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,
121 BPF_MAP_TYPE_PROG_ARRAY,
122 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
123 BPF_MAP_TYPE_PERCPU_HASH,
124 BPF_MAP_TYPE_PERCPU_ARRAY,
125 BPF_MAP_TYPE_STACK_TRACE,
126 BPF_MAP_TYPE_CGROUP_ARRAY,
127 BPF_MAP_TYPE_LRU_HASH,
128 BPF_MAP_TYPE_LRU_PERCPU_HASH,
129 BPF_MAP_TYPE_LPM_TRIE,
130 BPF_MAP_TYPE_ARRAY_OF_MAPS,
131 BPF_MAP_TYPE_HASH_OF_MAPS,
133 BPF_MAP_TYPE_SOCKMAP,
136 BPF_MAP_TYPE_SOCKHASH,
137 BPF_MAP_TYPE_CGROUP_STORAGE,
138 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
139 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
142 BPF_MAP_TYPE_SK_STORAGE,
143 BPF_MAP_TYPE_DEVMAP_HASH,
144 BPF_MAP_TYPE_STRUCT_OPS,
147 /* Note that tracing related programs such as
148 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
149 * are not subject to a stable API since kernel internal data
150 * structures can change from release to release and may
151 * therefore break existing tracing BPF programs. Tracing BPF
152 * programs correspond to /a/ specific kernel which is to be
153 * analyzed, and not /a/ specific kernel /and/ all future ones.
156 BPF_PROG_TYPE_UNSPEC,
157 BPF_PROG_TYPE_SOCKET_FILTER,
158 BPF_PROG_TYPE_KPROBE,
159 BPF_PROG_TYPE_SCHED_CLS,
160 BPF_PROG_TYPE_SCHED_ACT,
161 BPF_PROG_TYPE_TRACEPOINT,
163 BPF_PROG_TYPE_PERF_EVENT,
164 BPF_PROG_TYPE_CGROUP_SKB,
165 BPF_PROG_TYPE_CGROUP_SOCK,
166 BPF_PROG_TYPE_LWT_IN,
167 BPF_PROG_TYPE_LWT_OUT,
168 BPF_PROG_TYPE_LWT_XMIT,
169 BPF_PROG_TYPE_SOCK_OPS,
170 BPF_PROG_TYPE_SK_SKB,
171 BPF_PROG_TYPE_CGROUP_DEVICE,
172 BPF_PROG_TYPE_SK_MSG,
173 BPF_PROG_TYPE_RAW_TRACEPOINT,
174 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
175 BPF_PROG_TYPE_LWT_SEG6LOCAL,
176 BPF_PROG_TYPE_LIRC_MODE2,
177 BPF_PROG_TYPE_SK_REUSEPORT,
178 BPF_PROG_TYPE_FLOW_DISSECTOR,
179 BPF_PROG_TYPE_CGROUP_SYSCTL,
180 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
181 BPF_PROG_TYPE_CGROUP_SOCKOPT,
182 BPF_PROG_TYPE_TRACING,
183 BPF_PROG_TYPE_STRUCT_OPS,
188 enum bpf_attach_type {
189 BPF_CGROUP_INET_INGRESS,
190 BPF_CGROUP_INET_EGRESS,
191 BPF_CGROUP_INET_SOCK_CREATE,
193 BPF_SK_SKB_STREAM_PARSER,
194 BPF_SK_SKB_STREAM_VERDICT,
197 BPF_CGROUP_INET4_BIND,
198 BPF_CGROUP_INET6_BIND,
199 BPF_CGROUP_INET4_CONNECT,
200 BPF_CGROUP_INET6_CONNECT,
201 BPF_CGROUP_INET4_POST_BIND,
202 BPF_CGROUP_INET6_POST_BIND,
203 BPF_CGROUP_UDP4_SENDMSG,
204 BPF_CGROUP_UDP6_SENDMSG,
208 BPF_CGROUP_UDP4_RECVMSG,
209 BPF_CGROUP_UDP6_RECVMSG,
210 BPF_CGROUP_GETSOCKOPT,
211 BPF_CGROUP_SETSOCKOPT,
217 __MAX_BPF_ATTACH_TYPE
220 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
222 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
224 * NONE(default): No further bpf programs allowed in the subtree.
226 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
227 * the program in this cgroup yields to sub-cgroup program.
229 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
230 * that cgroup program gets run in addition to the program in this cgroup.
232 * Only one program is allowed to be attached to a cgroup with
233 * NONE or BPF_F_ALLOW_OVERRIDE flag.
234 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
235 * release old program and attach the new one. Attach flags has to match.
237 * Multiple programs are allowed to be attached to a cgroup with
238 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
239 * (those that were attached first, run first)
240 * The programs of sub-cgroup are executed first, then programs of
241 * this cgroup and then programs of parent cgroup.
242 * When children program makes decision (like picking TCP CA or sock bind)
243 * parent program has a chance to override it.
245 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
246 * programs for a cgroup. Though it's possible to replace an old program at
247 * any position by also specifying BPF_F_REPLACE flag and position itself in
248 * replace_bpf_fd attribute. Old program at this position will be released.
250 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
251 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
253 * cgrp1 (MULTI progs A, B) ->
254 * cgrp2 (OVERRIDE prog C) ->
255 * cgrp3 (MULTI prog D) ->
256 * cgrp4 (OVERRIDE prog E) ->
257 * cgrp5 (NONE prog F)
258 * the event in cgrp5 triggers execution of F,D,A,B in that order.
259 * if prog F is detached, the execution is E,D,A,B
260 * if prog F and D are detached, the execution is E,A,B
261 * if prog F, E and D are detached, the execution is C,A,B
263 * All eligible programs are executed regardless of return code from
266 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
267 #define BPF_F_ALLOW_MULTI (1U << 1)
268 #define BPF_F_REPLACE (1U << 2)
270 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
271 * verifier will perform strict alignment checking as if the kernel
272 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
273 * and NET_IP_ALIGN defined to 2.
275 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
277 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
278 * verifier will allow any alignment whatsoever. On platforms
279 * with strict alignment requirements for loads ands stores (such
280 * as sparc and mips) the verifier validates that all loads and
281 * stores provably follow this requirement. This flag turns that
282 * checking and enforcement off.
284 * It is mostly used for testing when we want to validate the
285 * context and memory access aspects of the verifier, but because
286 * of an unaligned access the alignment check would trigger before
287 * the one we are interested in.
289 #define BPF_F_ANY_ALIGNMENT (1U << 1)
291 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
292 * Verifier does sub-register def/use analysis and identifies instructions whose
293 * def only matters for low 32-bit, high 32-bit is never referenced later
294 * through implicit zero extension. Therefore verifier notifies JIT back-ends
295 * that it is safe to ignore clearing high 32-bit for these instructions. This
296 * saves some back-ends a lot of code-gen. However such optimization is not
297 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
298 * hence hasn't used verifier's analysis result. But, we really want to have a
299 * way to be able to verify the correctness of the described optimization on
300 * x86_64 on which testsuites are frequently exercised.
302 * So, this flag is introduced. Once it is set, verifier will randomize high
303 * 32-bit for those instructions who has been identified as safe to ignore them.
304 * Then, if verifier is not doing correct analysis, such randomization will
305 * regress tests to expose bugs.
307 #define BPF_F_TEST_RND_HI32 (1U << 2)
309 /* The verifier internal test flag. Behavior is undefined */
310 #define BPF_F_TEST_STATE_FREQ (1U << 3)
312 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
315 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
316 * insn[0].imm: map fd map fd
317 * insn[1].imm: 0 offset into value
320 * ldimm64 rewrite: address of map address of map[0]+offset
321 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
323 #define BPF_PSEUDO_MAP_FD 1
324 #define BPF_PSEUDO_MAP_VALUE 2
326 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
327 * offset to another bpf function
329 #define BPF_PSEUDO_CALL 1
331 /* flags for BPF_MAP_UPDATE_ELEM command */
333 BPF_ANY = 0, /* create new element or update existing */
334 BPF_NOEXIST = 1, /* create new element if it didn't exist */
335 BPF_EXIST = 2, /* update existing element */
336 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
339 /* flags for BPF_MAP_CREATE command */
341 BPF_F_NO_PREALLOC = (1U << 0),
342 /* Instead of having one common LRU list in the
343 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
344 * which can scale and perform better.
345 * Note, the LRU nodes (including free nodes) cannot be moved
346 * across different LRU lists.
348 BPF_F_NO_COMMON_LRU = (1U << 1),
349 /* Specify numa node during map creation */
350 BPF_F_NUMA_NODE = (1U << 2),
352 /* Flags for accessing BPF object from syscall side. */
353 BPF_F_RDONLY = (1U << 3),
354 BPF_F_WRONLY = (1U << 4),
356 /* Flag for stack_map, store build_id+offset instead of pointer */
357 BPF_F_STACK_BUILD_ID = (1U << 5),
359 /* Zero-initialize hash function seed. This should only be used for testing. */
360 BPF_F_ZERO_SEED = (1U << 6),
362 /* Flags for accessing BPF object from program side. */
363 BPF_F_RDONLY_PROG = (1U << 7),
364 BPF_F_WRONLY_PROG = (1U << 8),
366 /* Clone map from listener for newly accepted socket */
367 BPF_F_CLONE = (1U << 9),
369 /* Enable memory-mapping BPF map */
370 BPF_F_MMAPABLE = (1U << 10),
373 /* Flags for BPF_PROG_QUERY. */
375 /* Query effective (directly attached + inherited from ancestor cgroups)
376 * programs that will be executed for events within a cgroup.
377 * attach_flags with this flag are returned only for directly attached programs.
379 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
381 enum bpf_stack_build_id_status {
382 /* user space need an empty entry to identify end of a trace */
383 BPF_STACK_BUILD_ID_EMPTY = 0,
384 /* with valid build_id and offset */
385 BPF_STACK_BUILD_ID_VALID = 1,
386 /* couldn't get build_id, fallback to ip */
387 BPF_STACK_BUILD_ID_IP = 2,
390 #define BPF_BUILD_ID_SIZE 20
391 struct bpf_stack_build_id {
393 unsigned char build_id[BPF_BUILD_ID_SIZE];
400 #define BPF_OBJ_NAME_LEN 16U
403 struct { /* anonymous struct used by BPF_MAP_CREATE command */
404 __u32 map_type; /* one of enum bpf_map_type */
405 __u32 key_size; /* size of key in bytes */
406 __u32 value_size; /* size of value in bytes */
407 __u32 max_entries; /* max number of entries in a map */
408 __u32 map_flags; /* BPF_MAP_CREATE related
409 * flags defined above.
411 __u32 inner_map_fd; /* fd pointing to the inner map */
412 __u32 numa_node; /* numa node (effective only if
413 * BPF_F_NUMA_NODE is set).
415 char map_name[BPF_OBJ_NAME_LEN];
416 __u32 map_ifindex; /* ifindex of netdev to create on */
417 __u32 btf_fd; /* fd pointing to a BTF type data */
418 __u32 btf_key_type_id; /* BTF type_id of the key */
419 __u32 btf_value_type_id; /* BTF type_id of the value */
420 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
421 * struct stored as the
426 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
431 __aligned_u64 next_key;
436 struct { /* struct used by BPF_MAP_*_BATCH commands */
437 __aligned_u64 in_batch; /* start batch,
438 * NULL to start from beginning
440 __aligned_u64 out_batch; /* output: next start batch */
442 __aligned_u64 values;
443 __u32 count; /* input/output:
444 * input: # of key/value
446 * output: # of filled elements
453 struct { /* anonymous struct used by BPF_PROG_LOAD command */
454 __u32 prog_type; /* one of enum bpf_prog_type */
457 __aligned_u64 license;
458 __u32 log_level; /* verbosity level of verifier */
459 __u32 log_size; /* size of user buffer */
460 __aligned_u64 log_buf; /* user supplied buffer */
461 __u32 kern_version; /* not used */
463 char prog_name[BPF_OBJ_NAME_LEN];
464 __u32 prog_ifindex; /* ifindex of netdev to prep for */
465 /* For some prog types expected attach type must be known at
466 * load time to verify attach type specific parts of prog
467 * (context accesses, allowed helpers, etc).
469 __u32 expected_attach_type;
470 __u32 prog_btf_fd; /* fd pointing to BTF type data */
471 __u32 func_info_rec_size; /* userspace bpf_func_info size */
472 __aligned_u64 func_info; /* func info */
473 __u32 func_info_cnt; /* number of bpf_func_info records */
474 __u32 line_info_rec_size; /* userspace bpf_line_info size */
475 __aligned_u64 line_info; /* line info */
476 __u32 line_info_cnt; /* number of bpf_line_info records */
477 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
478 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
481 struct { /* anonymous struct used by BPF_OBJ_* commands */
482 __aligned_u64 pathname;
487 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
488 __u32 target_fd; /* container object to attach to */
489 __u32 attach_bpf_fd; /* eBPF program to attach */
492 __u32 replace_bpf_fd; /* previously attached eBPF
493 * program to replace if
494 * BPF_F_REPLACE is used
498 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
501 __u32 data_size_in; /* input: len of data_in */
502 __u32 data_size_out; /* input/output: len of data_out
503 * returns ENOSPC if data_out
506 __aligned_u64 data_in;
507 __aligned_u64 data_out;
510 __u32 ctx_size_in; /* input: len of ctx_in */
511 __u32 ctx_size_out; /* input/output: len of ctx_out
512 * returns ENOSPC if ctx_out
515 __aligned_u64 ctx_in;
516 __aligned_u64 ctx_out;
519 struct { /* anonymous struct used by BPF_*_GET_*_ID */
530 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
536 struct { /* anonymous struct used by BPF_PROG_QUERY command */
537 __u32 target_fd; /* container object to query */
541 __aligned_u64 prog_ids;
545 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
550 struct { /* anonymous struct for BPF_BTF_LOAD */
552 __aligned_u64 btf_log_buf;
559 __u32 pid; /* input: pid */
560 __u32 fd; /* input: fd */
561 __u32 flags; /* input: flags */
562 __u32 buf_len; /* input/output: buf len */
563 __aligned_u64 buf; /* input/output:
564 * tp_name for tracepoint
566 * filename for uprobe
568 __u32 prog_id; /* output: prod_id */
569 __u32 fd_type; /* output: BPF_FD_TYPE_* */
570 __u64 probe_offset; /* output: probe_offset */
571 __u64 probe_addr; /* output: probe_addr */
574 struct { /* struct used by BPF_LINK_CREATE command */
575 __u32 prog_fd; /* eBPF program to attach */
576 __u32 target_fd; /* object to attach to */
577 __u32 attach_type; /* attach type */
578 __u32 flags; /* extra flags */
580 } __attribute__((aligned(8)));
582 /* The description below is an attempt at providing documentation to eBPF
583 * developers about the multiple available eBPF helper functions. It can be
584 * parsed and used to produce a manual page. The workflow is the following,
585 * and requires the rst2man utility:
587 * $ ./scripts/bpf_helpers_doc.py \
588 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
589 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
590 * $ man /tmp/bpf-helpers.7
592 * Note that in order to produce this external documentation, some RST
593 * formatting is used in the descriptions to get "bold" and "italics" in
594 * manual pages. Also note that the few trailing white spaces are
595 * intentional, removing them would break paragraphs for rst2man.
597 * Start of BPF helper function descriptions:
599 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
601 * Perform a lookup in *map* for an entry associated to *key*.
603 * Map value associated to *key*, or **NULL** if no entry was
606 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
608 * Add or update the value of the entry associated to *key* in
609 * *map* with *value*. *flags* is one of:
612 * The entry for *key* must not exist in the map.
614 * The entry for *key* must already exist in the map.
616 * No condition on the existence of the entry for *key*.
618 * Flag value **BPF_NOEXIST** cannot be used for maps of types
619 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
620 * elements always exist), the helper would return an error.
622 * 0 on success, or a negative error in case of failure.
624 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
626 * Delete entry with *key* from *map*.
628 * 0 on success, or a negative error in case of failure.
630 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
632 * For tracing programs, safely attempt to read *size* bytes from
633 * kernel space address *unsafe_ptr* and store the data in *dst*.
635 * Generally, use bpf_probe_read_user() or bpf_probe_read_kernel()
638 * 0 on success, or a negative error in case of failure.
640 * u64 bpf_ktime_get_ns(void)
642 * Return the time elapsed since system boot, in nanoseconds.
646 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
648 * This helper is a "printk()-like" facility for debugging. It
649 * prints a message defined by format *fmt* (of size *fmt_size*)
650 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
651 * available. It can take up to three additional **u64**
652 * arguments (as an eBPF helpers, the total number of arguments is
655 * Each time the helper is called, it appends a line to the trace.
656 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
657 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
658 * The format of the trace is customizable, and the exact output
659 * one will get depends on the options set in
660 * *\/sys/kernel/debug/tracing/trace_options* (see also the
661 * *README* file under the same directory). However, it usually
662 * defaults to something like:
666 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
670 * * ``telnet`` is the name of the current task.
671 * * ``470`` is the PID of the current task.
672 * * ``001`` is the CPU number on which the task is
674 * * In ``.N..``, each character refers to a set of
675 * options (whether irqs are enabled, scheduling
676 * options, whether hard/softirqs are running, level of
677 * preempt_disabled respectively). **N** means that
678 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
680 * * ``419421.045894`` is a timestamp.
681 * * ``0x00000001`` is a fake value used by BPF for the
682 * instruction pointer register.
683 * * ``<formatted msg>`` is the message formatted with
686 * The conversion specifiers supported by *fmt* are similar, but
687 * more limited than for printk(). They are **%d**, **%i**,
688 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
689 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
690 * of field, padding with zeroes, etc.) is available, and the
691 * helper will return **-EINVAL** (but print nothing) if it
692 * encounters an unknown specifier.
694 * Also, note that **bpf_trace_printk**\ () is slow, and should
695 * only be used for debugging purposes. For this reason, a notice
696 * bloc (spanning several lines) is printed to kernel logs and
697 * states that the helper should not be used "for production use"
698 * the first time this helper is used (or more precisely, when
699 * **trace_printk**\ () buffers are allocated). For passing values
700 * to user space, perf events should be preferred.
702 * The number of bytes written to the buffer, or a negative error
703 * in case of failure.
705 * u32 bpf_get_prandom_u32(void)
707 * Get a pseudo-random number.
709 * From a security point of view, this helper uses its own
710 * pseudo-random internal state, and cannot be used to infer the
711 * seed of other random functions in the kernel. However, it is
712 * essential to note that the generator used by the helper is not
713 * cryptographically secure.
715 * A random 32-bit unsigned value.
717 * u32 bpf_get_smp_processor_id(void)
719 * Get the SMP (symmetric multiprocessing) processor id. Note that
720 * all programs run with preemption disabled, which means that the
721 * SMP processor id is stable during all the execution of the
724 * The SMP id of the processor running the program.
726 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
728 * Store *len* bytes from address *from* into the packet
729 * associated to *skb*, at *offset*. *flags* are a combination of
730 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
731 * checksum for the packet after storing the bytes) and
732 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
733 * **->swhash** and *skb*\ **->l4hash** to 0).
735 * A call to this helper is susceptible to change the underlying
736 * packet buffer. Therefore, at load time, all checks on pointers
737 * previously done by the verifier are invalidated and must be
738 * performed again, if the helper is used in combination with
739 * direct packet access.
741 * 0 on success, or a negative error in case of failure.
743 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
745 * Recompute the layer 3 (e.g. IP) checksum for the packet
746 * associated to *skb*. Computation is incremental, so the helper
747 * must know the former value of the header field that was
748 * modified (*from*), the new value of this field (*to*), and the
749 * number of bytes (2 or 4) for this field, stored in *size*.
750 * Alternatively, it is possible to store the difference between
751 * the previous and the new values of the header field in *to*, by
752 * setting *from* and *size* to 0. For both methods, *offset*
753 * indicates the location of the IP checksum within the packet.
755 * This helper works in combination with **bpf_csum_diff**\ (),
756 * which does not update the checksum in-place, but offers more
757 * flexibility and can handle sizes larger than 2 or 4 for the
758 * checksum to update.
760 * A call to this helper is susceptible to change the underlying
761 * packet buffer. Therefore, at load time, all checks on pointers
762 * previously done by the verifier are invalidated and must be
763 * performed again, if the helper is used in combination with
764 * direct packet access.
766 * 0 on success, or a negative error in case of failure.
768 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
770 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
771 * packet associated to *skb*. Computation is incremental, so the
772 * helper must know the former value of the header field that was
773 * modified (*from*), the new value of this field (*to*), and the
774 * number of bytes (2 or 4) for this field, stored on the lowest
775 * four bits of *flags*. Alternatively, it is possible to store
776 * the difference between the previous and the new values of the
777 * header field in *to*, by setting *from* and the four lowest
778 * bits of *flags* to 0. For both methods, *offset* indicates the
779 * location of the IP checksum within the packet. In addition to
780 * the size of the field, *flags* can be added (bitwise OR) actual
781 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
782 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
783 * for updates resulting in a null checksum the value is set to
784 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
785 * the checksum is to be computed against a pseudo-header.
787 * This helper works in combination with **bpf_csum_diff**\ (),
788 * which does not update the checksum in-place, but offers more
789 * flexibility and can handle sizes larger than 2 or 4 for the
790 * checksum to update.
792 * A call to this helper is susceptible to change the underlying
793 * packet buffer. Therefore, at load time, all checks on pointers
794 * previously done by the verifier are invalidated and must be
795 * performed again, if the helper is used in combination with
796 * direct packet access.
798 * 0 on success, or a negative error in case of failure.
800 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
802 * This special helper is used to trigger a "tail call", or in
803 * other words, to jump into another eBPF program. The same stack
804 * frame is used (but values on stack and in registers for the
805 * caller are not accessible to the callee). This mechanism allows
806 * for program chaining, either for raising the maximum number of
807 * available eBPF instructions, or to execute given programs in
808 * conditional blocks. For security reasons, there is an upper
809 * limit to the number of successive tail calls that can be
812 * Upon call of this helper, the program attempts to jump into a
813 * program referenced at index *index* in *prog_array_map*, a
814 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
815 * *ctx*, a pointer to the context.
817 * If the call succeeds, the kernel immediately runs the first
818 * instruction of the new program. This is not a function call,
819 * and it never returns to the previous program. If the call
820 * fails, then the helper has no effect, and the caller continues
821 * to run its subsequent instructions. A call can fail if the
822 * destination program for the jump does not exist (i.e. *index*
823 * is superior to the number of entries in *prog_array_map*), or
824 * if the maximum number of tail calls has been reached for this
825 * chain of programs. This limit is defined in the kernel by the
826 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
827 * which is currently set to 32.
829 * 0 on success, or a negative error in case of failure.
831 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
833 * Clone and redirect the packet associated to *skb* to another
834 * net device of index *ifindex*. Both ingress and egress
835 * interfaces can be used for redirection. The **BPF_F_INGRESS**
836 * value in *flags* is used to make the distinction (ingress path
837 * is selected if the flag is present, egress path otherwise).
838 * This is the only flag supported for now.
840 * In comparison with **bpf_redirect**\ () helper,
841 * **bpf_clone_redirect**\ () has the associated cost of
842 * duplicating the packet buffer, but this can be executed out of
843 * the eBPF program. Conversely, **bpf_redirect**\ () is more
844 * efficient, but it is handled through an action code where the
845 * redirection happens only after the eBPF program has returned.
847 * A call to this helper is susceptible to change the underlying
848 * packet buffer. Therefore, at load time, all checks on pointers
849 * previously done by the verifier are invalidated and must be
850 * performed again, if the helper is used in combination with
851 * direct packet access.
853 * 0 on success, or a negative error in case of failure.
855 * u64 bpf_get_current_pid_tgid(void)
857 * A 64-bit integer containing the current tgid and pid, and
859 * *current_task*\ **->tgid << 32 \|**
860 * *current_task*\ **->pid**.
862 * u64 bpf_get_current_uid_gid(void)
864 * A 64-bit integer containing the current GID and UID, and
865 * created as such: *current_gid* **<< 32 \|** *current_uid*.
867 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
869 * Copy the **comm** attribute of the current task into *buf* of
870 * *size_of_buf*. The **comm** attribute contains the name of
871 * the executable (excluding the path) for the current task. The
872 * *size_of_buf* must be strictly positive. On success, the
873 * helper makes sure that the *buf* is NUL-terminated. On failure,
874 * it is filled with zeroes.
876 * 0 on success, or a negative error in case of failure.
878 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
880 * Retrieve the classid for the current task, i.e. for the net_cls
881 * cgroup to which *skb* belongs.
883 * This helper can be used on TC egress path, but not on ingress.
885 * The net_cls cgroup provides an interface to tag network packets
886 * based on a user-provided identifier for all traffic coming from
887 * the tasks belonging to the related cgroup. See also the related
888 * kernel documentation, available from the Linux sources in file
889 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
891 * The Linux kernel has two versions for cgroups: there are
892 * cgroups v1 and cgroups v2. Both are available to users, who can
893 * use a mixture of them, but note that the net_cls cgroup is for
894 * cgroup v1 only. This makes it incompatible with BPF programs
895 * run on cgroups, which is a cgroup-v2-only feature (a socket can
896 * only hold data for one version of cgroups at a time).
898 * This helper is only available is the kernel was compiled with
899 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
900 * "**y**" or to "**m**".
902 * The classid, or 0 for the default unconfigured classid.
904 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
906 * Push a *vlan_tci* (VLAN tag control information) of protocol
907 * *vlan_proto* to the packet associated to *skb*, then update
908 * the checksum. Note that if *vlan_proto* is different from
909 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
910 * be **ETH_P_8021Q**.
912 * A call to this helper is susceptible to change the underlying
913 * packet buffer. Therefore, at load time, all checks on pointers
914 * previously done by the verifier are invalidated and must be
915 * performed again, if the helper is used in combination with
916 * direct packet access.
918 * 0 on success, or a negative error in case of failure.
920 * int bpf_skb_vlan_pop(struct sk_buff *skb)
922 * Pop a VLAN header from the packet associated to *skb*.
924 * A call to this helper is susceptible to change the underlying
925 * packet buffer. Therefore, at load time, all checks on pointers
926 * previously done by the verifier are invalidated and must be
927 * performed again, if the helper is used in combination with
928 * direct packet access.
930 * 0 on success, or a negative error in case of failure.
932 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
934 * Get tunnel metadata. This helper takes a pointer *key* to an
935 * empty **struct bpf_tunnel_key** of **size**, that will be
936 * filled with tunnel metadata for the packet associated to *skb*.
937 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
938 * indicates that the tunnel is based on IPv6 protocol instead of
941 * The **struct bpf_tunnel_key** is an object that generalizes the
942 * principal parameters used by various tunneling protocols into a
943 * single struct. This way, it can be used to easily make a
944 * decision based on the contents of the encapsulation header,
945 * "summarized" in this struct. In particular, it holds the IP
946 * address of the remote end (IPv4 or IPv6, depending on the case)
947 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
948 * this struct exposes the *key*\ **->tunnel_id**, which is
949 * generally mapped to a VNI (Virtual Network Identifier), making
950 * it programmable together with the **bpf_skb_set_tunnel_key**\
953 * Let's imagine that the following code is part of a program
954 * attached to the TC ingress interface, on one end of a GRE
955 * tunnel, and is supposed to filter out all messages coming from
956 * remote ends with IPv4 address other than 10.0.0.1:
961 * struct bpf_tunnel_key key = {};
963 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
965 * return TC_ACT_SHOT; // drop packet
967 * if (key.remote_ipv4 != 0x0a000001)
968 * return TC_ACT_SHOT; // drop packet
970 * return TC_ACT_OK; // accept packet
972 * This interface can also be used with all encapsulation devices
973 * that can operate in "collect metadata" mode: instead of having
974 * one network device per specific configuration, the "collect
975 * metadata" mode only requires a single device where the
976 * configuration can be extracted from this helper.
978 * This can be used together with various tunnels such as VXLan,
979 * Geneve, GRE or IP in IP (IPIP).
981 * 0 on success, or a negative error in case of failure.
983 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
985 * Populate tunnel metadata for packet associated to *skb.* The
986 * tunnel metadata is set to the contents of *key*, of *size*. The
987 * *flags* can be set to a combination of the following values:
989 * **BPF_F_TUNINFO_IPV6**
990 * Indicate that the tunnel is based on IPv6 protocol
992 * **BPF_F_ZERO_CSUM_TX**
993 * For IPv4 packets, add a flag to tunnel metadata
994 * indicating that checksum computation should be skipped
995 * and checksum set to zeroes.
996 * **BPF_F_DONT_FRAGMENT**
997 * Add a flag to tunnel metadata indicating that the
998 * packet should not be fragmented.
999 * **BPF_F_SEQ_NUMBER**
1000 * Add a flag to tunnel metadata indicating that a
1001 * sequence number should be added to tunnel header before
1002 * sending the packet. This flag was added for GRE
1003 * encapsulation, but might be used with other protocols
1004 * as well in the future.
1006 * Here is a typical usage on the transmit path:
1010 * struct bpf_tunnel_key key;
1012 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1013 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1015 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1016 * helper for additional information.
1018 * 0 on success, or a negative error in case of failure.
1020 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1022 * Read the value of a perf event counter. This helper relies on a
1023 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1024 * the perf event counter is selected when *map* is updated with
1025 * perf event file descriptors. The *map* is an array whose size
1026 * is the number of available CPUs, and each cell contains a value
1027 * relative to one CPU. The value to retrieve is indicated by
1028 * *flags*, that contains the index of the CPU to look up, masked
1029 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1030 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1031 * current CPU should be retrieved.
1033 * Note that before Linux 4.13, only hardware perf event can be
1036 * Also, be aware that the newer helper
1037 * **bpf_perf_event_read_value**\ () is recommended over
1038 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1039 * quirks where error and counter value are used as a return code
1040 * (which is wrong to do since ranges may overlap). This issue is
1041 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1042 * time provides more features over the **bpf_perf_event_read**\
1043 * () interface. Please refer to the description of
1044 * **bpf_perf_event_read_value**\ () for details.
1046 * The value of the perf event counter read from the map, or a
1047 * negative error code in case of failure.
1049 * int bpf_redirect(u32 ifindex, u64 flags)
1051 * Redirect the packet to another net device of index *ifindex*.
1052 * This helper is somewhat similar to **bpf_clone_redirect**\
1053 * (), except that the packet is not cloned, which provides
1054 * increased performance.
1056 * Except for XDP, both ingress and egress interfaces can be used
1057 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1058 * to make the distinction (ingress path is selected if the flag
1059 * is present, egress path otherwise). Currently, XDP only
1060 * supports redirection to the egress interface, and accepts no
1063 * The same effect can also be attained with the more generic
1064 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1065 * redirect target instead of providing it directly to the helper.
1067 * For XDP, the helper returns **XDP_REDIRECT** on success or
1068 * **XDP_ABORTED** on error. For other program types, the values
1069 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1072 * u32 bpf_get_route_realm(struct sk_buff *skb)
1074 * Retrieve the realm or the route, that is to say the
1075 * **tclassid** field of the destination for the *skb*. The
1076 * indentifier retrieved is a user-provided tag, similar to the
1077 * one used with the net_cls cgroup (see description for
1078 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1079 * held by a route (a destination entry), not by a task.
1081 * Retrieving this identifier works with the clsact TC egress hook
1082 * (see also **tc-bpf(8)**), or alternatively on conventional
1083 * classful egress qdiscs, but not on TC ingress path. In case of
1084 * clsact TC egress hook, this has the advantage that, internally,
1085 * the destination entry has not been dropped yet in the transmit
1086 * path. Therefore, the destination entry does not need to be
1087 * artificially held via **netif_keep_dst**\ () for a classful
1088 * qdisc until the *skb* is freed.
1090 * This helper is available only if the kernel was compiled with
1091 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1093 * The realm of the route for the packet associated to *skb*, or 0
1094 * if none was found.
1096 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1098 * Write raw *data* blob into a special BPF perf event held by
1099 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1100 * event must have the following attributes: **PERF_SAMPLE_RAW**
1101 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1102 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1104 * The *flags* are used to indicate the index in *map* for which
1105 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1106 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1107 * to indicate that the index of the current CPU core should be
1110 * The value to write, of *size*, is passed through eBPF stack and
1111 * pointed by *data*.
1113 * The context of the program *ctx* needs also be passed to the
1116 * On user space, a program willing to read the values needs to
1117 * call **perf_event_open**\ () on the perf event (either for
1118 * one or for all CPUs) and to store the file descriptor into the
1119 * *map*. This must be done before the eBPF program can send data
1120 * into it. An example is available in file
1121 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1122 * tree (the eBPF program counterpart is in
1123 * *samples/bpf/trace_output_kern.c*).
1125 * **bpf_perf_event_output**\ () achieves better performance
1126 * than **bpf_trace_printk**\ () for sharing data with user
1127 * space, and is much better suitable for streaming data from eBPF
1130 * Note that this helper is not restricted to tracing use cases
1131 * and can be used with programs attached to TC or XDP as well,
1132 * where it allows for passing data to user space listeners. Data
1135 * * Only custom structs,
1136 * * Only the packet payload, or
1137 * * A combination of both.
1139 * 0 on success, or a negative error in case of failure.
1141 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1143 * This helper was provided as an easy way to load data from a
1144 * packet. It can be used to load *len* bytes from *offset* from
1145 * the packet associated to *skb*, into the buffer pointed by
1148 * Since Linux 4.7, usage of this helper has mostly been replaced
1149 * by "direct packet access", enabling packet data to be
1150 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1151 * pointing respectively to the first byte of packet data and to
1152 * the byte after the last byte of packet data. However, it
1153 * remains useful if one wishes to read large quantities of data
1154 * at once from a packet into the eBPF stack.
1156 * 0 on success, or a negative error in case of failure.
1158 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1160 * Walk a user or a kernel stack and return its id. To achieve
1161 * this, the helper needs *ctx*, which is a pointer to the context
1162 * on which the tracing program is executed, and a pointer to a
1163 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1165 * The last argument, *flags*, holds the number of stack frames to
1166 * skip (from 0 to 255), masked with
1167 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1168 * a combination of the following flags:
1170 * **BPF_F_USER_STACK**
1171 * Collect a user space stack instead of a kernel stack.
1172 * **BPF_F_FAST_STACK_CMP**
1173 * Compare stacks by hash only.
1174 * **BPF_F_REUSE_STACKID**
1175 * If two different stacks hash into the same *stackid*,
1176 * discard the old one.
1178 * The stack id retrieved is a 32 bit long integer handle which
1179 * can be further combined with other data (including other stack
1180 * ids) and used as a key into maps. This can be useful for
1181 * generating a variety of graphs (such as flame graphs or off-cpu
1184 * For walking a stack, this helper is an improvement over
1185 * **bpf_probe_read**\ (), which can be used with unrolled loops
1186 * but is not efficient and consumes a lot of eBPF instructions.
1187 * Instead, **bpf_get_stackid**\ () can collect up to
1188 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1189 * this limit can be controlled with the **sysctl** program, and
1190 * that it should be manually increased in order to profile long
1191 * user stacks (such as stacks for Java programs). To do so, use:
1195 * # sysctl kernel.perf_event_max_stack=<new value>
1197 * The positive or null stack id on success, or a negative error
1198 * in case of failure.
1200 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1202 * Compute a checksum difference, from the raw buffer pointed by
1203 * *from*, of length *from_size* (that must be a multiple of 4),
1204 * towards the raw buffer pointed by *to*, of size *to_size*
1205 * (same remark). An optional *seed* can be added to the value
1206 * (this can be cascaded, the seed may come from a previous call
1209 * This is flexible enough to be used in several ways:
1211 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1212 * checksum, it can be used when pushing new data.
1213 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1214 * checksum, it can be used when removing data from a packet.
1215 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1216 * can be used to compute a diff. Note that *from_size* and
1217 * *to_size* do not need to be equal.
1219 * This helper can be used in combination with
1220 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1221 * which one can feed in the difference computed with
1222 * **bpf_csum_diff**\ ().
1224 * The checksum result, or a negative error code in case of
1227 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1229 * Retrieve tunnel options metadata for the packet associated to
1230 * *skb*, and store the raw tunnel option data to the buffer *opt*
1233 * This helper can be used with encapsulation devices that can
1234 * operate in "collect metadata" mode (please refer to the related
1235 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1236 * more details). A particular example where this can be used is
1237 * in combination with the Geneve encapsulation protocol, where it
1238 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1239 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1240 * the eBPF program. This allows for full customization of these
1243 * The size of the option data retrieved.
1245 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1247 * Set tunnel options metadata for the packet associated to *skb*
1248 * to the option data contained in the raw buffer *opt* of *size*.
1250 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1251 * helper for additional information.
1253 * 0 on success, or a negative error in case of failure.
1255 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1257 * Change the protocol of the *skb* to *proto*. Currently
1258 * supported are transition from IPv4 to IPv6, and from IPv6 to
1259 * IPv4. The helper takes care of the groundwork for the
1260 * transition, including resizing the socket buffer. The eBPF
1261 * program is expected to fill the new headers, if any, via
1262 * **skb_store_bytes**\ () and to recompute the checksums with
1263 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1264 * (). The main case for this helper is to perform NAT64
1265 * operations out of an eBPF program.
1267 * Internally, the GSO type is marked as dodgy so that headers are
1268 * checked and segments are recalculated by the GSO/GRO engine.
1269 * The size for GSO target is adapted as well.
1271 * All values for *flags* are reserved for future usage, and must
1274 * A call to this helper is susceptible to change the underlying
1275 * packet buffer. Therefore, at load time, all checks on pointers
1276 * previously done by the verifier are invalidated and must be
1277 * performed again, if the helper is used in combination with
1278 * direct packet access.
1280 * 0 on success, or a negative error in case of failure.
1282 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1284 * Change the packet type for the packet associated to *skb*. This
1285 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1286 * the eBPF program does not have a write access to *skb*\
1287 * **->pkt_type** beside this helper. Using a helper here allows
1288 * for graceful handling of errors.
1290 * The major use case is to change incoming *skb*s to
1291 * **PACKET_HOST** in a programmatic way instead of having to
1292 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1295 * Note that *type* only allows certain values. At this time, they
1300 * **PACKET_BROADCAST**
1301 * Send packet to all.
1302 * **PACKET_MULTICAST**
1303 * Send packet to group.
1304 * **PACKET_OTHERHOST**
1305 * Send packet to someone else.
1307 * 0 on success, or a negative error in case of failure.
1309 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1311 * Check whether *skb* is a descendant of the cgroup2 held by
1312 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1314 * The return value depends on the result of the test, and can be:
1316 * * 0, if the *skb* failed the cgroup2 descendant test.
1317 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1318 * * A negative error code, if an error occurred.
1320 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1322 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1323 * not set, in particular if the hash was cleared due to mangling,
1324 * recompute this hash. Later accesses to the hash can be done
1325 * directly with *skb*\ **->hash**.
1327 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1328 * prototype with **bpf_skb_change_proto**\ (), or calling
1329 * **bpf_skb_store_bytes**\ () with the
1330 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1331 * the hash and to trigger a new computation for the next call to
1332 * **bpf_get_hash_recalc**\ ().
1336 * u64 bpf_get_current_task(void)
1338 * A pointer to the current task struct.
1340 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1342 * Attempt in a safe way to write *len* bytes from the buffer
1343 * *src* to *dst* in memory. It only works for threads that are in
1344 * user context, and *dst* must be a valid user space address.
1346 * This helper should not be used to implement any kind of
1347 * security mechanism because of TOC-TOU attacks, but rather to
1348 * debug, divert, and manipulate execution of semi-cooperative
1351 * Keep in mind that this feature is meant for experiments, and it
1352 * has a risk of crashing the system and running programs.
1353 * Therefore, when an eBPF program using this helper is attached,
1354 * a warning including PID and process name is printed to kernel
1357 * 0 on success, or a negative error in case of failure.
1359 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1361 * Check whether the probe is being run is the context of a given
1362 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1363 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1365 * The return value depends on the result of the test, and can be:
1367 * * 0, if the *skb* task belongs to the cgroup2.
1368 * * 1, if the *skb* task does not belong to the cgroup2.
1369 * * A negative error code, if an error occurred.
1371 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1373 * Resize (trim or grow) the packet associated to *skb* to the
1374 * new *len*. The *flags* are reserved for future usage, and must
1377 * The basic idea is that the helper performs the needed work to
1378 * change the size of the packet, then the eBPF program rewrites
1379 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1380 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1381 * and others. This helper is a slow path utility intended for
1382 * replies with control messages. And because it is targeted for
1383 * slow path, the helper itself can afford to be slow: it
1384 * implicitly linearizes, unclones and drops offloads from the
1387 * A call to this helper is susceptible to change the underlying
1388 * packet buffer. Therefore, at load time, all checks on pointers
1389 * previously done by the verifier are invalidated and must be
1390 * performed again, if the helper is used in combination with
1391 * direct packet access.
1393 * 0 on success, or a negative error in case of failure.
1395 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1397 * Pull in non-linear data in case the *skb* is non-linear and not
1398 * all of *len* are part of the linear section. Make *len* bytes
1399 * from *skb* readable and writable. If a zero value is passed for
1400 * *len*, then the whole length of the *skb* is pulled.
1402 * This helper is only needed for reading and writing with direct
1405 * For direct packet access, testing that offsets to access
1406 * are within packet boundaries (test on *skb*\ **->data_end**) is
1407 * susceptible to fail if offsets are invalid, or if the requested
1408 * data is in non-linear parts of the *skb*. On failure the
1409 * program can just bail out, or in the case of a non-linear
1410 * buffer, use a helper to make the data available. The
1411 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1412 * the data. Another one consists in using **bpf_skb_pull_data**
1413 * to pull in once the non-linear parts, then retesting and
1414 * eventually access the data.
1416 * At the same time, this also makes sure the *skb* is uncloned,
1417 * which is a necessary condition for direct write. As this needs
1418 * to be an invariant for the write part only, the verifier
1419 * detects writes and adds a prologue that is calling
1420 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1421 * the very beginning in case it is indeed cloned.
1423 * A call to this helper is susceptible to change the underlying
1424 * packet buffer. Therefore, at load time, all checks on pointers
1425 * previously done by the verifier are invalidated and must be
1426 * performed again, if the helper is used in combination with
1427 * direct packet access.
1429 * 0 on success, or a negative error in case of failure.
1431 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1433 * Add the checksum *csum* into *skb*\ **->csum** in case the
1434 * driver has supplied a checksum for the entire packet into that
1435 * field. Return an error otherwise. This helper is intended to be
1436 * used in combination with **bpf_csum_diff**\ (), in particular
1437 * when the checksum needs to be updated after data has been
1438 * written into the packet through direct packet access.
1440 * The checksum on success, or a negative error code in case of
1443 * void bpf_set_hash_invalid(struct sk_buff *skb)
1445 * Invalidate the current *skb*\ **->hash**. It can be used after
1446 * mangling on headers through direct packet access, in order to
1447 * indicate that the hash is outdated and to trigger a
1448 * recalculation the next time the kernel tries to access this
1449 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1451 * int bpf_get_numa_node_id(void)
1453 * Return the id of the current NUMA node. The primary use case
1454 * for this helper is the selection of sockets for the local NUMA
1455 * node, when the program is attached to sockets using the
1456 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1457 * but the helper is also available to other eBPF program types,
1458 * similarly to **bpf_get_smp_processor_id**\ ().
1460 * The id of current NUMA node.
1462 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1464 * Grows headroom of packet associated to *skb* and adjusts the
1465 * offset of the MAC header accordingly, adding *len* bytes of
1466 * space. It automatically extends and reallocates memory as
1469 * This helper can be used on a layer 3 *skb* to push a MAC header
1470 * for redirection into a layer 2 device.
1472 * All values for *flags* are reserved for future usage, and must
1475 * A call to this helper is susceptible to change the underlying
1476 * packet buffer. Therefore, at load time, all checks on pointers
1477 * previously done by the verifier are invalidated and must be
1478 * performed again, if the helper is used in combination with
1479 * direct packet access.
1481 * 0 on success, or a negative error in case of failure.
1483 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1485 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1486 * it is possible to use a negative value for *delta*. This helper
1487 * can be used to prepare the packet for pushing or popping
1490 * A call to this helper is susceptible to change the underlying
1491 * packet buffer. Therefore, at load time, all checks on pointers
1492 * previously done by the verifier are invalidated and must be
1493 * performed again, if the helper is used in combination with
1494 * direct packet access.
1496 * 0 on success, or a negative error in case of failure.
1498 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1500 * Copy a NUL terminated string from an unsafe kernel address
1501 * *unsafe_ptr* to *dst*. See bpf_probe_read_kernel_str() for
1504 * Generally, use bpf_probe_read_user_str() or bpf_probe_read_kernel_str()
1507 * On success, the strictly positive length of the string,
1508 * including the trailing NUL character. On error, a negative
1511 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1513 * If the **struct sk_buff** pointed by *skb* has a known socket,
1514 * retrieve the cookie (generated by the kernel) of this socket.
1515 * If no cookie has been set yet, generate a new cookie. Once
1516 * generated, the socket cookie remains stable for the life of the
1517 * socket. This helper can be useful for monitoring per socket
1518 * networking traffic statistics as it provides a global socket
1519 * identifier that can be assumed unique.
1521 * A 8-byte long non-decreasing number on success, or 0 if the
1522 * socket field is missing inside *skb*.
1524 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1526 * Equivalent to bpf_get_socket_cookie() helper that accepts
1527 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1529 * A 8-byte long non-decreasing number.
1531 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1533 * Equivalent to bpf_get_socket_cookie() helper that accepts
1534 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1536 * A 8-byte long non-decreasing number.
1538 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1540 * The owner UID of the socket associated to *skb*. If the socket
1541 * is **NULL**, or if it is not a full socket (i.e. if it is a
1542 * time-wait or a request socket instead), **overflowuid** value
1543 * is returned (note that **overflowuid** might also be the actual
1544 * UID value for the socket).
1546 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1548 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1553 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1555 * Emulate a call to **setsockopt()** on the socket associated to
1556 * *bpf_socket*, which must be a full socket. The *level* at
1557 * which the option resides and the name *optname* of the option
1558 * must be specified, see **setsockopt(2)** for more information.
1559 * The option value of length *optlen* is pointed by *optval*.
1561 * This helper actually implements a subset of **setsockopt()**.
1562 * It supports the following *level*\ s:
1564 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1565 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1566 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1567 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1568 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1569 * **TCP_BPF_SNDCWND_CLAMP**.
1570 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1571 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1573 * 0 on success, or a negative error in case of failure.
1575 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1577 * Grow or shrink the room for data in the packet associated to
1578 * *skb* by *len_diff*, and according to the selected *mode*.
1580 * There are two supported modes at this time:
1582 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1583 * (room space is added or removed below the layer 2 header).
1585 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1586 * (room space is added or removed below the layer 3 header).
1588 * The following flags are supported at this time:
1590 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1591 * Adjusting mss in this way is not allowed for datagrams.
1593 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1594 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1595 * Any new space is reserved to hold a tunnel header.
1596 * Configure skb offsets and other fields accordingly.
1598 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1599 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1600 * Use with ENCAP_L3 flags to further specify the tunnel type.
1602 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1603 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1604 * type; *len* is the length of the inner MAC header.
1606 * A call to this helper is susceptible to change the underlying
1607 * packet buffer. Therefore, at load time, all checks on pointers
1608 * previously done by the verifier are invalidated and must be
1609 * performed again, if the helper is used in combination with
1610 * direct packet access.
1612 * 0 on success, or a negative error in case of failure.
1614 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1616 * Redirect the packet to the endpoint referenced by *map* at
1617 * index *key*. Depending on its type, this *map* can contain
1618 * references to net devices (for forwarding packets through other
1619 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1620 * but this is only implemented for native XDP (with driver
1621 * support) as of this writing).
1623 * The lower two bits of *flags* are used as the return code if
1624 * the map lookup fails. This is so that the return value can be
1625 * one of the XDP program return codes up to XDP_TX, as chosen by
1626 * the caller. Any higher bits in the *flags* argument must be
1629 * See also bpf_redirect(), which only supports redirecting to an
1630 * ifindex, but doesn't require a map to do so.
1632 * **XDP_REDIRECT** on success, or the value of the two lower bits
1633 * of the **flags* argument on error.
1635 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1637 * Redirect the packet to the socket referenced by *map* (of type
1638 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1639 * egress interfaces can be used for redirection. The
1640 * **BPF_F_INGRESS** value in *flags* is used to make the
1641 * distinction (ingress path is selected if the flag is present,
1642 * egress path otherwise). This is the only flag supported for now.
1644 * **SK_PASS** on success, or **SK_DROP** on error.
1646 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1648 * Add an entry to, or update a *map* referencing sockets. The
1649 * *skops* is used as a new value for the entry associated to
1650 * *key*. *flags* is one of:
1653 * The entry for *key* must not exist in the map.
1655 * The entry for *key* must already exist in the map.
1657 * No condition on the existence of the entry for *key*.
1659 * If the *map* has eBPF programs (parser and verdict), those will
1660 * be inherited by the socket being added. If the socket is
1661 * already attached to eBPF programs, this results in an error.
1663 * 0 on success, or a negative error in case of failure.
1665 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1667 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1668 * *delta* (which can be positive or negative). Note that this
1669 * operation modifies the address stored in *xdp_md*\ **->data**,
1670 * so the latter must be loaded only after the helper has been
1673 * The use of *xdp_md*\ **->data_meta** is optional and programs
1674 * are not required to use it. The rationale is that when the
1675 * packet is processed with XDP (e.g. as DoS filter), it is
1676 * possible to push further meta data along with it before passing
1677 * to the stack, and to give the guarantee that an ingress eBPF
1678 * program attached as a TC classifier on the same device can pick
1679 * this up for further post-processing. Since TC works with socket
1680 * buffers, it remains possible to set from XDP the **mark** or
1681 * **priority** pointers, or other pointers for the socket buffer.
1682 * Having this scratch space generic and programmable allows for
1683 * more flexibility as the user is free to store whatever meta
1686 * A call to this helper is susceptible to change the underlying
1687 * packet buffer. Therefore, at load time, all checks on pointers
1688 * previously done by the verifier are invalidated and must be
1689 * performed again, if the helper is used in combination with
1690 * direct packet access.
1692 * 0 on success, or a negative error in case of failure.
1694 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1696 * Read the value of a perf event counter, and store it into *buf*
1697 * of size *buf_size*. This helper relies on a *map* of type
1698 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1699 * counter is selected when *map* is updated with perf event file
1700 * descriptors. The *map* is an array whose size is the number of
1701 * available CPUs, and each cell contains a value relative to one
1702 * CPU. The value to retrieve is indicated by *flags*, that
1703 * contains the index of the CPU to look up, masked with
1704 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1705 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1706 * current CPU should be retrieved.
1708 * This helper behaves in a way close to
1709 * **bpf_perf_event_read**\ () helper, save that instead of
1710 * just returning the value observed, it fills the *buf*
1711 * structure. This allows for additional data to be retrieved: in
1712 * particular, the enabled and running times (in *buf*\
1713 * **->enabled** and *buf*\ **->running**, respectively) are
1714 * copied. In general, **bpf_perf_event_read_value**\ () is
1715 * recommended over **bpf_perf_event_read**\ (), which has some
1716 * ABI issues and provides fewer functionalities.
1718 * These values are interesting, because hardware PMU (Performance
1719 * Monitoring Unit) counters are limited resources. When there are
1720 * more PMU based perf events opened than available counters,
1721 * kernel will multiplex these events so each event gets certain
1722 * percentage (but not all) of the PMU time. In case that
1723 * multiplexing happens, the number of samples or counter value
1724 * will not reflect the case compared to when no multiplexing
1725 * occurs. This makes comparison between different runs difficult.
1726 * Typically, the counter value should be normalized before
1727 * comparing to other experiments. The usual normalization is done
1732 * normalized_counter = counter * t_enabled / t_running
1734 * Where t_enabled is the time enabled for event and t_running is
1735 * the time running for event since last normalization. The
1736 * enabled and running times are accumulated since the perf event
1737 * open. To achieve scaling factor between two invocations of an
1738 * eBPF program, users can can use CPU id as the key (which is
1739 * typical for perf array usage model) to remember the previous
1740 * value and do the calculation inside the eBPF program.
1742 * 0 on success, or a negative error in case of failure.
1744 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1746 * For en eBPF program attached to a perf event, retrieve the
1747 * value of the event counter associated to *ctx* and store it in
1748 * the structure pointed by *buf* and of size *buf_size*. Enabled
1749 * and running times are also stored in the structure (see
1750 * description of helper **bpf_perf_event_read_value**\ () for
1753 * 0 on success, or a negative error in case of failure.
1755 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1757 * Emulate a call to **getsockopt()** on the socket associated to
1758 * *bpf_socket*, which must be a full socket. The *level* at
1759 * which the option resides and the name *optname* of the option
1760 * must be specified, see **getsockopt(2)** for more information.
1761 * The retrieved value is stored in the structure pointed by
1762 * *opval* and of length *optlen*.
1764 * This helper actually implements a subset of **getsockopt()**.
1765 * It supports the following *level*\ s:
1767 * * **IPPROTO_TCP**, which supports *optname*
1768 * **TCP_CONGESTION**.
1769 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1770 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1772 * 0 on success, or a negative error in case of failure.
1774 * int bpf_override_return(struct pt_regs *regs, u64 rc)
1776 * Used for error injection, this helper uses kprobes to override
1777 * the return value of the probed function, and to set it to *rc*.
1778 * The first argument is the context *regs* on which the kprobe
1781 * This helper works by setting setting the PC (program counter)
1782 * to an override function which is run in place of the original
1783 * probed function. This means the probed function is not run at
1784 * all. The replacement function just returns with the required
1787 * This helper has security implications, and thus is subject to
1788 * restrictions. It is only available if the kernel was compiled
1789 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1790 * option, and in this case it only works on functions tagged with
1791 * **ALLOW_ERROR_INJECTION** in the kernel code.
1793 * Also, the helper is only available for the architectures having
1794 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1795 * x86 architecture is the only one to support this feature.
1799 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1801 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1802 * for the full TCP socket associated to *bpf_sock_ops* to
1805 * The primary use of this field is to determine if there should
1806 * be calls to eBPF programs of type
1807 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1808 * code. A program of the same type can change its value, per
1809 * connection and as necessary, when the connection is
1810 * established. This field is directly accessible for reading, but
1811 * this helper must be used for updates in order to return an
1812 * error if an eBPF program tries to set a callback that is not
1813 * supported in the current kernel.
1815 * *argval* is a flag array which can combine these flags:
1817 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1818 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1819 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1820 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1822 * Therefore, this function can be used to clear a callback flag by
1823 * setting the appropriate bit to zero. e.g. to disable the RTO
1826 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1827 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1829 * Here are some examples of where one could call such eBPF
1833 * * When a packet is retransmitted.
1834 * * When the connection terminates.
1835 * * When a packet is sent.
1836 * * When a packet is received.
1838 * Code **-EINVAL** if the socket is not a full TCP socket;
1839 * otherwise, a positive number containing the bits that could not
1840 * be set is returned (which comes down to 0 if all bits were set
1843 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1845 * This helper is used in programs implementing policies at the
1846 * socket level. If the message *msg* is allowed to pass (i.e. if
1847 * the verdict eBPF program returns **SK_PASS**), redirect it to
1848 * the socket referenced by *map* (of type
1849 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1850 * egress interfaces can be used for redirection. The
1851 * **BPF_F_INGRESS** value in *flags* is used to make the
1852 * distinction (ingress path is selected if the flag is present,
1853 * egress path otherwise). This is the only flag supported for now.
1855 * **SK_PASS** on success, or **SK_DROP** on error.
1857 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1859 * For socket policies, apply the verdict of the eBPF program to
1860 * the next *bytes* (number of bytes) of message *msg*.
1862 * For example, this helper can be used in the following cases:
1864 * * A single **sendmsg**\ () or **sendfile**\ () system call
1865 * contains multiple logical messages that the eBPF program is
1866 * supposed to read and for which it should apply a verdict.
1867 * * An eBPF program only cares to read the first *bytes* of a
1868 * *msg*. If the message has a large payload, then setting up
1869 * and calling the eBPF program repeatedly for all bytes, even
1870 * though the verdict is already known, would create unnecessary
1873 * When called from within an eBPF program, the helper sets a
1874 * counter internal to the BPF infrastructure, that is used to
1875 * apply the last verdict to the next *bytes*. If *bytes* is
1876 * smaller than the current data being processed from a
1877 * **sendmsg**\ () or **sendfile**\ () system call, the first
1878 * *bytes* will be sent and the eBPF program will be re-run with
1879 * the pointer for start of data pointing to byte number *bytes*
1880 * **+ 1**. If *bytes* is larger than the current data being
1881 * processed, then the eBPF verdict will be applied to multiple
1882 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1885 * Note that if a socket closes with the internal counter holding
1886 * a non-zero value, this is not a problem because data is not
1887 * being buffered for *bytes* and is sent as it is received.
1891 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1893 * For socket policies, prevent the execution of the verdict eBPF
1894 * program for message *msg* until *bytes* (byte number) have been
1897 * This can be used when one needs a specific number of bytes
1898 * before a verdict can be assigned, even if the data spans
1899 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1900 * case would be a user calling **sendmsg**\ () repeatedly with
1901 * 1-byte long message segments. Obviously, this is bad for
1902 * performance, but it is still valid. If the eBPF program needs
1903 * *bytes* bytes to validate a header, this helper can be used to
1904 * prevent the eBPF program to be called again until *bytes* have
1909 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1911 * For socket policies, pull in non-linear data from user space
1912 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1913 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1916 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1917 * *msg* it can only parse data that the (**data**, **data_end**)
1918 * pointers have already consumed. For **sendmsg**\ () hooks this
1919 * is likely the first scatterlist element. But for calls relying
1920 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1921 * be the range (**0**, **0**) because the data is shared with
1922 * user space and by default the objective is to avoid allowing
1923 * user space to modify data while (or after) eBPF verdict is
1924 * being decided. This helper can be used to pull in data and to
1925 * set the start and end pointer to given values. Data will be
1926 * copied if necessary (i.e. if data was not linear and if start
1927 * and end pointers do not point to the same chunk).
1929 * A call to this helper is susceptible to change the underlying
1930 * packet buffer. Therefore, at load time, all checks on pointers
1931 * previously done by the verifier are invalidated and must be
1932 * performed again, if the helper is used in combination with
1933 * direct packet access.
1935 * All values for *flags* are reserved for future usage, and must
1938 * 0 on success, or a negative error in case of failure.
1940 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1942 * Bind the socket associated to *ctx* to the address pointed by
1943 * *addr*, of length *addr_len*. This allows for making outgoing
1944 * connection from the desired IP address, which can be useful for
1945 * example when all processes inside a cgroup should use one
1946 * single IP address on a host that has multiple IP configured.
1948 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1949 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1950 * **AF_INET6**). Looking for a free port to bind to can be
1951 * expensive, therefore binding to port is not permitted by the
1952 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1953 * must be set to zero.
1955 * 0 on success, or a negative error in case of failure.
1957 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1959 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1960 * only possible to shrink the packet as of this writing,
1961 * therefore *delta* must be a negative integer.
1963 * A call to this helper is susceptible to change the underlying
1964 * packet buffer. Therefore, at load time, all checks on pointers
1965 * previously done by the verifier are invalidated and must be
1966 * performed again, if the helper is used in combination with
1967 * direct packet access.
1969 * 0 on success, or a negative error in case of failure.
1971 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1973 * Retrieve the XFRM state (IP transform framework, see also
1974 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1976 * The retrieved value is stored in the **struct bpf_xfrm_state**
1977 * pointed by *xfrm_state* and of length *size*.
1979 * All values for *flags* are reserved for future usage, and must
1982 * This helper is available only if the kernel was compiled with
1983 * **CONFIG_XFRM** configuration option.
1985 * 0 on success, or a negative error in case of failure.
1987 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
1989 * Return a user or a kernel stack in bpf program provided buffer.
1990 * To achieve this, the helper needs *ctx*, which is a pointer
1991 * to the context on which the tracing program is executed.
1992 * To store the stacktrace, the bpf program provides *buf* with
1993 * a nonnegative *size*.
1995 * The last argument, *flags*, holds the number of stack frames to
1996 * skip (from 0 to 255), masked with
1997 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1998 * the following flags:
2000 * **BPF_F_USER_STACK**
2001 * Collect a user space stack instead of a kernel stack.
2002 * **BPF_F_USER_BUILD_ID**
2003 * Collect buildid+offset instead of ips for user stack,
2004 * only valid if **BPF_F_USER_STACK** is also specified.
2006 * **bpf_get_stack**\ () can collect up to
2007 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2008 * to sufficient large buffer size. Note that
2009 * this limit can be controlled with the **sysctl** program, and
2010 * that it should be manually increased in order to profile long
2011 * user stacks (such as stacks for Java programs). To do so, use:
2015 * # sysctl kernel.perf_event_max_stack=<new value>
2017 * A non-negative value equal to or less than *size* on success,
2018 * or a negative error in case of failure.
2020 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2022 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2023 * it provides an easy way to load *len* bytes from *offset*
2024 * from the packet associated to *skb*, into the buffer pointed
2025 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2026 * a fifth argument *start_header* exists in order to select a
2027 * base offset to start from. *start_header* can be one of:
2029 * **BPF_HDR_START_MAC**
2030 * Base offset to load data from is *skb*'s mac header.
2031 * **BPF_HDR_START_NET**
2032 * Base offset to load data from is *skb*'s network header.
2034 * In general, "direct packet access" is the preferred method to
2035 * access packet data, however, this helper is in particular useful
2036 * in socket filters where *skb*\ **->data** does not always point
2037 * to the start of the mac header and where "direct packet access"
2040 * 0 on success, or a negative error in case of failure.
2042 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2044 * Do FIB lookup in kernel tables using parameters in *params*.
2045 * If lookup is successful and result shows packet is to be
2046 * forwarded, the neighbor tables are searched for the nexthop.
2047 * If successful (ie., FIB lookup shows forwarding and nexthop
2048 * is resolved), the nexthop address is returned in ipv4_dst
2049 * or ipv6_dst based on family, smac is set to mac address of
2050 * egress device, dmac is set to nexthop mac address, rt_metric
2051 * is set to metric from route (IPv4/IPv6 only), and ifindex
2052 * is set to the device index of the nexthop from the FIB lookup.
2054 * *plen* argument is the size of the passed in struct.
2055 * *flags* argument can be a combination of one or more of the
2058 * **BPF_FIB_LOOKUP_DIRECT**
2059 * Do a direct table lookup vs full lookup using FIB
2061 * **BPF_FIB_LOOKUP_OUTPUT**
2062 * Perform lookup from an egress perspective (default is
2065 * *ctx* is either **struct xdp_md** for XDP programs or
2066 * **struct sk_buff** tc cls_act programs.
2068 * * < 0 if any input argument is invalid
2069 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2070 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2071 * packet is not forwarded or needs assist from full stack
2073 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2075 * Add an entry to, or update a sockhash *map* referencing sockets.
2076 * The *skops* is used as a new value for the entry associated to
2077 * *key*. *flags* is one of:
2080 * The entry for *key* must not exist in the map.
2082 * The entry for *key* must already exist in the map.
2084 * No condition on the existence of the entry for *key*.
2086 * If the *map* has eBPF programs (parser and verdict), those will
2087 * be inherited by the socket being added. If the socket is
2088 * already attached to eBPF programs, this results in an error.
2090 * 0 on success, or a negative error in case of failure.
2092 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2094 * This helper is used in programs implementing policies at the
2095 * socket level. If the message *msg* is allowed to pass (i.e. if
2096 * the verdict eBPF program returns **SK_PASS**), redirect it to
2097 * the socket referenced by *map* (of type
2098 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2099 * egress interfaces can be used for redirection. The
2100 * **BPF_F_INGRESS** value in *flags* is used to make the
2101 * distinction (ingress path is selected if the flag is present,
2102 * egress path otherwise). This is the only flag supported for now.
2104 * **SK_PASS** on success, or **SK_DROP** on error.
2106 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2108 * This helper is used in programs implementing policies at the
2109 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2110 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2111 * to the socket referenced by *map* (of type
2112 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2113 * egress interfaces can be used for redirection. The
2114 * **BPF_F_INGRESS** value in *flags* is used to make the
2115 * distinction (ingress path is selected if the flag is present,
2116 * egress otherwise). This is the only flag supported for now.
2118 * **SK_PASS** on success, or **SK_DROP** on error.
2120 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2122 * Encapsulate the packet associated to *skb* within a Layer 3
2123 * protocol header. This header is provided in the buffer at
2124 * address *hdr*, with *len* its size in bytes. *type* indicates
2125 * the protocol of the header and can be one of:
2127 * **BPF_LWT_ENCAP_SEG6**
2128 * IPv6 encapsulation with Segment Routing Header
2129 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2130 * the IPv6 header is computed by the kernel.
2131 * **BPF_LWT_ENCAP_SEG6_INLINE**
2132 * Only works if *skb* contains an IPv6 packet. Insert a
2133 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2135 * **BPF_LWT_ENCAP_IP**
2136 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2137 * must be IPv4 or IPv6, followed by zero or more
2138 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2139 * total bytes in all prepended headers. Please note that
2140 * if **skb_is_gso**\ (*skb*) is true, no more than two
2141 * headers can be prepended, and the inner header, if
2142 * present, should be either GRE or UDP/GUE.
2144 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2145 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2146 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2147 * **BPF_PROG_TYPE_LWT_XMIT**.
2149 * A call to this helper is susceptible to change the underlying
2150 * packet buffer. Therefore, at load time, all checks on pointers
2151 * previously done by the verifier are invalidated and must be
2152 * performed again, if the helper is used in combination with
2153 * direct packet access.
2155 * 0 on success, or a negative error in case of failure.
2157 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2159 * Store *len* bytes from address *from* into the packet
2160 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2161 * inside the outermost IPv6 Segment Routing Header can be
2162 * modified through this helper.
2164 * A call to this helper is susceptible to change the underlying
2165 * packet buffer. Therefore, at load time, all checks on pointers
2166 * previously done by the verifier are invalidated and must be
2167 * performed again, if the helper is used in combination with
2168 * direct packet access.
2170 * 0 on success, or a negative error in case of failure.
2172 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2174 * Adjust the size allocated to TLVs in the outermost IPv6
2175 * Segment Routing Header contained in the packet associated to
2176 * *skb*, at position *offset* by *delta* bytes. Only offsets
2177 * after the segments are accepted. *delta* can be as well
2178 * positive (growing) as negative (shrinking).
2180 * A call to this helper is susceptible to change the underlying
2181 * packet buffer. Therefore, at load time, all checks on pointers
2182 * previously done by the verifier are invalidated and must be
2183 * performed again, if the helper is used in combination with
2184 * direct packet access.
2186 * 0 on success, or a negative error in case of failure.
2188 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2190 * Apply an IPv6 Segment Routing action of type *action* to the
2191 * packet associated to *skb*. Each action takes a parameter
2192 * contained at address *param*, and of length *param_len* bytes.
2193 * *action* can be one of:
2195 * **SEG6_LOCAL_ACTION_END_X**
2196 * End.X action: Endpoint with Layer-3 cross-connect.
2197 * Type of *param*: **struct in6_addr**.
2198 * **SEG6_LOCAL_ACTION_END_T**
2199 * End.T action: Endpoint with specific IPv6 table lookup.
2200 * Type of *param*: **int**.
2201 * **SEG6_LOCAL_ACTION_END_B6**
2202 * End.B6 action: Endpoint bound to an SRv6 policy.
2203 * Type of *param*: **struct ipv6_sr_hdr**.
2204 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2205 * End.B6.Encap action: Endpoint bound to an SRv6
2206 * encapsulation policy.
2207 * Type of *param*: **struct ipv6_sr_hdr**.
2209 * A call to this helper is susceptible to change the underlying
2210 * packet buffer. Therefore, at load time, all checks on pointers
2211 * previously done by the verifier are invalidated and must be
2212 * performed again, if the helper is used in combination with
2213 * direct packet access.
2215 * 0 on success, or a negative error in case of failure.
2217 * int bpf_rc_repeat(void *ctx)
2219 * This helper is used in programs implementing IR decoding, to
2220 * report a successfully decoded repeat key message. This delays
2221 * the generation of a key up event for previously generated
2224 * Some IR protocols like NEC have a special IR message for
2225 * repeating last button, for when a button is held down.
2227 * The *ctx* should point to the lirc sample as passed into
2230 * This helper is only available is the kernel was compiled with
2231 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2236 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2238 * This helper is used in programs implementing IR decoding, to
2239 * report a successfully decoded key press with *scancode*,
2240 * *toggle* value in the given *protocol*. The scancode will be
2241 * translated to a keycode using the rc keymap, and reported as
2242 * an input key down event. After a period a key up event is
2243 * generated. This period can be extended by calling either
2244 * **bpf_rc_keydown**\ () again with the same values, or calling
2245 * **bpf_rc_repeat**\ ().
2247 * Some protocols include a toggle bit, in case the button was
2248 * released and pressed again between consecutive scancodes.
2250 * The *ctx* should point to the lirc sample as passed into
2253 * The *protocol* is the decoded protocol number (see
2254 * **enum rc_proto** for some predefined values).
2256 * This helper is only available is the kernel was compiled with
2257 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2262 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2264 * Return the cgroup v2 id of the socket associated with the *skb*.
2265 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2266 * helper for cgroup v1 by providing a tag resp. identifier that
2267 * can be matched on or used for map lookups e.g. to implement
2268 * policy. The cgroup v2 id of a given path in the hierarchy is
2269 * exposed in user space through the f_handle API in order to get
2270 * to the same 64-bit id.
2272 * This helper can be used on TC egress path, but not on ingress,
2273 * and is available only if the kernel was compiled with the
2274 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2276 * The id is returned or 0 in case the id could not be retrieved.
2278 * u64 bpf_get_current_cgroup_id(void)
2280 * A 64-bit integer containing the current cgroup id based
2281 * on the cgroup within which the current task is running.
2283 * void *bpf_get_local_storage(void *map, u64 flags)
2285 * Get the pointer to the local storage area.
2286 * The type and the size of the local storage is defined
2287 * by the *map* argument.
2288 * The *flags* meaning is specific for each map type,
2289 * and has to be 0 for cgroup local storage.
2291 * Depending on the BPF program type, a local storage area
2292 * can be shared between multiple instances of the BPF program,
2293 * running simultaneously.
2295 * A user should care about the synchronization by himself.
2296 * For example, by using the **BPF_STX_XADD** instruction to alter
2299 * A pointer to the local storage area.
2301 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2303 * Select a **SO_REUSEPORT** socket from a
2304 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2305 * It checks the selected socket is matching the incoming
2306 * request in the socket buffer.
2308 * 0 on success, or a negative error in case of failure.
2310 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2312 * Return id of cgroup v2 that is ancestor of cgroup associated
2313 * with the *skb* at the *ancestor_level*. The root cgroup is at
2314 * *ancestor_level* zero and each step down the hierarchy
2315 * increments the level. If *ancestor_level* == level of cgroup
2316 * associated with *skb*, then return value will be same as that
2317 * of **bpf_skb_cgroup_id**\ ().
2319 * The helper is useful to implement policies based on cgroups
2320 * that are upper in hierarchy than immediate cgroup associated
2323 * The format of returned id and helper limitations are same as in
2324 * **bpf_skb_cgroup_id**\ ().
2326 * The id is returned or 0 in case the id could not be retrieved.
2328 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2330 * Look for TCP socket matching *tuple*, optionally in a child
2331 * network namespace *netns*. The return value must be checked,
2332 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2334 * The *ctx* should point to the context of the program, such as
2335 * the skb or socket (depending on the hook in use). This is used
2336 * to determine the base network namespace for the lookup.
2338 * *tuple_size* must be one of:
2340 * **sizeof**\ (*tuple*\ **->ipv4**)
2341 * Look for an IPv4 socket.
2342 * **sizeof**\ (*tuple*\ **->ipv6**)
2343 * Look for an IPv6 socket.
2345 * If the *netns* is a negative signed 32-bit integer, then the
2346 * socket lookup table in the netns associated with the *ctx* will
2347 * will be used. For the TC hooks, this is the netns of the device
2348 * in the skb. For socket hooks, this is the netns of the socket.
2349 * If *netns* is any other signed 32-bit value greater than or
2350 * equal to zero then it specifies the ID of the netns relative to
2351 * the netns associated with the *ctx*. *netns* values beyond the
2352 * range of 32-bit integers are reserved for future use.
2354 * All values for *flags* are reserved for future usage, and must
2357 * This helper is available only if the kernel was compiled with
2358 * **CONFIG_NET** configuration option.
2360 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2361 * For sockets with reuseport option, the **struct bpf_sock**
2362 * result is from *reuse*\ **->socks**\ [] using the hash of the
2365 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2367 * Look for UDP socket matching *tuple*, optionally in a child
2368 * network namespace *netns*. The return value must be checked,
2369 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2371 * The *ctx* should point to the context of the program, such as
2372 * the skb or socket (depending on the hook in use). This is used
2373 * to determine the base network namespace for the lookup.
2375 * *tuple_size* must be one of:
2377 * **sizeof**\ (*tuple*\ **->ipv4**)
2378 * Look for an IPv4 socket.
2379 * **sizeof**\ (*tuple*\ **->ipv6**)
2380 * Look for an IPv6 socket.
2382 * If the *netns* is a negative signed 32-bit integer, then the
2383 * socket lookup table in the netns associated with the *ctx* will
2384 * will be used. For the TC hooks, this is the netns of the device
2385 * in the skb. For socket hooks, this is the netns of the socket.
2386 * If *netns* is any other signed 32-bit value greater than or
2387 * equal to zero then it specifies the ID of the netns relative to
2388 * the netns associated with the *ctx*. *netns* values beyond the
2389 * range of 32-bit integers are reserved for future use.
2391 * All values for *flags* are reserved for future usage, and must
2394 * This helper is available only if the kernel was compiled with
2395 * **CONFIG_NET** configuration option.
2397 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2398 * For sockets with reuseport option, the **struct bpf_sock**
2399 * result is from *reuse*\ **->socks**\ [] using the hash of the
2402 * int bpf_sk_release(struct bpf_sock *sock)
2404 * Release the reference held by *sock*. *sock* must be a
2405 * non-**NULL** pointer that was returned from
2406 * **bpf_sk_lookup_xxx**\ ().
2408 * 0 on success, or a negative error in case of failure.
2410 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2412 * Push an element *value* in *map*. *flags* is one of:
2415 * If the queue/stack is full, the oldest element is
2416 * removed to make room for this.
2418 * 0 on success, or a negative error in case of failure.
2420 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2422 * Pop an element from *map*.
2424 * 0 on success, or a negative error in case of failure.
2426 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2428 * Get an element from *map* without removing it.
2430 * 0 on success, or a negative error in case of failure.
2432 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2434 * For socket policies, insert *len* bytes into *msg* at offset
2437 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2438 * *msg* it may want to insert metadata or options into the *msg*.
2439 * This can later be read and used by any of the lower layer BPF
2442 * This helper may fail if under memory pressure (a malloc
2443 * fails) in these cases BPF programs will get an appropriate
2444 * error and BPF programs will need to handle them.
2446 * 0 on success, or a negative error in case of failure.
2448 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2450 * Will remove *len* bytes from a *msg* starting at byte *start*.
2451 * This may result in **ENOMEM** errors under certain situations if
2452 * an allocation and copy are required due to a full ring buffer.
2453 * However, the helper will try to avoid doing the allocation
2454 * if possible. Other errors can occur if input parameters are
2455 * invalid either due to *start* byte not being valid part of *msg*
2456 * payload and/or *pop* value being to large.
2458 * 0 on success, or a negative error in case of failure.
2460 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2462 * This helper is used in programs implementing IR decoding, to
2463 * report a successfully decoded pointer movement.
2465 * The *ctx* should point to the lirc sample as passed into
2468 * This helper is only available is the kernel was compiled with
2469 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2474 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2476 * Acquire a spinlock represented by the pointer *lock*, which is
2477 * stored as part of a value of a map. Taking the lock allows to
2478 * safely update the rest of the fields in that value. The
2479 * spinlock can (and must) later be released with a call to
2480 * **bpf_spin_unlock**\ (\ *lock*\ ).
2482 * Spinlocks in BPF programs come with a number of restrictions
2485 * * **bpf_spin_lock** objects are only allowed inside maps of
2486 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2487 * list could be extended in the future).
2488 * * BTF description of the map is mandatory.
2489 * * The BPF program can take ONE lock at a time, since taking two
2490 * or more could cause dead locks.
2491 * * Only one **struct bpf_spin_lock** is allowed per map element.
2492 * * When the lock is taken, calls (either BPF to BPF or helpers)
2494 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2495 * allowed inside a spinlock-ed region.
2496 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2497 * the lock, on all execution paths, before it returns.
2498 * * The BPF program can access **struct bpf_spin_lock** only via
2499 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2500 * helpers. Loading or storing data into the **struct
2501 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2502 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2503 * of the map value must be a struct and have **struct
2504 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2505 * Nested lock inside another struct is not allowed.
2506 * * The **struct bpf_spin_lock** *lock* field in a map value must
2507 * be aligned on a multiple of 4 bytes in that value.
2508 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2509 * the **bpf_spin_lock** field to user space.
2510 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2511 * a BPF program, do not update the **bpf_spin_lock** field.
2512 * * **bpf_spin_lock** cannot be on the stack or inside a
2513 * networking packet (it can only be inside of a map values).
2514 * * **bpf_spin_lock** is available to root only.
2515 * * Tracing programs and socket filter programs cannot use
2516 * **bpf_spin_lock**\ () due to insufficient preemption checks
2517 * (but this may change in the future).
2518 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2522 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2524 * Release the *lock* previously locked by a call to
2525 * **bpf_spin_lock**\ (\ *lock*\ ).
2529 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2531 * This helper gets a **struct bpf_sock** pointer such
2532 * that all the fields in this **bpf_sock** can be accessed.
2534 * A **struct bpf_sock** pointer on success, or **NULL** in
2537 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2539 * This helper gets a **struct bpf_tcp_sock** pointer from a
2540 * **struct bpf_sock** pointer.
2542 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2545 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
2547 * Set ECN (Explicit Congestion Notification) field of IP header
2548 * to **CE** (Congestion Encountered) if current value is **ECT**
2549 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2552 * 1 if the **CE** flag is set (either by the current helper call
2553 * or because it was already present), 0 if it is not set.
2555 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2557 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2558 * **bpf_sk_release**\ () is unnecessary and not allowed.
2560 * A **struct bpf_sock** pointer on success, or **NULL** in
2563 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2565 * Look for TCP socket matching *tuple*, optionally in a child
2566 * network namespace *netns*. The return value must be checked,
2567 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2569 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2570 * that it also returns timewait or request sockets. Use
2571 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2574 * This helper is available only if the kernel was compiled with
2575 * **CONFIG_NET** configuration option.
2577 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2578 * For sockets with reuseport option, the **struct bpf_sock**
2579 * result is from *reuse*\ **->socks**\ [] using the hash of the
2582 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2584 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2585 * the listening socket in *sk*.
2587 * *iph* points to the start of the IPv4 or IPv6 header, while
2588 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2589 * **sizeof**\ (**struct ip6hdr**).
2591 * *th* points to the start of the TCP header, while *th_len*
2592 * contains **sizeof**\ (**struct tcphdr**).
2595 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2598 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2600 * Get name of sysctl in /proc/sys/ and copy it into provided by
2601 * program buffer *buf* of size *buf_len*.
2603 * The buffer is always NUL terminated, unless it's zero-sized.
2605 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2606 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2607 * only (e.g. "tcp_mem").
2609 * Number of character copied (not including the trailing NUL).
2611 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2612 * truncated name in this case).
2614 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2616 * Get current value of sysctl as it is presented in /proc/sys
2617 * (incl. newline, etc), and copy it as a string into provided
2618 * by program buffer *buf* of size *buf_len*.
2620 * The whole value is copied, no matter what file position user
2621 * space issued e.g. sys_read at.
2623 * The buffer is always NUL terminated, unless it's zero-sized.
2625 * Number of character copied (not including the trailing NUL).
2627 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2628 * truncated name in this case).
2630 * **-EINVAL** if current value was unavailable, e.g. because
2631 * sysctl is uninitialized and read returns -EIO for it.
2633 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2635 * Get new value being written by user space to sysctl (before
2636 * the actual write happens) and copy it as a string into
2637 * provided by program buffer *buf* of size *buf_len*.
2639 * User space may write new value at file position > 0.
2641 * The buffer is always NUL terminated, unless it's zero-sized.
2643 * Number of character copied (not including the trailing NUL).
2645 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2646 * truncated name in this case).
2648 * **-EINVAL** if sysctl is being read.
2650 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2652 * Override new value being written by user space to sysctl with
2653 * value provided by program in buffer *buf* of size *buf_len*.
2655 * *buf* should contain a string in same form as provided by user
2656 * space on sysctl write.
2658 * User space may write new value at file position > 0. To override
2659 * the whole sysctl value file position should be set to zero.
2663 * **-E2BIG** if the *buf_len* is too big.
2665 * **-EINVAL** if sysctl is being read.
2667 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2669 * Convert the initial part of the string from buffer *buf* of
2670 * size *buf_len* to a long integer according to the given base
2671 * and save the result in *res*.
2673 * The string may begin with an arbitrary amount of white space
2674 * (as determined by **isspace**\ (3)) followed by a single
2675 * optional '**-**' sign.
2677 * Five least significant bits of *flags* encode base, other bits
2678 * are currently unused.
2680 * Base must be either 8, 10, 16 or 0 to detect it automatically
2681 * similar to user space **strtol**\ (3).
2683 * Number of characters consumed on success. Must be positive but
2684 * no more than *buf_len*.
2686 * **-EINVAL** if no valid digits were found or unsupported base
2689 * **-ERANGE** if resulting value was out of range.
2691 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2693 * Convert the initial part of the string from buffer *buf* of
2694 * size *buf_len* to an unsigned long integer according to the
2695 * given base and save the result in *res*.
2697 * The string may begin with an arbitrary amount of white space
2698 * (as determined by **isspace**\ (3)).
2700 * Five least significant bits of *flags* encode base, other bits
2701 * are currently unused.
2703 * Base must be either 8, 10, 16 or 0 to detect it automatically
2704 * similar to user space **strtoul**\ (3).
2706 * Number of characters consumed on success. Must be positive but
2707 * no more than *buf_len*.
2709 * **-EINVAL** if no valid digits were found or unsupported base
2712 * **-ERANGE** if resulting value was out of range.
2714 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2716 * Get a bpf-local-storage from a *sk*.
2718 * Logically, it could be thought of getting the value from
2719 * a *map* with *sk* as the **key**. From this
2720 * perspective, the usage is not much different from
2721 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2722 * helper enforces the key must be a full socket and the map must
2723 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2725 * Underneath, the value is stored locally at *sk* instead of
2726 * the *map*. The *map* is used as the bpf-local-storage
2727 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2728 * searched against all bpf-local-storages residing at *sk*.
2730 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2731 * used such that a new bpf-local-storage will be
2732 * created if one does not exist. *value* can be used
2733 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2734 * the initial value of a bpf-local-storage. If *value* is
2735 * **NULL**, the new bpf-local-storage will be zero initialized.
2737 * A bpf-local-storage pointer is returned on success.
2739 * **NULL** if not found or there was an error in adding
2740 * a new bpf-local-storage.
2742 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2744 * Delete a bpf-local-storage from a *sk*.
2748 * **-ENOENT** if the bpf-local-storage cannot be found.
2750 * int bpf_send_signal(u32 sig)
2752 * Send signal *sig* to the process of the current task.
2753 * The signal may be delivered to any of this process's threads.
2755 * 0 on success or successfully queued.
2757 * **-EBUSY** if work queue under nmi is full.
2759 * **-EINVAL** if *sig* is invalid.
2761 * **-EPERM** if no permission to send the *sig*.
2763 * **-EAGAIN** if bpf program can try again.
2765 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2767 * Try to issue a SYN cookie for the packet with corresponding
2768 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2770 * *iph* points to the start of the IPv4 or IPv6 header, while
2771 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2772 * **sizeof**\ (**struct ip6hdr**).
2774 * *th* points to the start of the TCP header, while *th_len*
2775 * contains the length of the TCP header.
2778 * On success, lower 32 bits hold the generated SYN cookie in
2779 * followed by 16 bits which hold the MSS value for that cookie,
2780 * and the top 16 bits are unused.
2782 * On failure, the returned value is one of the following:
2784 * **-EINVAL** SYN cookie cannot be issued due to error
2786 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2788 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2790 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2792 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2794 * Write raw *data* blob into a special BPF perf event held by
2795 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2796 * event must have the following attributes: **PERF_SAMPLE_RAW**
2797 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2798 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2800 * The *flags* are used to indicate the index in *map* for which
2801 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2802 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2803 * to indicate that the index of the current CPU core should be
2806 * The value to write, of *size*, is passed through eBPF stack and
2807 * pointed by *data*.
2809 * *ctx* is a pointer to in-kernel struct sk_buff.
2811 * This helper is similar to **bpf_perf_event_output**\ () but
2812 * restricted to raw_tracepoint bpf programs.
2814 * 0 on success, or a negative error in case of failure.
2816 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2818 * Safely attempt to read *size* bytes from user space address
2819 * *unsafe_ptr* and store the data in *dst*.
2821 * 0 on success, or a negative error in case of failure.
2823 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2825 * Safely attempt to read *size* bytes from kernel space address
2826 * *unsafe_ptr* and store the data in *dst*.
2828 * 0 on success, or a negative error in case of failure.
2830 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2832 * Copy a NUL terminated string from an unsafe user address
2833 * *unsafe_ptr* to *dst*. The *size* should include the
2834 * terminating NUL byte. In case the string length is smaller than
2835 * *size*, the target is not padded with further NUL bytes. If the
2836 * string length is larger than *size*, just *size*-1 bytes are
2837 * copied and the last byte is set to NUL.
2839 * On success, the length of the copied string is returned. This
2840 * makes this helper useful in tracing programs for reading
2841 * strings, and more importantly to get its length at runtime. See
2842 * the following snippet:
2846 * SEC("kprobe/sys_open")
2847 * void bpf_sys_open(struct pt_regs *ctx)
2849 * char buf[PATHLEN]; // PATHLEN is defined to 256
2850 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2853 * // Consume buf, for example push it to
2854 * // userspace via bpf_perf_event_output(); we
2855 * // can use res (the string length) as event
2856 * // size, after checking its boundaries.
2859 * In comparison, using **bpf_probe_read_user()** helper here
2860 * instead to read the string would require to estimate the length
2861 * at compile time, and would often result in copying more memory
2864 * Another useful use case is when parsing individual process
2865 * arguments or individual environment variables navigating
2866 * *current*\ **->mm->arg_start** and *current*\
2867 * **->mm->env_start**: using this helper and the return value,
2868 * one can quickly iterate at the right offset of the memory area.
2870 * On success, the strictly positive length of the string,
2871 * including the trailing NUL character. On error, a negative
2874 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2876 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
2877 * to *dst*. Same semantics as with bpf_probe_read_user_str() apply.
2879 * On success, the strictly positive length of the string, including
2880 * the trailing NUL character. On error, a negative value.
2882 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2884 * Send out a tcp-ack. *tp* is the in-kernel struct tcp_sock.
2885 * *rcv_nxt* is the ack_seq to be sent out.
2887 * 0 on success, or a negative error in case of failure.
2889 * int bpf_send_signal_thread(u32 sig)
2891 * Send signal *sig* to the thread corresponding to the current task.
2893 * 0 on success or successfully queued.
2895 * **-EBUSY** if work queue under nmi is full.
2897 * **-EINVAL** if *sig* is invalid.
2899 * **-EPERM** if no permission to send the *sig*.
2901 * **-EAGAIN** if bpf program can try again.
2903 * u64 bpf_jiffies64(void)
2905 * Obtain the 64bit jiffies
2907 * The 64 bit jiffies
2909 * int bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
2911 * For an eBPF program attached to a perf event, retrieve the
2912 * branch records (struct perf_branch_entry) associated to *ctx*
2913 * and store it in the buffer pointed by *buf* up to size
2916 * On success, number of bytes written to *buf*. On error, a
2919 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
2920 * instead return the number of bytes required to store all the
2921 * branch entries. If this flag is set, *buf* may be NULL.
2923 * **-EINVAL** if arguments invalid or **size** not a multiple
2924 * of sizeof(struct perf_branch_entry).
2926 * **-ENOENT** if architecture does not support branch records.
2928 * int bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
2930 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
2931 * *namespace* will be returned in *nsdata*.
2933 * On failure, the returned value is one of the following:
2935 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
2936 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
2938 * **-ENOENT** if pidns does not exists for the current task.
2940 * int bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2942 * Write raw *data* blob into a special BPF perf event held by
2943 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2944 * event must have the following attributes: **PERF_SAMPLE_RAW**
2945 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2946 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2948 * The *flags* are used to indicate the index in *map* for which
2949 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2950 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2951 * to indicate that the index of the current CPU core should be
2954 * The value to write, of *size*, is passed through eBPF stack and
2955 * pointed by *data*.
2957 * *ctx* is a pointer to in-kernel struct xdp_buff.
2959 * This helper is similar to **bpf_perf_eventoutput**\ () but
2960 * restricted to raw_tracepoint bpf programs.
2962 * 0 on success, or a negative error in case of failure.
2964 * u64 bpf_get_netns_cookie(void *ctx)
2966 * Retrieve the cookie (generated by the kernel) of the network
2967 * namespace the input *ctx* is associated with. The network
2968 * namespace cookie remains stable for its lifetime and provides
2969 * a global identifier that can be assumed unique. If *ctx* is
2970 * NULL, then the helper returns the cookie for the initial
2971 * network namespace. The cookie itself is very similar to that
2972 * of bpf_get_socket_cookie() helper, but for network namespaces
2973 * instead of sockets.
2975 * A 8-byte long opaque number.
2977 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
2979 * Return id of cgroup v2 that is ancestor of the cgroup associated
2980 * with the current task at the *ancestor_level*. The root cgroup
2981 * is at *ancestor_level* zero and each step down the hierarchy
2982 * increments the level. If *ancestor_level* == level of cgroup
2983 * associated with the current task, then return value will be the
2984 * same as that of **bpf_get_current_cgroup_id**\ ().
2986 * The helper is useful to implement policies based on cgroups
2987 * that are upper in hierarchy than immediate cgroup associated
2988 * with the current task.
2990 * The format of returned id and helper limitations are same as in
2991 * **bpf_get_current_cgroup_id**\ ().
2993 * The id is returned or 0 in case the id could not be retrieved.
2995 * int bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags)
2997 * Assign the *sk* to the *skb*. When combined with appropriate
2998 * routing configuration to receive the packet towards the socket,
2999 * will cause *skb* to be delivered to the specified socket.
3000 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3001 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3002 * interfere with successful delivery to the socket.
3004 * This operation is only valid from TC ingress path.
3006 * The *flags* argument must be zero.
3008 * 0 on success, or a negative errno in case of failure.
3010 * * **-EINVAL** Unsupported flags specified.
3011 * * **-ENOENT** Socket is unavailable for assignment.
3012 * * **-ENETUNREACH** Socket is unreachable (wrong netns).
3013 * * **-EOPNOTSUPP** Unsupported operation, for example a
3014 * call from outside of TC ingress.
3015 * * **-ESOCKTNOSUPPORT** Socket type not supported (reuseport).
3017 #define __BPF_FUNC_MAPPER(FN) \
3019 FN(map_lookup_elem), \
3020 FN(map_update_elem), \
3021 FN(map_delete_elem), \
3025 FN(get_prandom_u32), \
3026 FN(get_smp_processor_id), \
3027 FN(skb_store_bytes), \
3028 FN(l3_csum_replace), \
3029 FN(l4_csum_replace), \
3031 FN(clone_redirect), \
3032 FN(get_current_pid_tgid), \
3033 FN(get_current_uid_gid), \
3034 FN(get_current_comm), \
3035 FN(get_cgroup_classid), \
3036 FN(skb_vlan_push), \
3038 FN(skb_get_tunnel_key), \
3039 FN(skb_set_tunnel_key), \
3040 FN(perf_event_read), \
3042 FN(get_route_realm), \
3043 FN(perf_event_output), \
3044 FN(skb_load_bytes), \
3047 FN(skb_get_tunnel_opt), \
3048 FN(skb_set_tunnel_opt), \
3049 FN(skb_change_proto), \
3050 FN(skb_change_type), \
3051 FN(skb_under_cgroup), \
3052 FN(get_hash_recalc), \
3053 FN(get_current_task), \
3054 FN(probe_write_user), \
3055 FN(current_task_under_cgroup), \
3056 FN(skb_change_tail), \
3057 FN(skb_pull_data), \
3059 FN(set_hash_invalid), \
3060 FN(get_numa_node_id), \
3061 FN(skb_change_head), \
3062 FN(xdp_adjust_head), \
3063 FN(probe_read_str), \
3064 FN(get_socket_cookie), \
3065 FN(get_socket_uid), \
3068 FN(skb_adjust_room), \
3070 FN(sk_redirect_map), \
3071 FN(sock_map_update), \
3072 FN(xdp_adjust_meta), \
3073 FN(perf_event_read_value), \
3074 FN(perf_prog_read_value), \
3076 FN(override_return), \
3077 FN(sock_ops_cb_flags_set), \
3078 FN(msg_redirect_map), \
3079 FN(msg_apply_bytes), \
3080 FN(msg_cork_bytes), \
3081 FN(msg_pull_data), \
3083 FN(xdp_adjust_tail), \
3084 FN(skb_get_xfrm_state), \
3086 FN(skb_load_bytes_relative), \
3088 FN(sock_hash_update), \
3089 FN(msg_redirect_hash), \
3090 FN(sk_redirect_hash), \
3091 FN(lwt_push_encap), \
3092 FN(lwt_seg6_store_bytes), \
3093 FN(lwt_seg6_adjust_srh), \
3094 FN(lwt_seg6_action), \
3097 FN(skb_cgroup_id), \
3098 FN(get_current_cgroup_id), \
3099 FN(get_local_storage), \
3100 FN(sk_select_reuseport), \
3101 FN(skb_ancestor_cgroup_id), \
3102 FN(sk_lookup_tcp), \
3103 FN(sk_lookup_udp), \
3105 FN(map_push_elem), \
3107 FN(map_peek_elem), \
3108 FN(msg_push_data), \
3110 FN(rc_pointer_rel), \
3115 FN(skb_ecn_set_ce), \
3116 FN(get_listener_sock), \
3117 FN(skc_lookup_tcp), \
3118 FN(tcp_check_syncookie), \
3119 FN(sysctl_get_name), \
3120 FN(sysctl_get_current_value), \
3121 FN(sysctl_get_new_value), \
3122 FN(sysctl_set_new_value), \
3125 FN(sk_storage_get), \
3126 FN(sk_storage_delete), \
3128 FN(tcp_gen_syncookie), \
3130 FN(probe_read_user), \
3131 FN(probe_read_kernel), \
3132 FN(probe_read_user_str), \
3133 FN(probe_read_kernel_str), \
3135 FN(send_signal_thread), \
3137 FN(read_branch_records), \
3138 FN(get_ns_current_pid_tgid), \
3140 FN(get_netns_cookie), \
3141 FN(get_current_ancestor_cgroup_id), \
3144 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
3145 * function eBPF program intends to call
3147 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
3149 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
3152 #undef __BPF_ENUM_FN
3154 /* All flags used by eBPF helper functions, placed here. */
3156 /* BPF_FUNC_skb_store_bytes flags. */
3158 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3159 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3162 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3163 * First 4 bits are for passing the header field size.
3166 BPF_F_HDR_FIELD_MASK = 0xfULL,
3169 /* BPF_FUNC_l4_csum_replace flags. */
3171 BPF_F_PSEUDO_HDR = (1ULL << 4),
3172 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3173 BPF_F_MARK_ENFORCE = (1ULL << 6),
3176 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
3178 BPF_F_INGRESS = (1ULL << 0),
3181 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
3183 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3186 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
3188 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3189 BPF_F_USER_STACK = (1ULL << 8),
3190 /* flags used by BPF_FUNC_get_stackid only. */
3191 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3192 BPF_F_REUSE_STACKID = (1ULL << 10),
3193 /* flags used by BPF_FUNC_get_stack only. */
3194 BPF_F_USER_BUILD_ID = (1ULL << 11),
3197 /* BPF_FUNC_skb_set_tunnel_key flags. */
3199 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3200 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3201 BPF_F_SEQ_NUMBER = (1ULL << 3),
3204 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3205 * BPF_FUNC_perf_event_read_value flags.
3208 BPF_F_INDEX_MASK = 0xffffffffULL,
3209 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
3210 /* BPF_FUNC_perf_event_output for sk_buff input context. */
3211 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3214 /* Current network namespace */
3216 BPF_F_CURRENT_NETNS = (-1L),
3219 /* BPF_FUNC_skb_adjust_room flags. */
3221 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3222 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3223 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3224 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3225 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
3229 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3230 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3233 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
3234 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3235 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
3237 /* BPF_FUNC_sysctl_get_name flags. */
3239 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3242 /* BPF_FUNC_sk_storage_get flags */
3244 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0),
3247 /* BPF_FUNC_read_branch_records flags. */
3249 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3252 /* Mode for BPF_FUNC_skb_adjust_room helper. */
3253 enum bpf_adj_room_mode {
3258 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3259 enum bpf_hdr_start_off {
3264 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3265 enum bpf_lwt_encap_mode {
3267 BPF_LWT_ENCAP_SEG6_INLINE,
3271 #define __bpf_md_ptr(type, name) \
3275 } __attribute__((aligned(8)))
3277 /* user accessible mirror of in-kernel sk_buff.
3278 * new fields can only be added to the end of this structure
3284 __u32 queue_mapping;
3290 __u32 ingress_ifindex;
3300 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
3302 __u32 remote_ip4; /* Stored in network byte order */
3303 __u32 local_ip4; /* Stored in network byte order */
3304 __u32 remote_ip6[4]; /* Stored in network byte order */
3305 __u32 local_ip6[4]; /* Stored in network byte order */
3306 __u32 remote_port; /* Stored in network byte order */
3307 __u32 local_port; /* stored in host byte order */
3311 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
3315 __bpf_md_ptr(struct bpf_sock *, sk);
3319 struct bpf_tunnel_key {
3323 __u32 remote_ipv6[4];
3327 __u16 tunnel_ext; /* Padding, future use. */
3331 /* user accessible mirror of in-kernel xfrm_state.
3332 * new fields can only be added to the end of this structure
3334 struct bpf_xfrm_state {
3336 __u32 spi; /* Stored in network byte order */
3338 __u16 ext; /* Padding, future use. */
3340 __u32 remote_ipv4; /* Stored in network byte order */
3341 __u32 remote_ipv6[4]; /* Stored in network byte order */
3345 /* Generic BPF return codes which all BPF program types may support.
3346 * The values are binary compatible with their TC_ACT_* counter-part to
3347 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3350 * XDP is handled seprately, see XDP_*.
3358 /* >127 are reserved for prog type specific return codes.
3360 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3361 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3362 * changed and should be routed based on its new L3 header.
3363 * (This is an L3 redirect, as opposed to L2 redirect
3364 * represented by BPF_REDIRECT above).
3366 BPF_LWT_REROUTE = 128,
3376 /* IP address also allows 1 and 2 bytes access */
3379 __u32 src_port; /* host byte order */
3380 __u32 dst_port; /* network byte order */
3386 struct bpf_tcp_sock {
3387 __u32 snd_cwnd; /* Sending congestion window */
3388 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3390 __u32 snd_ssthresh; /* Slow start size threshold */
3391 __u32 rcv_nxt; /* What we want to receive next */
3392 __u32 snd_nxt; /* Next sequence we send */
3393 __u32 snd_una; /* First byte we want an ack for */
3394 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3395 __u32 ecn_flags; /* ECN status bits. */
3396 __u32 rate_delivered; /* saved rate sample: packets delivered */
3397 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3398 __u32 packets_out; /* Packets which are "in flight" */
3399 __u32 retrans_out; /* Retransmitted packets out */
3400 __u32 total_retrans; /* Total retransmits for entire connection */
3401 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3402 * total number of segments in.
3404 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3405 * total number of data segments in.
3407 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3408 * The total number of segments sent.
3410 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3411 * total number of data segments sent.
3413 __u32 lost_out; /* Lost packets */
3414 __u32 sacked_out; /* SACK'd packets */
3415 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3416 * sum(delta(rcv_nxt)), or how many bytes
3419 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3420 * sum(delta(snd_una)), or how many bytes
3423 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3424 * total number of DSACK blocks received
3426 __u32 delivered; /* Total data packets delivered incl. rexmits */
3427 __u32 delivered_ce; /* Like the above but only ECE marked packets */
3428 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
3431 struct bpf_sock_tuple {
3448 struct bpf_xdp_sock {
3452 #define XDP_PACKET_HEADROOM 256
3454 /* User return codes for XDP prog type.
3455 * A valid XDP program must return one of these defined values. All other
3456 * return codes are reserved for future use. Unknown return codes will
3457 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3467 /* user accessible metadata for XDP packet hook
3468 * new fields must be added to the end of this structure
3474 /* Below access go through struct xdp_rxq_info */
3475 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3476 __u32 rx_queue_index; /* rxq->queue_index */
3484 /* user accessible metadata for SK_MSG packet hook, new fields must
3485 * be added to the end of this structure
3488 __bpf_md_ptr(void *, data);
3489 __bpf_md_ptr(void *, data_end);
3492 __u32 remote_ip4; /* Stored in network byte order */
3493 __u32 local_ip4; /* Stored in network byte order */
3494 __u32 remote_ip6[4]; /* Stored in network byte order */
3495 __u32 local_ip6[4]; /* Stored in network byte order */
3496 __u32 remote_port; /* Stored in network byte order */
3497 __u32 local_port; /* stored in host byte order */
3498 __u32 size; /* Total size of sk_msg */
3501 struct sk_reuseport_md {
3503 * Start of directly accessible data. It begins from
3504 * the tcp/udp header.
3506 __bpf_md_ptr(void *, data);
3507 /* End of directly accessible data */
3508 __bpf_md_ptr(void *, data_end);
3510 * Total length of packet (starting from the tcp/udp header).
3511 * Note that the directly accessible bytes (data_end - data)
3512 * could be less than this "len". Those bytes could be
3513 * indirectly read by a helper "bpf_skb_load_bytes()".
3517 * Eth protocol in the mac header (network byte order). e.g.
3518 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3521 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3522 __u32 bind_inany; /* Is sock bound to an INANY address? */
3523 __u32 hash; /* A hash of the packet 4 tuples */
3526 #define BPF_TAG_SIZE 8
3528 struct bpf_prog_info {
3531 __u8 tag[BPF_TAG_SIZE];
3532 __u32 jited_prog_len;
3533 __u32 xlated_prog_len;
3534 __aligned_u64 jited_prog_insns;
3535 __aligned_u64 xlated_prog_insns;
3536 __u64 load_time; /* ns since boottime */
3537 __u32 created_by_uid;
3539 __aligned_u64 map_ids;
3540 char name[BPF_OBJ_NAME_LEN];
3542 __u32 gpl_compatible:1;
3543 __u32 :31; /* alignment pad */
3546 __u32 nr_jited_ksyms;
3547 __u32 nr_jited_func_lens;
3548 __aligned_u64 jited_ksyms;
3549 __aligned_u64 jited_func_lens;
3551 __u32 func_info_rec_size;
3552 __aligned_u64 func_info;
3555 __aligned_u64 line_info;
3556 __aligned_u64 jited_line_info;
3557 __u32 nr_jited_line_info;
3558 __u32 line_info_rec_size;
3559 __u32 jited_line_info_rec_size;
3561 __aligned_u64 prog_tags;
3564 } __attribute__((aligned(8)));
3566 struct bpf_map_info {
3573 char name[BPF_OBJ_NAME_LEN];
3575 __u32 btf_vmlinux_value_type_id;
3579 __u32 btf_key_type_id;
3580 __u32 btf_value_type_id;
3581 } __attribute__((aligned(8)));
3583 struct bpf_btf_info {
3587 } __attribute__((aligned(8)));
3589 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3590 * by user and intended to be used by socket (e.g. to bind to, depends on
3591 * attach attach type).
3593 struct bpf_sock_addr {
3594 __u32 user_family; /* Allows 4-byte read, but no write. */
3595 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3596 * Stored in network byte order.
3598 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3599 * Stored in network byte order.
3601 __u32 user_port; /* Allows 4-byte read and write.
3602 * Stored in network byte order
3604 __u32 family; /* Allows 4-byte read, but no write */
3605 __u32 type; /* Allows 4-byte read, but no write */
3606 __u32 protocol; /* Allows 4-byte read, but no write */
3607 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3608 * Stored in network byte order.
3610 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3611 * Stored in network byte order.
3613 __bpf_md_ptr(struct bpf_sock *, sk);
3616 /* User bpf_sock_ops struct to access socket values and specify request ops
3617 * and their replies.
3618 * Some of this fields are in network (bigendian) byte order and may need
3619 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3620 * New fields can only be added at the end of this structure
3622 struct bpf_sock_ops {
3625 __u32 args[4]; /* Optionally passed to bpf program */
3626 __u32 reply; /* Returned by bpf program */
3627 __u32 replylong[4]; /* Optionally returned by bpf prog */
3630 __u32 remote_ip4; /* Stored in network byte order */
3631 __u32 local_ip4; /* Stored in network byte order */
3632 __u32 remote_ip6[4]; /* Stored in network byte order */
3633 __u32 local_ip6[4]; /* Stored in network byte order */
3634 __u32 remote_port; /* Stored in network byte order */
3635 __u32 local_port; /* stored in host byte order */
3636 __u32 is_fullsock; /* Some TCP fields are only valid if
3637 * there is a full socket. If not, the
3638 * fields read as zero.
3641 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
3642 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3651 __u32 rate_delivered;
3652 __u32 rate_interval_us;
3655 __u32 total_retrans;
3659 __u32 data_segs_out;
3663 __u64 bytes_received;
3665 __bpf_md_ptr(struct bpf_sock *, sk);
3668 /* Definitions for bpf_sock_ops_cb_flags */
3670 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
3671 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
3672 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
3673 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
3674 /* Mask of all currently supported cb flags */
3675 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF,
3678 /* List of known BPF sock_ops operators.
3679 * New entries can only be added at the end
3683 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3684 * -1 if default value should be used
3686 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3687 * window (in packets) or -1 if default
3688 * value should be used
3690 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3691 * active connection is initialized
3693 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3694 * active connection is
3697 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3698 * passive connection is
3701 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3704 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3705 * based on the path and may be
3706 * dependent on the congestion control
3707 * algorithm. In general it indicates
3708 * a congestion threshold. RTTs above
3709 * this indicate congestion
3711 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3712 * Arg1: value of icsk_retransmits
3713 * Arg2: value of icsk_rto
3714 * Arg3: whether RTO has expired
3716 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3717 * Arg1: sequence number of 1st byte
3719 * Arg3: return value of
3720 * tcp_transmit_skb (0 => success)
3722 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3726 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3727 * socket transition to LISTEN state.
3729 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3733 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3734 * changes between the TCP and BPF versions. Ideally this should never happen.
3735 * If it does, we need to add code to convert them before calling
3736 * the BPF sock_ops function.
3739 BPF_TCP_ESTABLISHED = 1,
3749 BPF_TCP_CLOSING, /* Now a valid state */
3750 BPF_TCP_NEW_SYN_RECV,
3752 BPF_TCP_MAX_STATES /* Leave at the end! */
3756 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
3757 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
3760 struct bpf_perf_event_value {
3767 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
3768 BPF_DEVCG_ACC_READ = (1ULL << 1),
3769 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
3773 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
3774 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
3777 struct bpf_cgroup_dev_ctx {
3778 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3784 struct bpf_raw_tracepoint_args {
3788 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3789 * OUTPUT: Do lookup from egress perspective; default is ingress
3792 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
3793 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
3797 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3798 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3799 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3800 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3801 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3802 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3803 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3804 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3805 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3808 struct bpf_fib_lookup {
3809 /* input: network family for lookup (AF_INET, AF_INET6)
3810 * output: network family of egress nexthop
3814 /* set if lookup is to consider L4 data - e.g., FIB rules */
3819 /* total length of packet from network header - used for MTU check */
3822 /* input: L3 device index for lookup
3823 * output: device index from FIB lookup
3828 /* inputs to lookup */
3829 __u8 tos; /* AF_INET */
3830 __be32 flowinfo; /* AF_INET6, flow_label + priority */
3832 /* output: metric of fib result (IPv4/IPv6 only) */
3838 __u32 ipv6_src[4]; /* in6_addr; network order */
3841 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3842 * network header. output: bpf_fib_lookup sets to gateway address
3843 * if FIB lookup returns gateway route
3847 __u32 ipv6_dst[4]; /* in6_addr; network order */
3851 __be16 h_vlan_proto;
3853 __u8 smac[6]; /* ETH_ALEN */
3854 __u8 dmac[6]; /* ETH_ALEN */
3857 enum bpf_task_fd_type {
3858 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3859 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3860 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3861 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3862 BPF_FD_TYPE_UPROBE, /* filename + offset */
3863 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3867 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
3868 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
3869 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
3872 struct bpf_flow_keys {
3875 __u16 addr_proto; /* ETH_P_* of valid addrs */
3889 __u32 ipv6_src[4]; /* in6_addr; network order */
3890 __u32 ipv6_dst[4]; /* in6_addr; network order */
3897 struct bpf_func_info {
3902 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3903 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3905 struct bpf_line_info {
3907 __u32 file_name_off;
3912 struct bpf_spin_lock {
3917 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3918 * Allows 1,2,4-byte read, but no write.
3920 __u32 file_pos; /* Sysctl file position to read from, write to.
3921 * Allows 1,2,4-byte read an 4-byte write.
3925 struct bpf_sockopt {
3926 __bpf_md_ptr(struct bpf_sock *, sk);
3927 __bpf_md_ptr(void *, optval);
3928 __bpf_md_ptr(void *, optval_end);
3936 struct bpf_pidns_info {
3940 #endif /* _UAPI__LINUX_BPF_H__ */