ specific kernel which is to be
152 * analyzed, and not /a/ specific kernel /and/ all future ones.
155 BPF_PROG_TYPE_UNSPEC,
156 BPF_PROG_TYPE_SOCKET_FILTER,
157 BPF_PROG_TYPE_KPROBE,
158 BPF_PROG_TYPE_SCHED_CLS,
159 BPF_PROG_TYPE_SCHED_ACT,
160 BPF_PROG_TYPE_TRACEPOINT,
162 BPF_PROG_TYPE_PERF_EVENT,
163 BPF_PROG_TYPE_CGROUP_SKB,
164 BPF_PROG_TYPE_CGROUP_SOCK,
165 BPF_PROG_TYPE_LWT_IN,
166 BPF_PROG_TYPE_LWT_OUT,
167 BPF_PROG_TYPE_LWT_XMIT,
168 BPF_PROG_TYPE_SOCK_OPS,
169 BPF_PROG_TYPE_SK_SKB,
170 BPF_PROG_TYPE_CGROUP_DEVICE,
171 BPF_PROG_TYPE_SK_MSG,
172 BPF_PROG_TYPE_RAW_TRACEPOINT,
173 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
174 BPF_PROG_TYPE_LWT_SEG6LOCAL,
175 BPF_PROG_TYPE_LIRC_MODE2,
176 BPF_PROG_TYPE_SK_REUSEPORT,
177 BPF_PROG_TYPE_FLOW_DISSECTOR,
178 BPF_PROG_TYPE_CGROUP_SYSCTL,
179 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
180 BPF_PROG_TYPE_CGROUP_SOCKOPT,
181 BPF_PROG_TYPE_TRACING,
182 BPF_PROG_TYPE_STRUCT_OPS,
186 enum bpf_attach_type {
187 BPF_CGROUP_INET_INGRESS,
188 BPF_CGROUP_INET_EGRESS,
189 BPF_CGROUP_INET_SOCK_CREATE,
191 BPF_SK_SKB_STREAM_PARSER,
192 BPF_SK_SKB_STREAM_VERDICT,
195 BPF_CGROUP_INET4_BIND,
196 BPF_CGROUP_INET6_BIND,
197 BPF_CGROUP_INET4_CONNECT,
198 BPF_CGROUP_INET6_CONNECT,
199 BPF_CGROUP_INET4_POST_BIND,
200 BPF_CGROUP_INET6_POST_BIND,
201 BPF_CGROUP_UDP4_SENDMSG,
202 BPF_CGROUP_UDP6_SENDMSG,
206 BPF_CGROUP_UDP4_RECVMSG,
207 BPF_CGROUP_UDP6_RECVMSG,
208 BPF_CGROUP_GETSOCKOPT,
209 BPF_CGROUP_SETSOCKOPT,
213 __MAX_BPF_ATTACH_TYPE
216 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
218 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
220 * NONE(default): No further bpf programs allowed in the subtree.
222 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
223 * the program in this cgroup yields to sub-cgroup program.
225 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
226 * that cgroup program gets run in addition to the program in this cgroup.
228 * Only one program is allowed to be attached to a cgroup with
229 * NONE or BPF_F_ALLOW_OVERRIDE flag.
230 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
231 * release old program and attach the new one. Attach flags has to match.
233 * Multiple programs are allowed to be attached to a cgroup with
234 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
235 * (those that were attached first, run first)
236 * The programs of sub-cgroup are executed first, then programs of
237 * this cgroup and then programs of parent cgroup.
238 * When children program makes decision (like picking TCP CA or sock bind)
239 * parent program has a chance to override it.
241 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
242 * programs for a cgroup. Though it's possible to replace an old program at
243 * any position by also specifying BPF_F_REPLACE flag and position itself in
244 * replace_bpf_fd attribute. Old program at this position will be released.
246 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
247 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
249 * cgrp1 (MULTI progs A, B) ->
250 * cgrp2 (OVERRIDE prog C) ->
251 * cgrp3 (MULTI prog D) ->
252 * cgrp4 (OVERRIDE prog E) ->
253 * cgrp5 (NONE prog F)
254 * the event in cgrp5 triggers execution of F,D,A,B in that order.
255 * if prog F is detached, the execution is E,D,A,B
256 * if prog F and D are detached, the execution is E,A,B
257 * if prog F, E and D are detached, the execution is C,A,B
259 * All eligible programs are executed regardless of return code from
262 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
263 #define BPF_F_ALLOW_MULTI (1U << 1)
264 #define BPF_F_REPLACE (1U << 2)
266 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
267 * verifier will perform strict alignment checking as if the kernel
268 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
269 * and NET_IP_ALIGN defined to 2.
271 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
273 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
274 * verifier will allow any alignment whatsoever. On platforms
275 * with strict alignment requirements for loads ands stores (such
276 * as sparc and mips) the verifier validates that all loads and
277 * stores provably follow this requirement. This flag turns that
278 * checking and enforcement off.
280 * It is mostly used for testing when we want to validate the
281 * context and memory access aspects of the verifier, but because
282 * of an unaligned access the alignment check would trigger before
283 * the one we are interested in.
285 #define BPF_F_ANY_ALIGNMENT (1U << 1)
287 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
288 * Verifier does sub-register def/use analysis and identifies instructions whose
289 * def only matters for low 32-bit, high 32-bit is never referenced later
290 * through implicit zero extension. Therefore verifier notifies JIT back-ends
291 * that it is safe to ignore clearing high 32-bit for these instructions. This
292 * saves some back-ends a lot of code-gen. However such optimization is not
293 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
294 * hence hasn't used verifier's analysis result. But, we really want to have a
295 * way to be able to verify the correctness of the described optimization on
296 * x86_64 on which testsuites are frequently exercised.
298 * So, this flag is introduced. Once it is set, verifier will randomize high
299 * 32-bit for those instructions who has been identified as safe to ignore them.
300 * Then, if verifier is not doing correct analysis, such randomization will
301 * regress tests to expose bugs.
303 #define BPF_F_TEST_RND_HI32 (1U << 2)
305 /* The verifier internal test flag. Behavior is undefined */
306 #define BPF_F_TEST_STATE_FREQ (1U << 3)
308 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
311 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
312 * insn[0].imm: map fd map fd
313 * insn[1].imm: 0 offset into value
316 * ldimm64 rewrite: address of map address of map[0]+offset
317 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
319 #define BPF_PSEUDO_MAP_FD 1
320 #define BPF_PSEUDO_MAP_VALUE 2
322 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
323 * offset to another bpf function
325 #define BPF_PSEUDO_CALL 1
327 /* flags for BPF_MAP_UPDATE_ELEM command */
329 BPF_ANY = 0, /* create new element or update existing */
330 BPF_NOEXIST = 1, /* create new element if it didn't exist */
331 BPF_EXIST = 2, /* update existing element */
332 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
335 /* flags for BPF_MAP_CREATE command */
337 BPF_F_NO_PREALLOC = (1U << 0),
338 /* Instead of having one common LRU list in the
339 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
340 * which can scale and perform better.
341 * Note, the LRU nodes (including free nodes) cannot be moved
342 * across different LRU lists.
344 BPF_F_NO_COMMON_LRU = (1U << 1),
345 /* Specify numa node during map creation */
346 BPF_F_NUMA_NODE = (1U << 2),
348 /* Flags for accessing BPF object from syscall side. */
349 BPF_F_RDONLY = (1U << 3),
350 BPF_F_WRONLY = (1U << 4),
352 /* Flag for stack_map, store build_id+offset instead of pointer */
353 BPF_F_STACK_BUILD_ID = (1U << 5),
355 /* Zero-initialize hash function seed. This should only be used for testing. */
356 BPF_F_ZERO_SEED = (1U << 6),
358 /* Flags for accessing BPF object from program side. */
359 BPF_F_RDONLY_PROG = (1U << 7),
360 BPF_F_WRONLY_PROG = (1U << 8),
362 /* Clone map from listener for newly accepted socket */
363 BPF_F_CLONE = (1U << 9),
365 /* Enable memory-mapping BPF map */
366 BPF_F_MMAPABLE = (1U << 10),
369 /* Flags for BPF_PROG_QUERY. */
371 /* Query effective (directly attached + inherited from ancestor cgroups)
372 * programs that will be executed for events within a cgroup.
373 * attach_flags with this flag are returned only for directly attached programs.
375 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
377 enum bpf_stack_build_id_status {
378 /* user space need an empty entry to identify end of a trace */
379 BPF_STACK_BUILD_ID_EMPTY = 0,
380 /* with valid build_id and offset */
381 BPF_STACK_BUILD_ID_VALID = 1,
382 /* couldn't get build_id, fallback to ip */
383 BPF_STACK_BUILD_ID_IP = 2,
386 #define BPF_BUILD_ID_SIZE 20
387 struct bpf_stack_build_id {
389 unsigned char build_id[BPF_BUILD_ID_SIZE];
396 #define BPF_OBJ_NAME_LEN 16U
399 struct { /* anonymous struct used by BPF_MAP_CREATE command */
400 __u32 map_type; /* one of enum bpf_map_type */
401 __u32 key_size; /* size of key in bytes */
402 __u32 value_size; /* size of value in bytes */
403 __u32 max_entries; /* max number of entries in a map */
404 __u32 map_flags; /* BPF_MAP_CREATE related
405 * flags defined above.
407 __u32 inner_map_fd; /* fd pointing to the inner map */
408 __u32 numa_node; /* numa node (effective only if
409 * BPF_F_NUMA_NODE is set).
411 char map_name[BPF_OBJ_NAME_LEN];
412 __u32 map_ifindex; /* ifindex of netdev to create on */
413 __u32 btf_fd; /* fd pointing to a BTF type data */
414 __u32 btf_key_type_id; /* BTF type_id of the key */
415 __u32 btf_value_type_id; /* BTF type_id of the value */
416 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
417 * struct stored as the
422 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
427 __aligned_u64 next_key;
432 struct { /* struct used by BPF_MAP_*_BATCH commands */
433 __aligned_u64 in_batch; /* start batch,
434 * NULL to start from beginning
436 __aligned_u64 out_batch; /* output: next start batch */
438 __aligned_u64 values;
439 __u32 count; /* input/output:
440 * input: # of key/value
442 * output: # of filled elements
449 struct { /* anonymous struct used by BPF_PROG_LOAD command */
450 __u32 prog_type; /* one of enum bpf_prog_type */
453 __aligned_u64 license;
454 __u32 log_level; /* verbosity level of verifier */
455 __u32 log_size; /* size of user buffer */
456 __aligned_u64 log_buf; /* user supplied buffer */
457 __u32 kern_version; /* not used */
459 char prog_name[BPF_OBJ_NAME_LEN];
460 __u32 prog_ifindex; /* ifindex of netdev to prep for */
461 /* For some prog types expected attach type must be known at
462 * load time to verify attach type specific parts of prog
463 * (context accesses, allowed helpers, etc).
465 __u32 expected_attach_type;
466 __u32 prog_btf_fd; /* fd pointing to BTF type data */
467 __u32 func_info_rec_size; /* userspace bpf_func_info size */
468 __aligned_u64 func_info; /* func info */
469 __u32 func_info_cnt; /* number of bpf_func_info records */
470 __u32 line_info_rec_size; /* userspace bpf_line_info size */
471 __aligned_u64 line_info; /* line info */
472 __u32 line_info_cnt; /* number of bpf_line_info records */
473 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
474 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
477 struct { /* anonymous struct used by BPF_OBJ_* commands */
478 __aligned_u64 pathname;
483 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
484 __u32 target_fd; /* container object to attach to */
485 __u32 attach_bpf_fd; /* eBPF program to attach */
488 __u32 replace_bpf_fd; /* previously attached eBPF
489 * program to replace if
490 * BPF_F_REPLACE is used
494 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
497 __u32 data_size_in; /* input: len of data_in */
498 __u32 data_size_out; /* input/output: len of data_out
499 * returns ENOSPC if data_out
502 __aligned_u64 data_in;
503 __aligned_u64 data_out;
506 __u32 ctx_size_in; /* input: len of ctx_in */
507 __u32 ctx_size_out; /* input/output: len of ctx_out
508 * returns ENOSPC if ctx_out
511 __aligned_u64 ctx_in;
512 __aligned_u64 ctx_out;
515 struct { /* anonymous struct used by BPF_*_GET_*_ID */
526 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
532 struct { /* anonymous struct used by BPF_PROG_QUERY command */
533 __u32 target_fd; /* container object to query */
537 __aligned_u64 prog_ids;
546 struct { /* anonymous struct for BPF_BTF_LOAD */
548 __aligned_u64 btf_log_buf;
555 __u32 pid; /* input: pid */
556 __u32 fd; /* input: fd */
557 __u32 flags; /* input: flags */
558 __u32 buf_len; /* input/output: buf len */
559 __aligned_u64 buf; /* input/output:
560 * tp_name for tracepoint
562 * filename for uprobe
564 __u32 prog_id; /* output: prod_id */
565 __u32 fd_type; /* output: BPF_FD_TYPE_* */
566 __u64 probe_offset; /* output: probe_offset */
567 __u64 probe_addr; /* output: probe_addr */
569 } __attribute__((aligned(8)));
571 /* The description below is an attempt at providing documentation to eBPF
572 * developers about the multiple available eBPF helper functions. It can be
573 * parsed and used to produce a manual page. The workflow is the following,
574 * and requires the rst2man utility:
576 * $ ./scripts/bpf_helpers_doc.py \
577 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
578 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
579 * $ man /tmp/bpf-helpers.7
581 * Note that in order to produce this external documentation, some RST
582 * formatting is used in the descriptions to get "bold" and "italics" in
583 * manual pages. Also note that the few trailing white spaces are
584 * intentional, removing them would break paragraphs for rst2man.
586 * Start of BPF helper function descriptions:
588 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
590 * Perform a lookup in *map* for an entry associated to *key*.
592 * Map value associated to *key*, or **NULL** if no entry was
595 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
597 * Add or update the value of the entry associated to *key* in
598 * *map* with *value*. *flags* is one of:
601 * The entry for *key* must not exist in the map.
603 * The entry for *key* must already exist in the map.
605 * No condition on the existence of the entry for *key*.
607 * Flag value **BPF_NOEXIST** cannot be used for maps of types
608 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
609 * elements always exist), the helper would return an error.
611 * 0 on success, or a negative error in case of failure.
613 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
615 * Delete entry with *key* from *map*.
617 * 0 on success, or a negative error in case of failure.
619 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
621 * For tracing programs, safely attempt to read *size* bytes from
622 * kernel space address *unsafe_ptr* and store the data in *dst*.
624 * Generally, use bpf_probe_read_user() or bpf_probe_read_kernel()
627 * 0 on success, or a negative error in case of failure.
629 * u64 bpf_ktime_get_ns(void)
631 * Return the time elapsed since system boot, in nanoseconds.
635 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
637 * This helper is a "printk()-like" facility for debugging. It
638 * prints a message defined by format *fmt* (of size *fmt_size*)
639 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
640 * available. It can take up to three additional **u64**
641 * arguments (as an eBPF helpers, the total number of arguments is
644 * Each time the helper is called, it appends a line to the trace.
645 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
646 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
647 * The format of the trace is customizable, and the exact output
648 * one will get depends on the options set in
649 * *\/sys/kernel/debug/tracing/trace_options* (see also the
650 * *README* file under the same directory). However, it usually
651 * defaults to something like:
655 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
659 * * ``telnet`` is the name of the current task.
660 * * ``470`` is the PID of the current task.
661 * * ``001`` is the CPU number on which the task is
663 * * In ``.N..``, each character refers to a set of
664 * options (whether irqs are enabled, scheduling
665 * options, whether hard/softirqs are running, level of
666 * preempt_disabled respectively). **N** means that
667 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
669 * * ``419421.045894`` is a timestamp.
670 * * ``0x00000001`` is a fake value used by BPF for the
671 * instruction pointer register.
672 * * ``<formatted msg>`` is the message formatted with
675 * The conversion specifiers supported by *fmt* are similar, but
676 * more limited than for printk(). They are **%d**, **%i**,
677 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
678 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
679 * of field, padding with zeroes, etc.) is available, and the
680 * helper will return **-EINVAL** (but print nothing) if it
681 * encounters an unknown specifier.
683 * Also, note that **bpf_trace_printk**\ () is slow, and should
684 * only be used for debugging purposes. For this reason, a notice
685 * bloc (spanning several lines) is printed to kernel logs and
686 * states that the helper should not be used "for production use"
687 * the first time this helper is used (or more precisely, when
688 * **trace_printk**\ () buffers are allocated). For passing values
689 * to user space, perf events should be preferred.
691 * The number of bytes written to the buffer, or a negative error
692 * in case of failure.
694 * u32 bpf_get_prandom_u32(void)
696 * Get a pseudo-random number.
698 * From a security point of view, this helper uses its own
699 * pseudo-random internal state, and cannot be used to infer the
700 * seed of other random functions in the kernel. However, it is
701 * essential to note that the generator used by the helper is not
702 * cryptographically secure.
704 * A random 32-bit unsigned value.
706 * u32 bpf_get_smp_processor_id(void)
708 * Get the SMP (symmetric multiprocessing) processor id. Note that
709 * all programs run with preemption disabled, which means that the
710 * SMP processor id is stable during all the execution of the
713 * The SMP id of the processor running the program.
715 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
717 * Store *len* bytes from address *from* into the packet
718 * associated to *skb*, at *offset*. *flags* are a combination of
719 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
720 * checksum for the packet after storing the bytes) and
721 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
722 * **->swhash** and *skb*\ **->l4hash** to 0).
724 * A call to this helper is susceptible to change the underlying
725 * packet buffer. Therefore, at load time, all checks on pointers
726 * previously done by the verifier are invalidated and must be
727 * performed again, if the helper is used in combination with
728 * direct packet access.
730 * 0 on success, or a negative error in case of failure.
732 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
734 * Recompute the layer 3 (e.g. IP) checksum for the packet
735 * associated to *skb*. Computation is incremental, so the helper
736 * must know the former value of the header field that was
737 * modified (*from*), the new value of this field (*to*), and the
738 * number of bytes (2 or 4) for this field, stored in *size*.
739 * Alternatively, it is possible to store the difference between
740 * the previous and the new values of the header field in *to*, by
741 * setting *from* and *size* to 0. For both methods, *offset*
742 * indicates the location of the IP checksum within the packet.
744 * This helper works in combination with **bpf_csum_diff**\ (),
745 * which does not update the checksum in-place, but offers more
746 * flexibility and can handle sizes larger than 2 or 4 for the
747 * checksum to update.
749 * A call to this helper is susceptible to change the underlying
750 * packet buffer. Therefore, at load time, all checks on pointers
751 * previously done by the verifier are invalidated and must be
752 * performed again, if the helper is used in combination with
753 * direct packet access.
755 * 0 on success, or a negative error in case of failure.
757 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
759 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
760 * packet associated to *skb*. Computation is incremental, so the
761 * helper must know the former value of the header field that was
762 * modified (*from*), the new value of this field (*to*), and the
763 * number of bytes (2 or 4) for this field, stored on the lowest
764 * four bits of *flags*. Alternatively, it is possible to store
765 * the difference between the previous and the new values of the
766 * header field in *to*, by setting *from* and the four lowest
767 * bits of *flags* to 0. For both methods, *offset* indicates the
768 * location of the IP checksum within the packet. In addition to
769 * the size of the field, *flags* can be added (bitwise OR) actual
770 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
771 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
772 * for updates resulting in a null checksum the value is set to
773 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
774 * the checksum is to be computed against a pseudo-header.
776 * This helper works in combination with **bpf_csum_diff**\ (),
777 * which does not update the checksum in-place, but offers more
778 * flexibility and can handle sizes larger than 2 or 4 for the
779 * checksum to update.
781 * A call to this helper is susceptible to change the underlying
782 * packet buffer. Therefore, at load time, all checks on pointers
783 * previously done by the verifier are invalidated and must be
784 * performed again, if the helper is used in combination with
785 * direct packet access.
787 * 0 on success, or a negative error in case of failure.
789 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
791 * This special helper is used to trigger a "tail call", or in
792 * other words, to jump into another eBPF program. The same stack
793 * frame is used (but values on stack and in registers for the
794 * caller are not accessible to the callee). This mechanism allows
795 * for program chaining, either for raising the maximum number of
796 * available eBPF instructions, or to execute given programs in
797 * conditional blocks. For security reasons, there is an upper
798 * limit to the number of successive tail calls that can be
801 * Upon call of this helper, the program attempts to jump into a
802 * program referenced at index *index* in *prog_array_map*, a
803 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
804 * *ctx*, a pointer to the context.
806 * If the call succeeds, the kernel immediately runs the first
807 * instruction of the new program. This is not a function call,
808 * and it never returns to the previous program. If the call
809 * fails, then the helper has no effect, and the caller continues
810 * to run its subsequent instructions. A call can fail if the
811 * destination program for the jump does not exist (i.e. *index*
812 * is superior to the number of entries in *prog_array_map*), or
813 * if the maximum number of tail calls has been reached for this
814 * chain of programs. This limit is defined in the kernel by the
815 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
816 * which is currently set to 32.
818 * 0 on success, or a negative error in case of failure.
820 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
822 * Clone and redirect the packet associated to *skb* to another
823 * net device of index *ifindex*. Both ingress and egress
824 * interfaces can be used for redirection. The **BPF_F_INGRESS**
825 * value in *flags* is used to make the distinction (ingress path
826 * is selected if the flag is present, egress path otherwise).
827 * This is the only flag supported for now.
829 * In comparison with **bpf_redirect**\ () helper,
830 * **bpf_clone_redirect**\ () has the associated cost of
831 * duplicating the packet buffer, but this can be executed out of
832 * the eBPF program. Conversely, **bpf_redirect**\ () is more
833 * efficient, but it is handled through an action code where the
834 * redirection happens only after the eBPF program has returned.
836 * A call to this helper is susceptible to change the underlying
837 * packet buffer. Therefore, at load time, all checks on pointers
838 * previously done by the verifier are invalidated and must be
839 * performed again, if the helper is used in combination with
840 * direct packet access.
842 * 0 on success, or a negative error in case of failure.
844 * u64 bpf_get_current_pid_tgid(void)
846 * A 64-bit integer containing the current tgid and pid, and
848 * *current_task*\ **->tgid << 32 \|**
849 * *current_task*\ **->pid**.
851 * u64 bpf_get_current_uid_gid(void)
853 * A 64-bit integer containing the current GID and UID, and
854 * created as such: *current_gid* **<< 32 \|** *current_uid*.
856 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
858 * Copy the **comm** attribute of the current task into *buf* of
859 * *size_of_buf*. The **comm** attribute contains the name of
860 * the executable (excluding the path) for the current task. The
861 * *size_of_buf* must be strictly positive. On success, the
862 * helper makes sure that the *buf* is NUL-terminated. On failure,
863 * it is filled with zeroes.
865 * 0 on success, or a negative error in case of failure.
867 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
869 * Retrieve the classid for the current task, i.e. for the net_cls
870 * cgroup to which *skb* belongs.
872 * This helper can be used on TC egress path, but not on ingress.
874 * The net_cls cgroup provides an interface to tag network packets
875 * based on a user-provided identifier for all traffic coming from
876 * the tasks belonging to the related cgroup. See also the related
877 * kernel documentation, available from the Linux sources in file
878 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
880 * The Linux kernel has two versions for cgroups: there are
881 * cgroups v1 and cgroups v2. Both are available to users, who can
882 * use a mixture of them, but note that the net_cls cgroup is for
883 * cgroup v1 only. This makes it incompatible with BPF programs
884 * run on cgroups, which is a cgroup-v2-only feature (a socket can
885 * only hold data for one version of cgroups at a time).
887 * This helper is only available is the kernel was compiled with
888 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
889 * "**y**" or to "**m**".
891 * The classid, or 0 for the default unconfigured classid.
893 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
895 * Push a *vlan_tci* (VLAN tag control information) of protocol
896 * *vlan_proto* to the packet associated to *skb*, then update
897 * the checksum. Note that if *vlan_proto* is different from
898 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
899 * be **ETH_P_8021Q**.
901 * A call to this helper is susceptible to change the underlying
902 * packet buffer. Therefore, at load time, all checks on pointers
903 * previously done by the verifier are invalidated and must be
904 * performed again, if the helper is used in combination with
905 * direct packet access.
907 * 0 on success, or a negative error in case of failure.
909 * int bpf_skb_vlan_pop(struct sk_buff *skb)
911 * Pop a VLAN header from the packet associated to *skb*.
913 * A call to this helper is susceptible to change the underlying
914 * packet buffer. Therefore, at load time, all checks on pointers
915 * previously done by the verifier are invalidated and must be
916 * performed again, if the helper is used in combination with
917 * direct packet access.
919 * 0 on success, or a negative error in case of failure.
921 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
923 * Get tunnel metadata. This helper takes a pointer *key* to an
924 * empty **struct bpf_tunnel_key** of **size**, that will be
925 * filled with tunnel metadata for the packet associated to *skb*.
926 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
927 * indicates that the tunnel is based on IPv6 protocol instead of
930 * The **struct bpf_tunnel_key** is an object that generalizes the
931 * principal parameters used by various tunneling protocols into a
932 * single struct. This way, it can be used to easily make a
933 * decision based on the contents of the encapsulation header,
934 * "summarized" in this struct. In particular, it holds the IP
935 * address of the remote end (IPv4 or IPv6, depending on the case)
936 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
937 * this struct exposes the *key*\ **->tunnel_id**, which is
938 * generally mapped to a VNI (Virtual Network Identifier), making
939 * it programmable together with the **bpf_skb_set_tunnel_key**\
942 * Let's imagine that the following code is part of a program
943 * attached to the TC ingress interface, on one end of a GRE
944 * tunnel, and is supposed to filter out all messages coming from
945 * remote ends with IPv4 address other than 10.0.0.1:
950 * struct bpf_tunnel_key key = {};
952 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
954 * return TC_ACT_SHOT; // drop packet
956 * if (key.remote_ipv4 != 0x0a000001)
957 * return TC_ACT_SHOT; // drop packet
959 * return TC_ACT_OK; // accept packet
961 * This interface can also be used with all encapsulation devices
962 * that can operate in "collect metadata" mode: instead of having
963 * one network device per specific configuration, the "collect
964 * metadata" mode only requires a single device where the
965 * configuration can be extracted from this helper.
967 * This can be used together with various tunnels such as VXLan,
968 * Geneve, GRE or IP in IP (IPIP).
970 * 0 on success, or a negative error in case of failure.
972 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
974 * Populate tunnel metadata for packet associated to *skb.* The
975 * tunnel metadata is set to the contents of *key*, of *size*. The
976 * *flags* can be set to a combination of the following values:
978 * **BPF_F_TUNINFO_IPV6**
979 * Indicate that the tunnel is based on IPv6 protocol
981 * **BPF_F_ZERO_CSUM_TX**
982 * For IPv4 packets, add a flag to tunnel metadata
983 * indicating that checksum computation should be skipped
984 * and checksum set to zeroes.
985 * **BPF_F_DONT_FRAGMENT**
986 * Add a flag to tunnel metadata indicating that the
987 * packet should not be fragmented.
988 * **BPF_F_SEQ_NUMBER**
989 * Add a flag to tunnel metadata indicating that a
990 * sequence number should be added to tunnel header before
991 * sending the packet. This flag was added for GRE
992 * encapsulation, but might be used with other protocols
993 * as well in the future.
995 * Here is a typical usage on the transmit path:
999 * struct bpf_tunnel_key key;
1001 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1002 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1004 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1005 * helper for additional information.
1007 * 0 on success, or a negative error in case of failure.
1009 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1011 * Read the value of a perf event counter. This helper relies on a
1012 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1013 * the perf event counter is selected when *map* is updated with
1014 * perf event file descriptors. The *map* is an array whose size
1015 * is the number of available CPUs, and each cell contains a value
1016 * relative to one CPU. The value to retrieve is indicated by
1017 * *flags*, that contains the index of the CPU to look up, masked
1018 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1019 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1020 * current CPU should be retrieved.
1022 * Note that before Linux 4.13, only hardware perf event can be
1025 * Also, be aware that the newer helper
1026 * **bpf_perf_event_read_value**\ () is recommended over
1027 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1028 * quirks where error and counter value are used as a return code
1029 * (which is wrong to do since ranges may overlap). This issue is
1030 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1031 * time provides more features over the **bpf_perf_event_read**\
1032 * () interface. Please refer to the description of
1033 * **bpf_perf_event_read_value**\ () for details.
1035 * The value of the perf event counter read from the map, or a
1036 * negative error code in case of failure.
1038 * int bpf_redirect(u32 ifindex, u64 flags)
1040 * Redirect the packet to another net device of index *ifindex*.
1041 * This helper is somewhat similar to **bpf_clone_redirect**\
1042 * (), except that the packet is not cloned, which provides
1043 * increased performance.
1045 * Except for XDP, both ingress and egress interfaces can be used
1046 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1047 * to make the distinction (ingress path is selected if the flag
1048 * is present, egress path otherwise). Currently, XDP only
1049 * supports redirection to the egress interface, and accepts no
1052 * The same effect can also be attained with the more generic
1053 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1054 * redirect target instead of providing it directly to the helper.
1056 * For XDP, the helper returns **XDP_REDIRECT** on success or
1057 * **XDP_ABORTED** on error. For other program types, the values
1058 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1061 * u32 bpf_get_route_realm(struct sk_buff *skb)
1063 * Retrieve the realm or the route, that is to say the
1064 * **tclassid** field of the destination for the *skb*. The
1065 * indentifier retrieved is a user-provided tag, similar to the
1066 * one used with the net_cls cgroup (see description for
1067 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1068 * held by a route (a destination entry), not by a task.
1070 * Retrieving this identifier works with the clsact TC egress hook
1071 * (see also **tc-bpf(8)**), or alternatively on conventional
1072 * classful egress qdiscs, but not on TC ingress path. In case of
1073 * clsact TC egress hook, this has the advantage that, internally,
1074 * the destination entry has not been dropped yet in the transmit
1075 * path. Therefore, the destination entry does not need to be
1076 * artificially held via **netif_keep_dst**\ () for a classful
1077 * qdisc until the *skb* is freed.
1079 * This helper is available only if the kernel was compiled with
1080 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1082 * The realm of the route for the packet associated to *skb*, or 0
1083 * if none was found.
1085 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1087 * Write raw *data* blob into a special BPF perf event held by
1088 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1089 * event must have the following attributes: **PERF_SAMPLE_RAW**
1090 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1091 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1093 * The *flags* are used to indicate the index in *map* for which
1094 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1095 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1096 * to indicate that the index of the current CPU core should be
1099 * The value to write, of *size*, is passed through eBPF stack and
1100 * pointed by *data*.
1102 * The context of the program *ctx* needs also be passed to the
1105 * On user space, a program willing to read the values needs to
1106 * call **perf_event_open**\ () on the perf event (either for
1107 * one or for all CPUs) and to store the file descriptor into the
1108 * *map*. This must be done before the eBPF program can send data
1109 * into it. An example is available in file
1110 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1111 * tree (the eBPF program counterpart is in
1112 * *samples/bpf/trace_output_kern.c*).
1114 * **bpf_perf_event_output**\ () achieves better performance
1115 * than **bpf_trace_printk**\ () for sharing data with user
1116 * space, and is much better suitable for streaming data from eBPF
1119 * Note that this helper is not restricted to tracing use cases
1120 * and can be used with programs attached to TC or XDP as well,
1121 * where it allows for passing data to user space listeners. Data
1124 * * Only custom structs,
1125 * * Only the packet payload, or
1126 * * A combination of both.
1128 * 0 on success, or a negative error in case of failure.
1130 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1132 * This helper was provided as an easy way to load data from a
1133 * packet. It can be used to load *len* bytes from *offset* from
1134 * the packet associated to *skb*, into the buffer pointed by
1137 * Since Linux 4.7, usage of this helper has mostly been replaced
1138 * by "direct packet access", enabling packet data to be
1139 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1140 * pointing respectively to the first byte of packet data and to
1141 * the byte after the last byte of packet data. However, it
1142 * remains useful if one wishes to read large quantities of data
1143 * at once from a packet into the eBPF stack.
1145 * 0 on success, or a negative error in case of failure.
1147 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1149 * Walk a user or a kernel stack and return its id. To achieve
1150 * this, the helper needs *ctx*, which is a pointer to the context
1151 * on which the tracing program is executed, and a pointer to a
1152 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1154 * The last argument, *flags*, holds the number of stack frames to
1155 * skip (from 0 to 255), masked with
1156 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1157 * a combination of the following flags:
1159 * **BPF_F_USER_STACK**
1160 * Collect a user space stack instead of a kernel stack.
1161 * **BPF_F_FAST_STACK_CMP**
1162 * Compare stacks by hash only.
1163 * **BPF_F_REUSE_STACKID**
1164 * If two different stacks hash into the same *stackid*,
1165 * discard the old one.
1167 * The stack id retrieved is a 32 bit long integer handle which
1168 * can be further combined with other data (including other stack
1169 * ids) and used as a key into maps. This can be useful for
1170 * generating a variety of graphs (such as flame graphs or off-cpu
1173 * For walking a stack, this helper is an improvement over
1174 * **bpf_probe_read**\ (), which can be used with unrolled loops
1175 * but is not efficient and consumes a lot of eBPF instructions.
1176 * Instead, **bpf_get_stackid**\ () can collect up to
1177 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1178 * this limit can be controlled with the **sysctl** program, and
1179 * that it should be manually increased in order to profile long
1180 * user stacks (such as stacks for Java programs). To do so, use:
1184 * # sysctl kernel.perf_event_max_stack=<new value>
1186 * The positive or null stack id on success, or a negative error
1187 * in case of failure.
1189 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1191 * Compute a checksum difference, from the raw buffer pointed by
1192 * *from*, of length *from_size* (that must be a multiple of 4),
1193 * towards the raw buffer pointed by *to*, of size *to_size*
1194 * (same remark). An optional *seed* can be added to the value
1195 * (this can be cascaded, the seed may come from a previous call
1198 * This is flexible enough to be used in several ways:
1200 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1201 * checksum, it can be used when pushing new data.
1202 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1203 * checksum, it can be used when removing data from a packet.
1204 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1205 * can be used to compute a diff. Note that *from_size* and
1206 * *to_size* do not need to be equal.
1208 * This helper can be used in combination with
1209 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1210 * which one can feed in the difference computed with
1211 * **bpf_csum_diff**\ ().
1213 * The checksum result, or a negative error code in case of
1216 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1218 * Retrieve tunnel options metadata for the packet associated to
1219 * *skb*, and store the raw tunnel option data to the buffer *opt*
1222 * This helper can be used with encapsulation devices that can
1223 * operate in "collect metadata" mode (please refer to the related
1224 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1225 * more details). A particular example where this can be used is
1226 * in combination with the Geneve encapsulation protocol, where it
1227 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1228 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1229 * the eBPF program. This allows for full customization of these
1232 * The size of the option data retrieved.
1234 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1236 * Set tunnel options metadata for the packet associated to *skb*
1237 * to the option data contained in the raw buffer *opt* of *size*.
1239 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1240 * helper for additional information.
1242 * 0 on success, or a negative error in case of failure.
1244 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1246 * Change the protocol of the *skb* to *proto*. Currently
1247 * supported are transition from IPv4 to IPv6, and from IPv6 to
1248 * IPv4. The helper takes care of the groundwork for the
1249 * transition, including resizing the socket buffer. The eBPF
1250 * program is expected to fill the new headers, if any, via
1251 * **skb_store_bytes**\ () and to recompute the checksums with
1252 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1253 * (). The main case for this helper is to perform NAT64
1254 * operations out of an eBPF program.
1256 * Internally, the GSO type is marked as dodgy so that headers are
1257 * checked and segments are recalculated by the GSO/GRO engine.
1258 * The size for GSO target is adapted as well.
1260 * All values for *flags* are reserved for future usage, and must
1263 * A call to this helper is susceptible to change the underlying
1264 * packet buffer. Therefore, at load time, all checks on pointers
1265 * previously done by the verifier are invalidated and must be
1266 * performed again, if the helper is used in combination with
1267 * direct packet access.
1269 * 0 on success, or a negative error in case of failure.
1271 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1273 * Change the packet type for the packet associated to *skb*. This
1274 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1275 * the eBPF program does not have a write access to *skb*\
1276 * **->pkt_type** beside this helper. Using a helper here allows
1277 * for graceful handling of errors.
1279 * The major use case is to change incoming *skb*s to
1280 * **PACKET_HOST** in a programmatic way instead of having to
1281 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1284 * Note that *type* only allows certain values. At this time, they
1289 * **PACKET_BROADCAST**
1290 * Send packet to all.
1291 * **PACKET_MULTICAST**
1292 * Send packet to group.
1293 * **PACKET_OTHERHOST**
1294 * Send packet to someone else.
1296 * 0 on success, or a negative error in case of failure.
1298 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1300 * Check whether *skb* is a descendant of the cgroup2 held by
1301 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1303 * The return value depends on the result of the test, and can be:
1305 * * 0, if the *skb* failed the cgroup2 descendant test.
1306 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1307 * * A negative error code, if an error occurred.
1309 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1311 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1312 * not set, in particular if the hash was cleared due to mangling,
1313 * recompute this hash. Later accesses to the hash can be done
1314 * directly with *skb*\ **->hash**.
1316 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1317 * prototype with **bpf_skb_change_proto**\ (), or calling
1318 * **bpf_skb_store_bytes**\ () with the
1319 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1320 * the hash and to trigger a new computation for the next call to
1321 * **bpf_get_hash_recalc**\ ().
1325 * u64 bpf_get_current_task(void)
1327 * A pointer to the current task struct.
1329 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1331 * Attempt in a safe way to write *len* bytes from the buffer
1332 * *src* to *dst* in memory. It only works for threads that are in
1333 * user context, and *dst* must be a valid user space address.
1335 * This helper should not be used to implement any kind of
1336 * security mechanism because of TOC-TOU attacks, but rather to
1337 * debug, divert, and manipulate execution of semi-cooperative
1340 * Keep in mind that this feature is meant for experiments, and it
1341 * has a risk of crashing the system and running programs.
1342 * Therefore, when an eBPF program using this helper is attached,
1343 * a warning including PID and process name is printed to kernel
1346 * 0 on success, or a negative error in case of failure.
1348 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1350 * Check whether the probe is being run is the context of a given
1351 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1352 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1354 * The return value depends on the result of the test, and can be:
1356 * * 0, if the *skb* task belongs to the cgroup2.
1357 * * 1, if the *skb* task does not belong to the cgroup2.
1358 * * A negative error code, if an error occurred.
1360 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1362 * Resize (trim or grow) the packet associated to *skb* to the
1363 * new *len*. The *flags* are reserved for future usage, and must
1366 * The basic idea is that the helper performs the needed work to
1367 * change the size of the packet, then the eBPF program rewrites
1368 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1369 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1370 * and others. This helper is a slow path utility intended for
1371 * replies with control messages. And because it is targeted for
1372 * slow path, the helper itself can afford to be slow: it
1373 * implicitly linearizes, unclones and drops offloads from the
1376 * A call to this helper is susceptible to change the underlying
1377 * packet buffer. Therefore, at load time, all checks on pointers
1378 * previously done by the verifier are invalidated and must be
1379 * performed again, if the helper is used in combination with
1380 * direct packet access.
1382 * 0 on success, or a negative error in case of failure.
1384 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1386 * Pull in non-linear data in case the *skb* is non-linear and not
1387 * all of *len* are part of the linear section. Make *len* bytes
1388 * from *skb* readable and writable. If a zero value is passed for
1389 * *len*, then the whole length of the *skb* is pulled.
1391 * This helper is only needed for reading and writing with direct
1394 * For direct packet access, testing that offsets to access
1395 * are within packet boundaries (test on *skb*\ **->data_end**) is
1396 * susceptible to fail if offsets are invalid, or if the requested
1397 * data is in non-linear parts of the *skb*. On failure the
1398 * program can just bail out, or in the case of a non-linear
1399 * buffer, use a helper to make the data available. The
1400 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1401 * the data. Another one consists in using **bpf_skb_pull_data**
1402 * to pull in once the non-linear parts, then retesting and
1403 * eventually access the data.
1405 * At the same time, this also makes sure the *skb* is uncloned,
1406 * which is a necessary condition for direct write. As this needs
1407 * to be an invariant for the write part only, the verifier
1408 * detects writes and adds a prologue that is calling
1409 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1410 * the very beginning in case it is indeed cloned.
1412 * A call to this helper is susceptible to change the underlying
1413 * packet buffer. Therefore, at load time, all checks on pointers
1414 * previously done by the verifier are invalidated and must be
1415 * performed again, if the helper is used in combination with
1416 * direct packet access.
1418 * 0 on success, or a negative error in case of failure.
1420 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1422 * Add the checksum *csum* into *skb*\ **->csum** in case the
1423 * driver has supplied a checksum for the entire packet into that
1424 * field. Return an error otherwise. This helper is intended to be
1425 * used in combination with **bpf_csum_diff**\ (), in particular
1426 * when the checksum needs to be updated after data has been
1427 * written into the packet through direct packet access.
1429 * The checksum on success, or a negative error code in case of
1432 * void bpf_set_hash_invalid(struct sk_buff *skb)
1434 * Invalidate the current *skb*\ **->hash**. It can be used after
1435 * mangling on headers through direct packet access, in order to
1436 * indicate that the hash is outdated and to trigger a
1437 * recalculation the next time the kernel tries to access this
1438 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1440 * int bpf_get_numa_node_id(void)
1442 * Return the id of the current NUMA node. The primary use case
1443 * for this helper is the selection of sockets for the local NUMA
1444 * node, when the program is attached to sockets using the
1445 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1446 * but the helper is also available to other eBPF program types,
1447 * similarly to **bpf_get_smp_processor_id**\ ().
1449 * The id of current NUMA node.
1451 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1453 * Grows headroom of packet associated to *skb* and adjusts the
1454 * offset of the MAC header accordingly, adding *len* bytes of
1455 * space. It automatically extends and reallocates memory as
1458 * This helper can be used on a layer 3 *skb* to push a MAC header
1459 * for redirection into a layer 2 device.
1461 * All values for *flags* are reserved for future usage, and must
1464 * A call to this helper is susceptible to change the underlying
1465 * packet buffer. Therefore, at load time, all checks on pointers
1466 * previously done by the verifier are invalidated and must be
1467 * performed again, if the helper is used in combination with
1468 * direct packet access.
1470 * 0 on success, or a negative error in case of failure.
1472 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1474 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1475 * it is possible to use a negative value for *delta*. This helper
1476 * can be used to prepare the packet for pushing or popping
1479 * A call to this helper is susceptible to change the underlying
1480 * packet buffer. Therefore, at load time, all checks on pointers
1481 * previously done by the verifier are invalidated and must be
1482 * performed again, if the helper is used in combination with
1483 * direct packet access.
1485 * 0 on success, or a negative error in case of failure.
1487 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1489 * Copy a NUL terminated string from an unsafe kernel address
1490 * *unsafe_ptr* to *dst*. See bpf_probe_read_kernel_str() for
1493 * Generally, use bpf_probe_read_user_str() or bpf_probe_read_kernel_str()
1496 * On success, the strictly positive length of the string,
1497 * including the trailing NUL character. On error, a negative
1500 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1502 * If the **struct sk_buff** pointed by *skb* has a known socket,
1503 * retrieve the cookie (generated by the kernel) of this socket.
1504 * If no cookie has been set yet, generate a new cookie. Once
1505 * generated, the socket cookie remains stable for the life of the
1506 * socket. This helper can be useful for monitoring per socket
1507 * networking traffic statistics as it provides a global socket
1508 * identifier that can be assumed unique.
1510 * A 8-byte long non-decreasing number on success, or 0 if the
1511 * socket field is missing inside *skb*.
1513 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1515 * Equivalent to bpf_get_socket_cookie() helper that accepts
1516 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1518 * A 8-byte long non-decreasing number.
1520 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1522 * Equivalent to bpf_get_socket_cookie() helper that accepts
1523 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1525 * A 8-byte long non-decreasing number.
1527 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1529 * The owner UID of the socket associated to *skb*. If the socket
1530 * is **NULL**, or if it is not a full socket (i.e. if it is a
1531 * time-wait or a request socket instead), **overflowuid** value
1532 * is returned (note that **overflowuid** might also be the actual
1533 * UID value for the socket).
1535 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1537 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1542 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1544 * Emulate a call to **setsockopt()** on the socket associated to
1545 * *bpf_socket*, which must be a full socket. The *level* at
1546 * which the option resides and the name *optname* of the option
1547 * must be specified, see **setsockopt(2)** for more information.
1548 * The option value of length *optlen* is pointed by *optval*.
1550 * This helper actually implements a subset of **setsockopt()**.
1551 * It supports the following *level*\ s:
1553 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1554 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1555 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1556 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1557 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1558 * **TCP_BPF_SNDCWND_CLAMP**.
1559 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1560 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1562 * 0 on success, or a negative error in case of failure.
1564 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1566 * Grow or shrink the room for data in the packet associated to
1567 * *skb* by *len_diff*, and according to the selected *mode*.
1569 * There are two supported modes at this time:
1571 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1572 * (room space is added or removed below the layer 2 header).
1574 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1575 * (room space is added or removed below the layer 3 header).
1577 * The following flags are supported at this time:
1579 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1580 * Adjusting mss in this way is not allowed for datagrams.
1582 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1583 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1584 * Any new space is reserved to hold a tunnel header.
1585 * Configure skb offsets and other fields accordingly.
1587 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1588 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1589 * Use with ENCAP_L3 flags to further specify the tunnel type.
1591 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1592 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1593 * type; *len* is the length of the inner MAC header.
1595 * A call to this helper is susceptible to change the underlying
1596 * packet buffer. Therefore, at load time, all checks on pointers
1597 * previously done by the verifier are invalidated and must be
1598 * performed again, if the helper is used in combination with
1599 * direct packet access.
1601 * 0 on success, or a negative error in case of failure.
1603 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1605 * Redirect the packet to the endpoint referenced by *map* at
1606 * index *key*. Depending on its type, this *map* can contain
1607 * references to net devices (for forwarding packets through other
1608 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1609 * but this is only implemented for native XDP (with driver
1610 * support) as of this writing).
1612 * The lower two bits of *flags* are used as the return code if
1613 * the map lookup fails. This is so that the return value can be
1614 * one of the XDP program return codes up to XDP_TX, as chosen by
1615 * the caller. Any higher bits in the *flags* argument must be
1618 * See also bpf_redirect(), which only supports redirecting to an
1619 * ifindex, but doesn't require a map to do so.
1621 * **XDP_REDIRECT** on success, or the value of the two lower bits
1622 * of the **flags* argument on error.
1624 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1626 * Redirect the packet to the socket referenced by *map* (of type
1627 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1628 * egress interfaces can be used for redirection. The
1629 * **BPF_F_INGRESS** value in *flags* is used to make the
1630 * distinction (ingress path is selected if the flag is present,
1631 * egress path otherwise). This is the only flag supported for now.
1633 * **SK_PASS** on success, or **SK_DROP** on error.
1635 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1637 * Add an entry to, or update a *map* referencing sockets. The
1638 * *skops* is used as a new value for the entry associated to
1639 * *key*. *flags* is one of:
1642 * The entry for *key* must not exist in the map.
1644 * The entry for *key* must already exist in the map.
1646 * No condition on the existence of the entry for *key*.
1648 * If the *map* has eBPF programs (parser and verdict), those will
1649 * be inherited by the socket being added. If the socket is
1650 * already attached to eBPF programs, this results in an error.
1652 * 0 on success, or a negative error in case of failure.
1654 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1656 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1657 * *delta* (which can be positive or negative). Note that this
1658 * operation modifies the address stored in *xdp_md*\ **->data**,
1659 * so the latter must be loaded only after the helper has been
1662 * The use of *xdp_md*\ **->data_meta** is optional and programs
1663 * are not required to use it. The rationale is that when the
1664 * packet is processed with XDP (e.g. as DoS filter), it is
1665 * possible to push further meta data along with it before passing
1666 * to the stack, and to give the guarantee that an ingress eBPF
1667 * program attached as a TC classifier on the same device can pick
1668 * this up for further post-processing. Since TC works with socket
1669 * buffers, it remains possible to set from XDP the **mark** or
1670 * **priority** pointers, or other pointers for the socket buffer.
1671 * Having this scratch space generic and programmable allows for
1672 * more flexibility as the user is free to store whatever meta
1675 * A call to this helper is susceptible to change the underlying
1676 * packet buffer. Therefore, at load time, all checks on pointers
1677 * previously done by the verifier are invalidated and must be
1678 * performed again, if the helper is used in combination with
1679 * direct packet access.
1681 * 0 on success, or a negative error in case of failure.
1683 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1685 * Read the value of a perf event counter, and store it into *buf*
1686 * of size *buf_size*. This helper relies on a *map* of type
1687 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1688 * counter is selected when *map* is updated with perf event file
1689 * descriptors. The *map* is an array whose size is the number of
1690 * available CPUs, and each cell contains a value relative to one
1691 * CPU. The value to retrieve is indicated by *flags*, that
1692 * contains the index of the CPU to look up, masked with
1693 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1694 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1695 * current CPU should be retrieved.
1697 * This helper behaves in a way close to
1698 * **bpf_perf_event_read**\ () helper, save that instead of
1699 * just returning the value observed, it fills the *buf*
1700 * structure. This allows for additional data to be retrieved: in
1701 * particular, the enabled and running times (in *buf*\
1702 * **->enabled** and *buf*\ **->running**, respectively) are
1703 * copied. In general, **bpf_perf_event_read_value**\ () is
1704 * recommended over **bpf_perf_event_read**\ (), which has some
1705 * ABI issues and provides fewer functionalities.
1707 * These values are interesting, because hardware PMU (Performance
1708 * Monitoring Unit) counters are limited resources. When there are
1709 * more PMU based perf events opened than available counters,
1710 * kernel will multiplex these events so each event gets certain
1711 * percentage (but not all) of the PMU time. In case that
1712 * multiplexing happens, the number of samples or counter value
1713 * will not reflect the case compared to when no multiplexing
1714 * occurs. This makes comparison between different runs difficult.
1715 * Typically, the counter value should be normalized before
1716 * comparing to other experiments. The usual normalization is done
1721 * normalized_counter = counter * t_enabled / t_running
1723 * Where t_enabled is the time enabled for event and t_running is
1724 * the time running for event since last normalization. The
1725 * enabled and running times are accumulated since the perf event
1726 * open. To achieve scaling factor between two invocations of an
1727 * eBPF program, users can can use CPU id as the key (which is
1728 * typical for perf array usage model) to remember the previous
1729 * value and do the calculation inside the eBPF program.
1731 * 0 on success, or a negative error in case of failure.
1733 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1735 * For en eBPF program attached to a perf event, retrieve the
1736 * value of the event counter associated to *ctx* and store it in
1737 * the structure pointed by *buf* and of size *buf_size*. Enabled
1738 * and running times are also stored in the structure (see
1739 * description of helper **bpf_perf_event_read_value**\ () for
1742 * 0 on success, or a negative error in case of failure.
1744 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1746 * Emulate a call to **getsockopt()** on the socket associated to
1747 * *bpf_socket*, which must be a full socket. The *level* at
1748 * which the option resides and the name *optname* of the option
1749 * must be specified, see **getsockopt(2)** for more information.
1750 * The retrieved value is stored in the structure pointed by
1751 * *opval* and of length *optlen*.
1753 * This helper actually implements a subset of **getsockopt()**.
1754 * It supports the following *level*\ s:
1756 * * **IPPROTO_TCP**, which supports *optname*
1757 * **TCP_CONGESTION**.
1758 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1759 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1761 * 0 on success, or a negative error in case of failure.
1763 * int bpf_override_return(struct pt_regs *regs, u64 rc)
1765 * Used for error injection, this helper uses kprobes to override
1766 * the return value of the probed function, and to set it to *rc*.
1767 * The first argument is the context *regs* on which the kprobe
1770 * This helper works by setting setting the PC (program counter)
1771 * to an override function which is run in place of the original
1772 * probed function. This means the probed function is not run at
1773 * all. The replacement function just returns with the required
1776 * This helper has security implications, and thus is subject to
1777 * restrictions. It is only available if the kernel was compiled
1778 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1779 * option, and in this case it only works on functions tagged with
1780 * **ALLOW_ERROR_INJECTION** in the kernel code.
1782 * Also, the helper is only available for the architectures having
1783 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1784 * x86 architecture is the only one to support this feature.
1788 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1790 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1791 * for the full TCP socket associated to *bpf_sock_ops* to
1794 * The primary use of this field is to determine if there should
1795 * be calls to eBPF programs of type
1796 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1797 * code. A program of the same type can change its value, per
1798 * connection and as necessary, when the connection is
1799 * established. This field is directly accessible for reading, but
1800 * this helper must be used for updates in order to return an
1801 * error if an eBPF program tries to set a callback that is not
1802 * supported in the current kernel.
1804 * *argval* is a flag array which can combine these flags:
1806 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1807 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1808 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1809 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1811 * Therefore, this function can be used to clear a callback flag by
1812 * setting the appropriate bit to zero. e.g. to disable the RTO
1815 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1816 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1818 * Here are some examples of where one could call such eBPF
1822 * * When a packet is retransmitted.
1823 * * When the connection terminates.
1824 * * When a packet is sent.
1825 * * When a packet is received.
1827 * Code **-EINVAL** if the socket is not a full TCP socket;
1828 * otherwise, a positive number containing the bits that could not
1829 * be set is returned (which comes down to 0 if all bits were set
1832 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1834 * This helper is used in programs implementing policies at the
1835 * socket level. If the message *msg* is allowed to pass (i.e. if
1836 * the verdict eBPF program returns **SK_PASS**), redirect it to
1837 * the socket referenced by *map* (of type
1838 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1839 * egress interfaces can be used for redirection. The
1840 * **BPF_F_INGRESS** value in *flags* is used to make the
1841 * distinction (ingress path is selected if the flag is present,
1842 * egress path otherwise). This is the only flag supported for now.
1844 * **SK_PASS** on success, or **SK_DROP** on error.
1846 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1848 * For socket policies, apply the verdict of the eBPF program to
1849 * the next *bytes* (number of bytes) of message *msg*.
1851 * For example, this helper can be used in the following cases:
1853 * * A single **sendmsg**\ () or **sendfile**\ () system call
1854 * contains multiple logical messages that the eBPF program is
1855 * supposed to read and for which it should apply a verdict.
1856 * * An eBPF program only cares to read the first *bytes* of a
1857 * *msg*. If the message has a large payload, then setting up
1858 * and calling the eBPF program repeatedly for all bytes, even
1859 * though the verdict is already known, would create unnecessary
1862 * When called from within an eBPF program, the helper sets a
1863 * counter internal to the BPF infrastructure, that is used to
1864 * apply the last verdict to the next *bytes*. If *bytes* is
1865 * smaller than the current data being processed from a
1866 * **sendmsg**\ () or **sendfile**\ () system call, the first
1867 * *bytes* will be sent and the eBPF program will be re-run with
1868 * the pointer for start of data pointing to byte number *bytes*
1869 * **+ 1**. If *bytes* is larger than the current data being
1870 * processed, then the eBPF verdict will be applied to multiple
1871 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1874 * Note that if a socket closes with the internal counter holding
1875 * a non-zero value, this is not a problem because data is not
1876 * being buffered for *bytes* and is sent as it is received.
1880 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1882 * For socket policies, prevent the execution of the verdict eBPF
1883 * program for message *msg* until *bytes* (byte number) have been
1886 * This can be used when one needs a specific number of bytes
1887 * before a verdict can be assigned, even if the data spans
1888 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1889 * case would be a user calling **sendmsg**\ () repeatedly with
1890 * 1-byte long message segments. Obviously, this is bad for
1891 * performance, but it is still valid. If the eBPF program needs
1892 * *bytes* bytes to validate a header, this helper can be used to
1893 * prevent the eBPF program to be called again until *bytes* have
1898 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1900 * For socket policies, pull in non-linear data from user space
1901 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1902 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1905 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1906 * *msg* it can only parse data that the (**data**, **data_end**)
1907 * pointers have already consumed. For **sendmsg**\ () hooks this
1908 * is likely the first scatterlist element. But for calls relying
1909 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1910 * be the range (**0**, **0**) because the data is shared with
1911 * user space and by default the objective is to avoid allowing
1912 * user space to modify data while (or after) eBPF verdict is
1913 * being decided. This helper can be used to pull in data and to
1914 * set the start and end pointer to given values. Data will be
1915 * copied if necessary (i.e. if data was not linear and if start
1916 * and end pointers do not point to the same chunk).
1918 * A call to this helper is susceptible to change the underlying
1919 * packet buffer. Therefore, at load time, all checks on pointers
1920 * previously done by the verifier are invalidated and must be
1921 * performed again, if the helper is used in combination with
1922 * direct packet access.
1924 * All values for *flags* are reserved for future usage, and must
1927 * 0 on success, or a negative error in case of failure.
1929 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1931 * Bind the socket associated to *ctx* to the address pointed by
1932 * *addr*, of length *addr_len*. This allows for making outgoing
1933 * connection from the desired IP address, which can be useful for
1934 * example when all processes inside a cgroup should use one
1935 * single IP address on a host that has multiple IP configured.
1937 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1938 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1939 * **AF_INET6**). Looking for a free port to bind to can be
1940 * expensive, therefore binding to port is not permitted by the
1941 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1942 * must be set to zero.
1944 * 0 on success, or a negative error in case of failure.
1946 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1948 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1949 * only possible to shrink the packet as of this writing,
1950 * therefore *delta* must be a negative integer.
1952 * A call to this helper is susceptible to change the underlying
1953 * packet buffer. Therefore, at load time, all checks on pointers
1954 * previously done by the verifier are invalidated and must be
1955 * performed again, if the helper is used in combination with
1956 * direct packet access.
1958 * 0 on success, or a negative error in case of failure.
1960 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1962 * Retrieve the XFRM state (IP transform framework, see also
1963 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1965 * The retrieved value is stored in the **struct bpf_xfrm_state**
1966 * pointed by *xfrm_state* and of length *size*.
1968 * All values for *flags* are reserved for future usage, and must
1971 * This helper is available only if the kernel was compiled with
1972 * **CONFIG_XFRM** configuration option.
1974 * 0 on success, or a negative error in case of failure.
1976 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
1978 * Return a user or a kernel stack in bpf program provided buffer.
1979 * To achieve this, the helper needs *ctx*, which is a pointer
1980 * to the context on which the tracing program is executed.
1981 * To store the stacktrace, the bpf program provides *buf* with
1982 * a nonnegative *size*.
1984 * The last argument, *flags*, holds the number of stack frames to
1985 * skip (from 0 to 255), masked with
1986 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1987 * the following flags:
1989 * **BPF_F_USER_STACK**
1990 * Collect a user space stack instead of a kernel stack.
1991 * **BPF_F_USER_BUILD_ID**
1992 * Collect buildid+offset instead of ips for user stack,
1993 * only valid if **BPF_F_USER_STACK** is also specified.
1995 * **bpf_get_stack**\ () can collect up to
1996 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1997 * to sufficient large buffer size. Note that
1998 * this limit can be controlled with the **sysctl** program, and
1999 * that it should be manually increased in order to profile long
2000 * user stacks (such as stacks for Java programs). To do so, use:
2004 * # sysctl kernel.perf_event_max_stack=<new value>
2006 * A non-negative value equal to or less than *size* on success,
2007 * or a negative error in case of failure.
2009 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2011 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2012 * it provides an easy way to load *len* bytes from *offset*
2013 * from the packet associated to *skb*, into the buffer pointed
2014 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2015 * a fifth argument *start_header* exists in order to select a
2016 * base offset to start from. *start_header* can be one of:
2018 * **BPF_HDR_START_MAC**
2019 * Base offset to load data from is *skb*'s mac header.
2020 * **BPF_HDR_START_NET**
2021 * Base offset to load data from is *skb*'s network header.
2023 * In general, "direct packet access" is the preferred method to
2024 * access packet data, however, this helper is in particular useful
2025 * in socket filters where *skb*\ **->data** does not always point
2026 * to the start of the mac header and where "direct packet access"
2029 * 0 on success, or a negative error in case of failure.
2031 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2033 * Do FIB lookup in kernel tables using parameters in *params*.
2034 * If lookup is successful and result shows packet is to be
2035 * forwarded, the neighbor tables are searched for the nexthop.
2036 * If successful (ie., FIB lookup shows forwarding and nexthop
2037 * is resolved), the nexthop address is returned in ipv4_dst
2038 * or ipv6_dst based on family, smac is set to mac address of
2039 * egress device, dmac is set to nexthop mac address, rt_metric
2040 * is set to metric from route (IPv4/IPv6 only), and ifindex
2041 * is set to the device index of the nexthop from the FIB lookup.
2043 * *plen* argument is the size of the passed in struct.
2044 * *flags* argument can be a combination of one or more of the
2047 * **BPF_FIB_LOOKUP_DIRECT**
2048 * Do a direct table lookup vs full lookup using FIB
2050 * **BPF_FIB_LOOKUP_OUTPUT**
2051 * Perform lookup from an egress perspective (default is
2054 * *ctx* is either **struct xdp_md** for XDP programs or
2055 * **struct sk_buff** tc cls_act programs.
2057 * * < 0 if any input argument is invalid
2058 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2059 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2060 * packet is not forwarded or needs assist from full stack
2062 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2064 * Add an entry to, or update a sockhash *map* referencing sockets.
2065 * The *skops* is used as a new value for the entry associated to
2066 * *key*. *flags* is one of:
2069 * The entry for *key* must not exist in the map.
2071 * The entry for *key* must already exist in the map.
2073 * No condition on the existence of the entry for *key*.
2075 * If the *map* has eBPF programs (parser and verdict), those will
2076 * be inherited by the socket being added. If the socket is
2077 * already attached to eBPF programs, this results in an error.
2079 * 0 on success, or a negative error in case of failure.
2081 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2083 * This helper is used in programs implementing policies at the
2084 * socket level. If the message *msg* is allowed to pass (i.e. if
2085 * the verdict eBPF program returns **SK_PASS**), redirect it to
2086 * the socket referenced by *map* (of type
2087 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2088 * egress interfaces can be used for redirection. The
2089 * **BPF_F_INGRESS** value in *flags* is used to make the
2090 * distinction (ingress path is selected if the flag is present,
2091 * egress path otherwise). This is the only flag supported for now.
2093 * **SK_PASS** on success, or **SK_DROP** on error.
2095 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2097 * This helper is used in programs implementing policies at the
2098 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2099 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2100 * to the socket referenced by *map* (of type
2101 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2102 * egress interfaces can be used for redirection. The
2103 * **BPF_F_INGRESS** value in *flags* is used to make the
2104 * distinction (ingress path is selected if the flag is present,
2105 * egress otherwise). This is the only flag supported for now.
2107 * **SK_PASS** on success, or **SK_DROP** on error.
2109 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2111 * Encapsulate the packet associated to *skb* within a Layer 3
2112 * protocol header. This header is provided in the buffer at
2113 * address *hdr*, with *len* its size in bytes. *type* indicates
2114 * the protocol of the header and can be one of:
2116 * **BPF_LWT_ENCAP_SEG6**
2117 * IPv6 encapsulation with Segment Routing Header
2118 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2119 * the IPv6 header is computed by the kernel.
2120 * **BPF_LWT_ENCAP_SEG6_INLINE**
2121 * Only works if *skb* contains an IPv6 packet. Insert a
2122 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2124 * **BPF_LWT_ENCAP_IP**
2125 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2126 * must be IPv4 or IPv6, followed by zero or more
2127 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2128 * total bytes in all prepended headers. Please note that
2129 * if **skb_is_gso**\ (*skb*) is true, no more than two
2130 * headers can be prepended, and the inner header, if
2131 * present, should be either GRE or UDP/GUE.
2133 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2134 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2135 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2136 * **BPF_PROG_TYPE_LWT_XMIT**.
2138 * A call to this helper is susceptible to change the underlying
2139 * packet buffer. Therefore, at load time, all checks on pointers
2140 * previously done by the verifier are invalidated and must be
2141 * performed again, if the helper is used in combination with
2142 * direct packet access.
2144 * 0 on success, or a negative error in case of failure.
2146 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2148 * Store *len* bytes from address *from* into the packet
2149 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2150 * inside the outermost IPv6 Segment Routing Header can be
2151 * modified through this helper.
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_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2163 * Adjust the size allocated to TLVs in the outermost IPv6
2164 * Segment Routing Header contained in the packet associated to
2165 * *skb*, at position *offset* by *delta* bytes. Only offsets
2166 * after the segments are accepted. *delta* can be as well
2167 * positive (growing) as negative (shrinking).
2169 * A call to this helper is susceptible to change the underlying
2170 * packet buffer. Therefore, at load time, all checks on pointers
2171 * previously done by the verifier are invalidated and must be
2172 * performed again, if the helper is used in combination with
2173 * direct packet access.
2175 * 0 on success, or a negative error in case of failure.
2177 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2179 * Apply an IPv6 Segment Routing action of type *action* to the
2180 * packet associated to *skb*. Each action takes a parameter
2181 * contained at address *param*, and of length *param_len* bytes.
2182 * *action* can be one of:
2184 * **SEG6_LOCAL_ACTION_END_X**
2185 * End.X action: Endpoint with Layer-3 cross-connect.
2186 * Type of *param*: **struct in6_addr**.
2187 * **SEG6_LOCAL_ACTION_END_T**
2188 * End.T action: Endpoint with specific IPv6 table lookup.
2189 * Type of *param*: **int**.
2190 * **SEG6_LOCAL_ACTION_END_B6**
2191 * End.B6 action: Endpoint bound to an SRv6 policy.
2192 * Type of *param*: **struct ipv6_sr_hdr**.
2193 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2194 * End.B6.Encap action: Endpoint bound to an SRv6
2195 * encapsulation policy.
2196 * Type of *param*: **struct ipv6_sr_hdr**.
2198 * A call to this helper is susceptible to change the underlying
2199 * packet buffer. Therefore, at load time, all checks on pointers
2200 * previously done by the verifier are invalidated and must be
2201 * performed again, if the helper is used in combination with
2202 * direct packet access.
2204 * 0 on success, or a negative error in case of failure.
2206 * int bpf_rc_repeat(void *ctx)
2208 * This helper is used in programs implementing IR decoding, to
2209 * report a successfully decoded repeat key message. This delays
2210 * the generation of a key up event for previously generated
2213 * Some IR protocols like NEC have a special IR message for
2214 * repeating last button, for when a button is held down.
2216 * The *ctx* should point to the lirc sample as passed into
2219 * This helper is only available is the kernel was compiled with
2220 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2225 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2227 * This helper is used in programs implementing IR decoding, to
2228 * report a successfully decoded key press with *scancode*,
2229 * *toggle* value in the given *protocol*. The scancode will be
2230 * translated to a keycode using the rc keymap, and reported as
2231 * an input key down event. After a period a key up event is
2232 * generated. This period can be extended by calling either
2233 * **bpf_rc_keydown**\ () again with the same values, or calling
2234 * **bpf_rc_repeat**\ ().
2236 * Some protocols include a toggle bit, in case the button was
2237 * released and pressed again between consecutive scancodes.
2239 * The *ctx* should point to the lirc sample as passed into
2242 * The *protocol* is the decoded protocol number (see
2243 * **enum rc_proto** for some predefined values).
2245 * This helper is only available is the kernel was compiled with
2246 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2251 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2253 * Return the cgroup v2 id of the socket associated with the *skb*.
2254 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2255 * helper for cgroup v1 by providing a tag resp. identifier that
2256 * can be matched on or used for map lookups e.g. to implement
2257 * policy. The cgroup v2 id of a given path in the hierarchy is
2258 * exposed in user space through the f_handle API in order to get
2259 * to the same 64-bit id.
2261 * This helper can be used on TC egress path, but not on ingress,
2262 * and is available only if the kernel was compiled with the
2263 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2265 * The id is returned or 0 in case the id could not be retrieved.
2267 * u64 bpf_get_current_cgroup_id(void)
2269 * A 64-bit integer containing the current cgroup id based
2270 * on the cgroup within which the current task is running.
2272 * void *bpf_get_local_storage(void *map, u64 flags)
2274 * Get the pointer to the local storage area.
2275 * The type and the size of the local storage is defined
2276 * by the *map* argument.
2277 * The *flags* meaning is specific for each map type,
2278 * and has to be 0 for cgroup local storage.
2280 * Depending on the BPF program type, a local storage area
2281 * can be shared between multiple instances of the BPF program,
2282 * running simultaneously.
2284 * A user should care about the synchronization by himself.
2285 * For example, by using the **BPF_STX_XADD** instruction to alter
2288 * A pointer to the local storage area.
2290 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2292 * Select a **SO_REUSEPORT** socket from a
2293 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2294 * It checks the selected socket is matching the incoming
2295 * request in the socket buffer.
2297 * 0 on success, or a negative error in case of failure.
2299 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2301 * Return id of cgroup v2 that is ancestor of cgroup associated
2302 * with the *skb* at the *ancestor_level*. The root cgroup is at
2303 * *ancestor_level* zero and each step down the hierarchy
2304 * increments the level. If *ancestor_level* == level of cgroup
2305 * associated with *skb*, then return value will be same as that
2306 * of **bpf_skb_cgroup_id**\ ().
2308 * The helper is useful to implement policies based on cgroups
2309 * that are upper in hierarchy than immediate cgroup associated
2312 * The format of returned id and helper limitations are same as in
2313 * **bpf_skb_cgroup_id**\ ().
2315 * The id is returned or 0 in case the id could not be retrieved.
2317 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2319 * Look for TCP socket matching *tuple*, optionally in a child
2320 * network namespace *netns*. The return value must be checked,
2321 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2323 * The *ctx* should point to the context of the program, such as
2324 * the skb or socket (depending on the hook in use). This is used
2325 * to determine the base network namespace for the lookup.
2327 * *tuple_size* must be one of:
2329 * **sizeof**\ (*tuple*\ **->ipv4**)
2330 * Look for an IPv4 socket.
2331 * **sizeof**\ (*tuple*\ **->ipv6**)
2332 * Look for an IPv6 socket.
2334 * If the *netns* is a negative signed 32-bit integer, then the
2335 * socket lookup table in the netns associated with the *ctx* will
2336 * will be used. For the TC hooks, this is the netns of the device
2337 * in the skb. For socket hooks, this is the netns of the socket.
2338 * If *netns* is any other signed 32-bit value greater than or
2339 * equal to zero then it specifies the ID of the netns relative to
2340 * the netns associated with the *ctx*. *netns* values beyond the
2341 * range of 32-bit integers are reserved for future use.
2343 * All values for *flags* are reserved for future usage, and must
2346 * This helper is available only if the kernel was compiled with
2347 * **CONFIG_NET** configuration option.
2349 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2350 * For sockets with reuseport option, the **struct bpf_sock**
2351 * result is from *reuse*\ **->socks**\ [] using the hash of the
2354 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2356 * Look for UDP socket matching *tuple*, optionally in a child
2357 * network namespace *netns*. The return value must be checked,
2358 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2360 * The *ctx* should point to the context of the program, such as
2361 * the skb or socket (depending on the hook in use). This is used
2362 * to determine the base network namespace for the lookup.
2364 * *tuple_size* must be one of:
2366 * **sizeof**\ (*tuple*\ **->ipv4**)
2367 * Look for an IPv4 socket.
2368 * **sizeof**\ (*tuple*\ **->ipv6**)
2369 * Look for an IPv6 socket.
2371 * If the *netns* is a negative signed 32-bit integer, then the
2372 * socket lookup table in the netns associated with the *ctx* will
2373 * will be used. For the TC hooks, this is the netns of the device
2374 * in the skb. For socket hooks, this is the netns of the socket.
2375 * If *netns* is any other signed 32-bit value greater than or
2376 * equal to zero then it specifies the ID of the netns relative to
2377 * the netns associated with the *ctx*. *netns* values beyond the
2378 * range of 32-bit integers are reserved for future use.
2380 * All values for *flags* are reserved for future usage, and must
2383 * This helper is available only if the kernel was compiled with
2384 * **CONFIG_NET** configuration option.
2386 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2387 * For sockets with reuseport option, the **struct bpf_sock**
2388 * result is from *reuse*\ **->socks**\ [] using the hash of the
2391 * int bpf_sk_release(struct bpf_sock *sock)
2393 * Release the reference held by *sock*. *sock* must be a
2394 * non-**NULL** pointer that was returned from
2395 * **bpf_sk_lookup_xxx**\ ().
2397 * 0 on success, or a negative error in case of failure.
2399 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2401 * Push an element *value* in *map*. *flags* is one of:
2404 * If the queue/stack is full, the oldest element is
2405 * removed to make room for this.
2407 * 0 on success, or a negative error in case of failure.
2409 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2411 * Pop an element from *map*.
2413 * 0 on success, or a negative error in case of failure.
2415 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2417 * Get an element from *map* without removing it.
2419 * 0 on success, or a negative error in case of failure.
2421 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2423 * For socket policies, insert *len* bytes into *msg* at offset
2426 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2427 * *msg* it may want to insert metadata or options into the *msg*.
2428 * This can later be read and used by any of the lower layer BPF
2431 * This helper may fail if under memory pressure (a malloc
2432 * fails) in these cases BPF programs will get an appropriate
2433 * error and BPF programs will need to handle them.
2435 * 0 on success, or a negative error in case of failure.
2437 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2439 * Will remove *len* bytes from a *msg* starting at byte *start*.
2440 * This may result in **ENOMEM** errors under certain situations if
2441 * an allocation and copy are required due to a full ring buffer.
2442 * However, the helper will try to avoid doing the allocation
2443 * if possible. Other errors can occur if input parameters are
2444 * invalid either due to *start* byte not being valid part of *msg*
2445 * payload and/or *pop* value being to large.
2447 * 0 on success, or a negative error in case of failure.
2449 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2451 * This helper is used in programs implementing IR decoding, to
2452 * report a successfully decoded pointer movement.
2454 * The *ctx* should point to the lirc sample as passed into
2457 * This helper is only available is the kernel was compiled with
2458 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2463 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2465 * Acquire a spinlock represented by the pointer *lock*, which is
2466 * stored as part of a value of a map. Taking the lock allows to
2467 * safely update the rest of the fields in that value. The
2468 * spinlock can (and must) later be released with a call to
2469 * **bpf_spin_unlock**\ (\ *lock*\ ).
2471 * Spinlocks in BPF programs come with a number of restrictions
2474 * * **bpf_spin_lock** objects are only allowed inside maps of
2475 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2476 * list could be extended in the future).
2477 * * BTF description of the map is mandatory.
2478 * * The BPF program can take ONE lock at a time, since taking two
2479 * or more could cause dead locks.
2480 * * Only one **struct bpf_spin_lock** is allowed per map element.
2481 * * When the lock is taken, calls (either BPF to BPF or helpers)
2483 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2484 * allowed inside a spinlock-ed region.
2485 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2486 * the lock, on all execution paths, before it returns.
2487 * * The BPF program can access **struct bpf_spin_lock** only via
2488 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2489 * helpers. Loading or storing data into the **struct
2490 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2491 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2492 * of the map value must be a struct and have **struct
2493 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2494 * Nested lock inside another struct is not allowed.
2495 * * The **struct bpf_spin_lock** *lock* field in a map value must
2496 * be aligned on a multiple of 4 bytes in that value.
2497 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2498 * the **bpf_spin_lock** field to user space.
2499 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2500 * a BPF program, do not update the **bpf_spin_lock** field.
2501 * * **bpf_spin_lock** cannot be on the stack or inside a
2502 * networking packet (it can only be inside of a map values).
2503 * * **bpf_spin_lock** is available to root only.
2504 * * Tracing programs and socket filter programs cannot use
2505 * **bpf_spin_lock**\ () due to insufficient preemption checks
2506 * (but this may change in the future).
2507 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2511 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2513 * Release the *lock* previously locked by a call to
2514 * **bpf_spin_lock**\ (\ *lock*\ ).
2518 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2520 * This helper gets a **struct bpf_sock** pointer such
2521 * that all the fields in this **bpf_sock** can be accessed.
2523 * A **struct bpf_sock** pointer on success, or **NULL** in
2526 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2528 * This helper gets a **struct bpf_tcp_sock** pointer from a
2529 * **struct bpf_sock** pointer.
2531 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2534 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
2536 * Set ECN (Explicit Congestion Notification) field of IP header
2537 * to **CE** (Congestion Encountered) if current value is **ECT**
2538 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2541 * 1 if the **CE** flag is set (either by the current helper call
2542 * or because it was already present), 0 if it is not set.
2544 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2546 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2547 * **bpf_sk_release**\ () is unnecessary and not allowed.
2549 * A **struct bpf_sock** pointer on success, or **NULL** in
2552 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2554 * Look for TCP socket matching *tuple*, optionally in a child
2555 * network namespace *netns*. The return value must be checked,
2556 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2558 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2559 * that it also returns timewait or request sockets. Use
2560 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2563 * This helper is available only if the kernel was compiled with
2564 * **CONFIG_NET** configuration option.
2566 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2567 * For sockets with reuseport option, the **struct bpf_sock**
2568 * result is from *reuse*\ **->socks**\ [] using the hash of the
2571 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2573 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2574 * the listening socket in *sk*.
2576 * *iph* points to the start of the IPv4 or IPv6 header, while
2577 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2578 * **sizeof**\ (**struct ip6hdr**).
2580 * *th* points to the start of the TCP header, while *th_len*
2581 * contains **sizeof**\ (**struct tcphdr**).
2584 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2587 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2589 * Get name of sysctl in /proc/sys/ and copy it into provided by
2590 * program buffer *buf* of size *buf_len*.
2592 * The buffer is always NUL terminated, unless it's zero-sized.
2594 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2595 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2596 * only (e.g. "tcp_mem").
2598 * Number of character copied (not including the trailing NUL).
2600 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2601 * truncated name in this case).
2603 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2605 * Get current value of sysctl as it is presented in /proc/sys
2606 * (incl. newline, etc), and copy it as a string into provided
2607 * by program buffer *buf* of size *buf_len*.
2609 * The whole value is copied, no matter what file position user
2610 * space issued e.g. sys_read at.
2612 * The buffer is always NUL terminated, unless it's zero-sized.
2614 * Number of character copied (not including the trailing NUL).
2616 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2617 * truncated name in this case).
2619 * **-EINVAL** if current value was unavailable, e.g. because
2620 * sysctl is uninitialized and read returns -EIO for it.
2622 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2624 * Get new value being written by user space to sysctl (before
2625 * the actual write happens) and copy it as a string into
2626 * provided by program buffer *buf* of size *buf_len*.
2628 * User space may write new value at file position > 0.
2630 * The buffer is always NUL terminated, unless it's zero-sized.
2632 * Number of character copied (not including the trailing NUL).
2634 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2635 * truncated name in this case).
2637 * **-EINVAL** if sysctl is being read.
2639 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2641 * Override new value being written by user space to sysctl with
2642 * value provided by program in buffer *buf* of size *buf_len*.
2644 * *buf* should contain a string in same form as provided by user
2645 * space on sysctl write.
2647 * User space may write new value at file position > 0. To override
2648 * the whole sysctl value file position should be set to zero.
2652 * **-E2BIG** if the *buf_len* is too big.
2654 * **-EINVAL** if sysctl is being read.
2656 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2658 * Convert the initial part of the string from buffer *buf* of
2659 * size *buf_len* to a long integer according to the given base
2660 * and save the result in *res*.
2662 * The string may begin with an arbitrary amount of white space
2663 * (as determined by **isspace**\ (3)) followed by a single
2664 * optional '**-**' sign.
2666 * Five least significant bits of *flags* encode base, other bits
2667 * are currently unused.
2669 * Base must be either 8, 10, 16 or 0 to detect it automatically
2670 * similar to user space **strtol**\ (3).
2672 * Number of characters consumed on success. Must be positive but
2673 * no more than *buf_len*.
2675 * **-EINVAL** if no valid digits were found or unsupported base
2678 * **-ERANGE** if resulting value was out of range.
2680 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2682 * Convert the initial part of the string from buffer *buf* of
2683 * size *buf_len* to an unsigned long integer according to the
2684 * given base and save the result in *res*.
2686 * The string may begin with an arbitrary amount of white space
2687 * (as determined by **isspace**\ (3)).
2689 * Five least significant bits of *flags* encode base, other bits
2690 * are currently unused.
2692 * Base must be either 8, 10, 16 or 0 to detect it automatically
2693 * similar to user space **strtoul**\ (3).
2695 * Number of characters consumed on success. Must be positive but
2696 * no more than *buf_len*.
2698 * **-EINVAL** if no valid digits were found or unsupported base
2701 * **-ERANGE** if resulting value was out of range.
2703 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2705 * Get a bpf-local-storage from a *sk*.
2707 * Logically, it could be thought of getting the value from
2708 * a *map* with *sk* as the **key**. From this
2709 * perspective, the usage is not much different from
2710 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2711 * helper enforces the key must be a full socket and the map must
2712 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2714 * Underneath, the value is stored locally at *sk* instead of
2715 * the *map*. The *map* is used as the bpf-local-storage
2716 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2717 * searched against all bpf-local-storages residing at *sk*.
2719 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2720 * used such that a new bpf-local-storage will be
2721 * created if one does not exist. *value* can be used
2722 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2723 * the initial value of a bpf-local-storage. If *value* is
2724 * **NULL**, the new bpf-local-storage will be zero initialized.
2726 * A bpf-local-storage pointer is returned on success.
2728 * **NULL** if not found or there was an error in adding
2729 * a new bpf-local-storage.
2731 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2733 * Delete a bpf-local-storage from a *sk*.
2737 * **-ENOENT** if the bpf-local-storage cannot be found.
2739 * int bpf_send_signal(u32 sig)
2741 * Send signal *sig* to the process of the current task.
2742 * The signal may be delivered to any of this process's threads.
2744 * 0 on success or successfully queued.
2746 * **-EBUSY** if work queue under nmi is full.
2748 * **-EINVAL** if *sig* is invalid.
2750 * **-EPERM** if no permission to send the *sig*.
2752 * **-EAGAIN** if bpf program can try again.
2754 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2756 * Try to issue a SYN cookie for the packet with corresponding
2757 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2759 * *iph* points to the start of the IPv4 or IPv6 header, while
2760 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2761 * **sizeof**\ (**struct ip6hdr**).
2763 * *th* points to the start of the TCP header, while *th_len*
2764 * contains the length of the TCP header.
2767 * On success, lower 32 bits hold the generated SYN cookie in
2768 * followed by 16 bits which hold the MSS value for that cookie,
2769 * and the top 16 bits are unused.
2771 * On failure, the returned value is one of the following:
2773 * **-EINVAL** SYN cookie cannot be issued due to error
2775 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2777 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2779 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2781 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2783 * Write raw *data* blob into a special BPF perf event held by
2784 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2785 * event must have the following attributes: **PERF_SAMPLE_RAW**
2786 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2787 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2789 * The *flags* are used to indicate the index in *map* for which
2790 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2791 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2792 * to indicate that the index of the current CPU core should be
2795 * The value to write, of *size*, is passed through eBPF stack and
2796 * pointed by *data*.
2798 * *ctx* is a pointer to in-kernel struct sk_buff.
2800 * This helper is similar to **bpf_perf_event_output**\ () but
2801 * restricted to raw_tracepoint bpf programs.
2803 * 0 on success, or a negative error in case of failure.
2805 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2807 * Safely attempt to read *size* bytes from user space address
2808 * *unsafe_ptr* and store the data in *dst*.
2810 * 0 on success, or a negative error in case of failure.
2812 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2814 * Safely attempt to read *size* bytes from kernel space address
2815 * *unsafe_ptr* and store the data in *dst*.
2817 * 0 on success, or a negative error in case of failure.
2819 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2821 * Copy a NUL terminated string from an unsafe user address
2822 * *unsafe_ptr* to *dst*. The *size* should include the
2823 * terminating NUL byte. In case the string length is smaller than
2824 * *size*, the target is not padded with further NUL bytes. If the
2825 * string length is larger than *size*, just *size*-1 bytes are
2826 * copied and the last byte is set to NUL.
2828 * On success, the length of the copied string is returned. This
2829 * makes this helper useful in tracing programs for reading
2830 * strings, and more importantly to get its length at runtime. See
2831 * the following snippet:
2835 * SEC("kprobe/sys_open")
2836 * void bpf_sys_open(struct pt_regs *ctx)
2838 * char buf[PATHLEN]; // PATHLEN is defined to 256
2839 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2842 * // Consume buf, for example push it to
2843 * // userspace via bpf_perf_event_output(); we
2844 * // can use res (the string length) as event
2845 * // size, after checking its boundaries.
2848 * In comparison, using **bpf_probe_read_user()** helper here
2849 * instead to read the string would require to estimate the length
2850 * at compile time, and would often result in copying more memory
2853 * Another useful use case is when parsing individual process
2854 * arguments or individual environment variables navigating
2855 * *current*\ **->mm->arg_start** and *current*\
2856 * **->mm->env_start**: using this helper and the return value,
2857 * one can quickly iterate at the right offset of the memory area.
2859 * On success, the strictly positive length of the string,
2860 * including the trailing NUL character. On error, a negative
2863 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2865 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
2866 * to *dst*. Same semantics as with bpf_probe_read_user_str() apply.
2868 * On success, the strictly positive length of the string, including
2869 * the trailing NUL character. On error, a negative value.
2871 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2873 * Send out a tcp-ack. *tp* is the in-kernel struct tcp_sock.
2874 * *rcv_nxt* is the ack_seq to be sent out.
2876 * 0 on success, or a negative error in case of failure.
2878 * int bpf_send_signal_thread(u32 sig)
2880 * Send signal *sig* to the thread corresponding to the current task.
2882 * 0 on success or successfully queued.
2884 * **-EBUSY** if work queue under nmi is full.
2886 * **-EINVAL** if *sig* is invalid.
2888 * **-EPERM** if no permission to send the *sig*.
2890 * **-EAGAIN** if bpf program can try again.
2892 * u64 bpf_jiffies64(void)
2894 * Obtain the 64bit jiffies
2896 * The 64 bit jiffies
2898 * int bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
2900 * For an eBPF program attached to a perf event, retrieve the
2901 * branch records (struct perf_branch_entry) associated to *ctx*
2902 * and store it in the buffer pointed by *buf* up to size
2905 * On success, number of bytes written to *buf*. On error, a
2908 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
2909 * instead return the number of bytes required to store all the
2910 * branch entries. If this flag is set, *buf* may be NULL.
2912 * **-EINVAL** if arguments invalid or **size** not a multiple
2913 * of sizeof(struct perf_branch_entry).
2915 * **-ENOENT** if architecture does not support branch records.
2917 #define __BPF_FUNC_MAPPER(FN) \
2919 FN(map_lookup_elem), \
2920 FN(map_update_elem), \
2921 FN(map_delete_elem), \
2925 FN(get_prandom_u32), \
2926 FN(get_smp_processor_id), \
2927 FN(skb_store_bytes), \
2928 FN(l3_csum_replace), \
2929 FN(l4_csum_replace), \
2931 FN(clone_redirect), \
2932 FN(get_current_pid_tgid), \
2933 FN(get_current_uid_gid), \
2934 FN(get_current_comm), \
2935 FN(get_cgroup_classid), \
2936 FN(skb_vlan_push), \
2938 FN(skb_get_tunnel_key), \
2939 FN(skb_set_tunnel_key), \
2940 FN(perf_event_read), \
2942 FN(get_route_realm), \
2943 FN(perf_event_output), \
2944 FN(skb_load_bytes), \
2947 FN(skb_get_tunnel_opt), \
2948 FN(skb_set_tunnel_opt), \
2949 FN(skb_change_proto), \
2950 FN(skb_change_type), \
2951 FN(skb_under_cgroup), \
2952 FN(get_hash_recalc), \
2953 FN(get_current_task), \
2954 FN(probe_write_user), \
2955 FN(current_task_under_cgroup), \
2956 FN(skb_change_tail), \
2957 FN(skb_pull_data), \
2959 FN(set_hash_invalid), \
2960 FN(get_numa_node_id), \
2961 FN(skb_change_head), \
2962 FN(xdp_adjust_head), \
2963 FN(probe_read_str), \
2964 FN(get_socket_cookie), \
2965 FN(get_socket_uid), \
2968 FN(skb_adjust_room), \
2970 FN(sk_redirect_map), \
2971 FN(sock_map_update), \
2972 FN(xdp_adjust_meta), \
2973 FN(perf_event_read_value), \
2974 FN(perf_prog_read_value), \
2976 FN(override_return), \
2977 FN(sock_ops_cb_flags_set), \
2978 FN(msg_redirect_map), \
2979 FN(msg_apply_bytes), \
2980 FN(msg_cork_bytes), \
2981 FN(msg_pull_data), \
2983 FN(xdp_adjust_tail), \
2984 FN(skb_get_xfrm_state), \
2986 FN(skb_load_bytes_relative), \
2988 FN(sock_hash_update), \
2989 FN(msg_redirect_hash), \
2990 FN(sk_redirect_hash), \
2991 FN(lwt_push_encap), \
2992 FN(lwt_seg6_store_bytes), \
2993 FN(lwt_seg6_adjust_srh), \
2994 FN(lwt_seg6_action), \
2997 FN(skb_cgroup_id), \
2998 FN(get_current_cgroup_id), \
2999 FN(get_local_storage), \
3000 FN(sk_select_reuseport), \
3001 FN(skb_ancestor_cgroup_id), \
3002 FN(sk_lookup_tcp), \
3003 FN(sk_lookup_udp), \
3005 FN(map_push_elem), \
3007 FN(map_peek_elem), \
3008 FN(msg_push_data), \
3010 FN(rc_pointer_rel), \
3015 FN(skb_ecn_set_ce), \
3016 FN(get_listener_sock), \
3017 FN(skc_lookup_tcp), \
3018 FN(tcp_check_syncookie), \
3019 FN(sysctl_get_name), \
3020 FN(sysctl_get_current_value), \
3021 FN(sysctl_get_new_value), \
3022 FN(sysctl_set_new_value), \
3025 FN(sk_storage_get), \
3026 FN(sk_storage_delete), \
3028 FN(tcp_gen_syncookie), \
3030 FN(probe_read_user), \
3031 FN(probe_read_kernel), \
3032 FN(probe_read_user_str), \
3033 FN(probe_read_kernel_str), \
3035 FN(send_signal_thread), \
3037 FN(read_branch_records),
3039 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
3040 * function eBPF program intends to call
3042 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
3044 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
3047 #undef __BPF_ENUM_FN
3049 /* All flags used by eBPF helper functions, placed here. */
3051 /* BPF_FUNC_skb_store_bytes flags. */
3053 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3054 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3057 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3058 * First 4 bits are for passing the header field size.
3061 BPF_F_HDR_FIELD_MASK = 0xfULL,
3064 /* BPF_FUNC_l4_csum_replace flags. */
3066 BPF_F_PSEUDO_HDR = (1ULL << 4),
3067 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3068 BPF_F_MARK_ENFORCE = (1ULL << 6),
3071 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
3073 BPF_F_INGRESS = (1ULL << 0),
3076 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
3078 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3081 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
3083 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3084 BPF_F_USER_STACK = (1ULL << 8),
3085 /* flags used by BPF_FUNC_get_stackid only. */
3086 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3087 BPF_F_REUSE_STACKID = (1ULL << 10),
3088 /* flags used by BPF_FUNC_get_stack only. */
3089 BPF_F_USER_BUILD_ID = (1ULL << 11),
3092 /* BPF_FUNC_skb_set_tunnel_key flags. */
3094 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3095 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3096 BPF_F_SEQ_NUMBER = (1ULL << 3),
3099 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3100 * BPF_FUNC_perf_event_read_value flags.
3103 BPF_F_INDEX_MASK = 0xffffffffULL,
3104 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
3105 /* BPF_FUNC_perf_event_output for sk_buff input context. */
3106 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3109 /* Current network namespace */
3111 BPF_F_CURRENT_NETNS = (-1L),
3114 /* BPF_FUNC_skb_adjust_room flags. */
3116 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3117 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3118 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3119 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3120 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
3124 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3125 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3128 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
3129 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3130 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
3132 /* BPF_FUNC_sysctl_get_name flags. */
3134 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3137 /* BPF_FUNC_sk_storage_get flags */
3139 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0),
3142 /* BPF_FUNC_read_branch_records flags. */
3144 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3147 /* Mode for BPF_FUNC_skb_adjust_room helper. */
3148 enum bpf_adj_room_mode {
3153 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3154 enum bpf_hdr_start_off {
3159 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3160 enum bpf_lwt_encap_mode {
3162 BPF_LWT_ENCAP_SEG6_INLINE,
3166 #define __bpf_md_ptr(type, name) \
3170 } __attribute__((aligned(8)))
3172 /* user accessible mirror of in-kernel sk_buff.
3173 * new fields can only be added to the end of this structure
3179 __u32 queue_mapping;
3185 __u32 ingress_ifindex;
3195 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
3197 __u32 remote_ip4; /* Stored in network byte order */
3198 __u32 local_ip4; /* Stored in network byte order */
3199 __u32 remote_ip6[4]; /* Stored in network byte order */
3200 __u32 local_ip6[4]; /* Stored in network byte order */
3201 __u32 remote_port; /* Stored in network byte order */
3202 __u32 local_port; /* stored in host byte order */
3206 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
3210 __bpf_md_ptr(struct bpf_sock *, sk);
3214 struct bpf_tunnel_key {
3218 __u32 remote_ipv6[4];
3222 __u16 tunnel_ext; /* Padding, future use. */
3226 /* user accessible mirror of in-kernel xfrm_state.
3227 * new fields can only be added to the end of this structure
3229 struct bpf_xfrm_state {
3231 __u32 spi; /* Stored in network byte order */
3233 __u16 ext; /* Padding, future use. */
3235 __u32 remote_ipv4; /* Stored in network byte order */
3236 __u32 remote_ipv6[4]; /* Stored in network byte order */
3240 /* Generic BPF return codes which all BPF program types may support.
3241 * The values are binary compatible with their TC_ACT_* counter-part to
3242 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3245 * XDP is handled seprately, see XDP_*.
3253 /* >127 are reserved for prog type specific return codes.
3255 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3256 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3257 * changed and should be routed based on its new L3 header.
3258 * (This is an L3 redirect, as opposed to L2 redirect
3259 * represented by BPF_REDIRECT above).
3261 BPF_LWT_REROUTE = 128,
3271 /* IP address also allows 1 and 2 bytes access */
3274 __u32 src_port; /* host byte order */
3275 __u32 dst_port; /* network byte order */
3281 struct bpf_tcp_sock {
3282 __u32 snd_cwnd; /* Sending congestion window */
3283 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3285 __u32 snd_ssthresh; /* Slow start size threshold */
3286 __u32 rcv_nxt; /* What we want to receive next */
3287 __u32 snd_nxt; /* Next sequence we send */
3288 __u32 snd_una; /* First byte we want an ack for */
3289 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3290 __u32 ecn_flags; /* ECN status bits. */
3291 __u32 rate_delivered; /* saved rate sample: packets delivered */
3292 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3293 __u32 packets_out; /* Packets which are "in flight" */
3294 __u32 retrans_out; /* Retransmitted packets out */
3295 __u32 total_retrans; /* Total retransmits for entire connection */
3296 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3297 * total number of segments in.
3299 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3300 * total number of data segments in.
3302 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3303 * The total number of segments sent.
3305 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3306 * total number of data segments sent.
3308 __u32 lost_out; /* Lost packets */
3309 __u32 sacked_out; /* SACK'd packets */
3310 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3311 * sum(delta(rcv_nxt)), or how many bytes
3314 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3315 * sum(delta(snd_una)), or how many bytes
3318 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3319 * total number of DSACK blocks received
3321 __u32 delivered; /* Total data packets delivered incl. rexmits */
3322 __u32 delivered_ce; /* Like the above but only ECE marked packets */
3323 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
3326 struct bpf_sock_tuple {
3343 struct bpf_xdp_sock {
3347 #define XDP_PACKET_HEADROOM 256
3349 /* User return codes for XDP prog type.
3350 * A valid XDP program must return one of these defined values. All other
3351 * return codes are reserved for future use. Unknown return codes will
3352 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3362 /* user accessible metadata for XDP packet hook
3363 * new fields must be added to the end of this structure
3369 /* Below access go through struct xdp_rxq_info */
3370 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3371 __u32 rx_queue_index; /* rxq->queue_index */
3379 /* user accessible metadata for SK_MSG packet hook, new fields must
3380 * be added to the end of this structure
3383 __bpf_md_ptr(void *, data);
3384 __bpf_md_ptr(void *, data_end);
3387 __u32 remote_ip4; /* Stored in network byte order */
3388 __u32 local_ip4; /* Stored in network byte order */
3389 __u32 remote_ip6[4]; /* Stored in network byte order */
3390 __u32 local_ip6[4]; /* Stored in network byte order */
3391 __u32 remote_port; /* Stored in network byte order */
3392 __u32 local_port; /* stored in host byte order */
3393 __u32 size; /* Total size of sk_msg */
3396 struct sk_reuseport_md {
3398 * Start of directly accessible data. It begins from
3399 * the tcp/udp header.
3401 __bpf_md_ptr(void *, data);
3402 /* End of directly accessible data */
3403 __bpf_md_ptr(void *, data_end);
3405 * Total length of packet (starting from the tcp/udp header).
3406 * Note that the directly accessible bytes (data_end - data)
3407 * could be less than this "len". Those bytes could be
3408 * indirectly read by a helper "bpf_skb_load_bytes()".
3412 * Eth protocol in the mac header (network byte order). e.g.
3413 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3416 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3417 __u32 bind_inany; /* Is sock bound to an INANY address? */
3418 __u32 hash; /* A hash of the packet 4 tuples */
3421 #define BPF_TAG_SIZE 8
3423 struct bpf_prog_info {
3426 __u8 tag[BPF_TAG_SIZE];
3427 __u32 jited_prog_len;
3428 __u32 xlated_prog_len;
3429 __aligned_u64 jited_prog_insns;
3430 __aligned_u64 xlated_prog_insns;
3431 __u64 load_time; /* ns since boottime */
3432 __u32 created_by_uid;
3434 __aligned_u64 map_ids;
3435 char name[BPF_OBJ_NAME_LEN];
3437 __u32 gpl_compatible:1;
3438 __u32 :31; /* alignment pad */
3441 __u32 nr_jited_ksyms;
3442 __u32 nr_jited_func_lens;
3443 __aligned_u64 jited_ksyms;
3444 __aligned_u64 jited_func_lens;
3446 __u32 func_info_rec_size;
3447 __aligned_u64 func_info;
3450 __aligned_u64 line_info;
3451 __aligned_u64 jited_line_info;
3452 __u32 nr_jited_line_info;
3453 __u32 line_info_rec_size;
3454 __u32 jited_line_info_rec_size;
3456 __aligned_u64 prog_tags;
3459 } __attribute__((aligned(8)));
3461 struct bpf_map_info {
3468 char name[BPF_OBJ_NAME_LEN];
3470 __u32 btf_vmlinux_value_type_id;
3474 __u32 btf_key_type_id;
3475 __u32 btf_value_type_id;
3476 } __attribute__((aligned(8)));
3478 struct bpf_btf_info {
3482 } __attribute__((aligned(8)));
3484 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3485 * by user and intended to be used by socket (e.g. to bind to, depends on
3486 * attach attach type).
3488 struct bpf_sock_addr {
3489 __u32 user_family; /* Allows 4-byte read, but no write. */
3490 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3491 * Stored in network byte order.
3493 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3494 * Stored in network byte order.
3496 __u32 user_port; /* Allows 4-byte read and write.
3497 * Stored in network byte order
3499 __u32 family; /* Allows 4-byte read, but no write */
3500 __u32 type; /* Allows 4-byte read, but no write */
3501 __u32 protocol; /* Allows 4-byte read, but no write */
3502 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3503 * Stored in network byte order.
3505 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3506 * Stored in network byte order.
3508 __bpf_md_ptr(struct bpf_sock *, sk);
3511 /* User bpf_sock_ops struct to access socket values and specify request ops
3512 * and their replies.
3513 * Some of this fields are in network (bigendian) byte order and may need
3514 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3515 * New fields can only be added at the end of this structure
3517 struct bpf_sock_ops {
3520 __u32 args[4]; /* Optionally passed to bpf program */
3521 __u32 reply; /* Returned by bpf program */
3522 __u32 replylong[4]; /* Optionally returned by bpf prog */
3525 __u32 remote_ip4; /* Stored in network byte order */
3526 __u32 local_ip4; /* Stored in network byte order */
3527 __u32 remote_ip6[4]; /* Stored in network byte order */
3528 __u32 local_ip6[4]; /* Stored in network byte order */
3529 __u32 remote_port; /* Stored in network byte order */
3530 __u32 local_port; /* stored in host byte order */
3531 __u32 is_fullsock; /* Some TCP fields are only valid if
3532 * there is a full socket. If not, the
3533 * fields read as zero.
3536 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
3537 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3546 __u32 rate_delivered;
3547 __u32 rate_interval_us;
3550 __u32 total_retrans;
3554 __u32 data_segs_out;
3558 __u64 bytes_received;
3560 __bpf_md_ptr(struct bpf_sock *, sk);
3563 /* Definitions for bpf_sock_ops_cb_flags */
3565 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
3566 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
3567 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
3568 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
3569 /* Mask of all currently supported cb flags */
3570 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF,
3573 /* List of known BPF sock_ops operators.
3574 * New entries can only be added at the end
3578 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3579 * -1 if default value should be used
3581 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3582 * window (in packets) or -1 if default
3583 * value should be used
3585 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3586 * active connection is initialized
3588 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3589 * active connection is
3592 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3593 * passive connection is
3596 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3599 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3600 * based on the path and may be
3601 * dependent on the congestion control
3602 * algorithm. In general it indicates
3603 * a congestion threshold. RTTs above
3604 * this indicate congestion
3606 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3607 * Arg1: value of icsk_retransmits
3608 * Arg2: value of icsk_rto
3609 * Arg3: whether RTO has expired
3611 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3612 * Arg1: sequence number of 1st byte
3614 * Arg3: return value of
3615 * tcp_transmit_skb (0 => success)
3617 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3621 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3622 * socket transition to LISTEN state.
3624 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3628 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3629 * changes between the TCP and BPF versions. Ideally this should never happen.
3630 * If it does, we need to add code to convert them before calling
3631 * the BPF sock_ops function.
3634 BPF_TCP_ESTABLISHED = 1,
3644 BPF_TCP_CLOSING, /* Now a valid state */
3645 BPF_TCP_NEW_SYN_RECV,
3647 BPF_TCP_MAX_STATES /* Leave at the end! */
3651 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
3652 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
3655 struct bpf_perf_event_value {
3662 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
3663 BPF_DEVCG_ACC_READ = (1ULL << 1),
3664 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
3668 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
3669 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
3672 struct bpf_cgroup_dev_ctx {
3673 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3679 struct bpf_raw_tracepoint_args {
3683 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3684 * OUTPUT: Do lookup from egress perspective; default is ingress
3687 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
3688 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
3692 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3693 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3694 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3695 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3696 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3697 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3698 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3699 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3700 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3703 struct bpf_fib_lookup {
3704 /* input: network family for lookup (AF_INET, AF_INET6)
3705 * output: network family of egress nexthop
3709 /* set if lookup is to consider L4 data - e.g., FIB rules */
3714 /* total length of packet from network header - used for MTU check */
3717 /* input: L3 device index for lookup
3718 * output: device index from FIB lookup
3723 /* inputs to lookup */
3724 __u8 tos; /* AF_INET */
3725 __be32 flowinfo; /* AF_INET6, flow_label + priority */
3727 /* output: metric of fib result (IPv4/IPv6 only) */
3733 __u32 ipv6_src[4]; /* in6_addr; network order */
3736 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3737 * network header. output: bpf_fib_lookup sets to gateway address
3738 * if FIB lookup returns gateway route
3742 __u32 ipv6_dst[4]; /* in6_addr; network order */
3746 __be16 h_vlan_proto;
3748 __u8 smac[6]; /* ETH_ALEN */
3749 __u8 dmac[6]; /* ETH_ALEN */
3752 enum bpf_task_fd_type {
3753 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3754 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3755 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3756 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3757 BPF_FD_TYPE_UPROBE, /* filename + offset */
3758 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3762 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
3763 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
3764 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
3767 struct bpf_flow_keys {
3770 __u16 addr_proto; /* ETH_P_* of valid addrs */
3784 __u32 ipv6_src[4]; /* in6_addr; network order */
3785 __u32 ipv6_dst[4]; /* in6_addr; network order */
3792 struct bpf_func_info {
3797 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3798 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3800 struct bpf_line_info {
3802 __u32 file_name_off;
3807 struct bpf_spin_lock {
3812 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3813 * Allows 1,2,4-byte read, but no write.
3815 __u32 file_pos; /* Sysctl file position to read from, write to.
3816 * Allows 1,2,4-byte read an 4-byte write.
3820 struct bpf_sockopt {
3821 __bpf_md_ptr(struct bpf_sock *, sk);
3822 __bpf_md_ptr(void *, optval);
3823 __bpf_md_ptr(void *, optval_end);
3831 #endif /* _UAPI__LINUX_BPF_H__ */