1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_XADD 0xc0 /* exclusive add */
25 #define BPF_MOV 0xb0 /* mov reg to reg */
26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28 /* change endianness of a register */
29 #define BPF_END 0xd0 /* flags for endianness conversion: */
30 #define BPF_TO_LE 0x00 /* convert to little-endian */
31 #define BPF_TO_BE 0x08 /* convert to big-endian */
32 #define BPF_FROM_LE BPF_TO_LE
33 #define BPF_FROM_BE BPF_TO_BE
36 #define BPF_JNE 0x50 /* jump != */
37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
43 #define BPF_CALL 0x80 /* function call */
44 #define BPF_EXIT 0x90 /* function return */
46 /* Register numbers */
62 /* BPF has 10 general purpose 64-bit registers and stack frame. */
63 #define MAX_BPF_REG __MAX_BPF_REG
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74 struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data[0]; /* Arbitrary size */
79 struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
84 /* BPF syscall commands, see bpf(2) man-page for details. */
99 BPF_PROG_GET_FD_BY_ID,
100 BPF_MAP_GET_FD_BY_ID,
101 BPF_OBJ_GET_INFO_BY_FD,
103 BPF_RAW_TRACEPOINT_OPEN,
105 BPF_BTF_GET_FD_BY_ID,
107 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
116 BPF_MAP_TYPE_PROG_ARRAY,
117 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
118 BPF_MAP_TYPE_PERCPU_HASH,
119 BPF_MAP_TYPE_PERCPU_ARRAY,
120 BPF_MAP_TYPE_STACK_TRACE,
121 BPF_MAP_TYPE_CGROUP_ARRAY,
122 BPF_MAP_TYPE_LRU_HASH,
123 BPF_MAP_TYPE_LRU_PERCPU_HASH,
124 BPF_MAP_TYPE_LPM_TRIE,
125 BPF_MAP_TYPE_ARRAY_OF_MAPS,
126 BPF_MAP_TYPE_HASH_OF_MAPS,
128 BPF_MAP_TYPE_SOCKMAP,
131 BPF_MAP_TYPE_SOCKHASH,
132 BPF_MAP_TYPE_CGROUP_STORAGE,
133 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
134 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
137 BPF_MAP_TYPE_SK_STORAGE,
138 BPF_MAP_TYPE_DEVMAP_HASH,
139 BPF_MAP_TYPE_STRUCT_OPS,
142 /* Note that tracing related programs such as
143 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
144 * are not subject to a stable API since kernel internal data
145 * structures can change from release to release and may
146 * therefore break existing tracing BPF programs. Tracing BPF
147 * programs correspond to /a/ specific kernel which is to be
148 * analyzed, and not /a/ specific kernel /and/ all future ones.
151 BPF_PROG_TYPE_UNSPEC,
152 BPF_PROG_TYPE_SOCKET_FILTER,
153 BPF_PROG_TYPE_KPROBE,
154 BPF_PROG_TYPE_SCHED_CLS,
155 BPF_PROG_TYPE_SCHED_ACT,
156 BPF_PROG_TYPE_TRACEPOINT,
158 BPF_PROG_TYPE_PERF_EVENT,
159 BPF_PROG_TYPE_CGROUP_SKB,
160 BPF_PROG_TYPE_CGROUP_SOCK,
161 BPF_PROG_TYPE_LWT_IN,
162 BPF_PROG_TYPE_LWT_OUT,
163 BPF_PROG_TYPE_LWT_XMIT,
164 BPF_PROG_TYPE_SOCK_OPS,
165 BPF_PROG_TYPE_SK_SKB,
166 BPF_PROG_TYPE_CGROUP_DEVICE,
167 BPF_PROG_TYPE_SK_MSG,
168 BPF_PROG_TYPE_RAW_TRACEPOINT,
169 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
170 BPF_PROG_TYPE_LWT_SEG6LOCAL,
171 BPF_PROG_TYPE_LIRC_MODE2,
172 BPF_PROG_TYPE_SK_REUSEPORT,
173 BPF_PROG_TYPE_FLOW_DISSECTOR,
174 BPF_PROG_TYPE_CGROUP_SYSCTL,
175 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
176 BPF_PROG_TYPE_CGROUP_SOCKOPT,
177 BPF_PROG_TYPE_TRACING,
178 BPF_PROG_TYPE_STRUCT_OPS,
181 enum bpf_attach_type {
182 BPF_CGROUP_INET_INGRESS,
183 BPF_CGROUP_INET_EGRESS,
184 BPF_CGROUP_INET_SOCK_CREATE,
186 BPF_SK_SKB_STREAM_PARSER,
187 BPF_SK_SKB_STREAM_VERDICT,
190 BPF_CGROUP_INET4_BIND,
191 BPF_CGROUP_INET6_BIND,
192 BPF_CGROUP_INET4_CONNECT,
193 BPF_CGROUP_INET6_CONNECT,
194 BPF_CGROUP_INET4_POST_BIND,
195 BPF_CGROUP_INET6_POST_BIND,
196 BPF_CGROUP_UDP4_SENDMSG,
197 BPF_CGROUP_UDP6_SENDMSG,
201 BPF_CGROUP_UDP4_RECVMSG,
202 BPF_CGROUP_UDP6_RECVMSG,
203 BPF_CGROUP_GETSOCKOPT,
204 BPF_CGROUP_SETSOCKOPT,
208 __MAX_BPF_ATTACH_TYPE
211 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
213 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
215 * NONE(default): No further bpf programs allowed in the subtree.
217 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
218 * the program in this cgroup yields to sub-cgroup program.
220 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
221 * that cgroup program gets run in addition to the program in this cgroup.
223 * Only one program is allowed to be attached to a cgroup with
224 * NONE or BPF_F_ALLOW_OVERRIDE flag.
225 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
226 * release old program and attach the new one. Attach flags has to match.
228 * Multiple programs are allowed to be attached to a cgroup with
229 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
230 * (those that were attached first, run first)
231 * The programs of sub-cgroup are executed first, then programs of
232 * this cgroup and then programs of parent cgroup.
233 * When children program makes decision (like picking TCP CA or sock bind)
234 * parent program has a chance to override it.
236 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
237 * programs for a cgroup. Though it's possible to replace an old program at
238 * any position by also specifying BPF_F_REPLACE flag and position itself in
239 * replace_bpf_fd attribute. Old program at this position will be released.
241 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
242 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
244 * cgrp1 (MULTI progs A, B) ->
245 * cgrp2 (OVERRIDE prog C) ->
246 * cgrp3 (MULTI prog D) ->
247 * cgrp4 (OVERRIDE prog E) ->
248 * cgrp5 (NONE prog F)
249 * the event in cgrp5 triggers execution of F,D,A,B in that order.
250 * if prog F is detached, the execution is E,D,A,B
251 * if prog F and D are detached, the execution is E,A,B
252 * if prog F, E and D are detached, the execution is C,A,B
254 * All eligible programs are executed regardless of return code from
257 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
258 #define BPF_F_ALLOW_MULTI (1U << 1)
259 #define BPF_F_REPLACE (1U << 2)
261 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
262 * verifier will perform strict alignment checking as if the kernel
263 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
264 * and NET_IP_ALIGN defined to 2.
266 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
268 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
269 * verifier will allow any alignment whatsoever. On platforms
270 * with strict alignment requirements for loads ands stores (such
271 * as sparc and mips) the verifier validates that all loads and
272 * stores provably follow this requirement. This flag turns that
273 * checking and enforcement off.
275 * It is mostly used for testing when we want to validate the
276 * context and memory access aspects of the verifier, but because
277 * of an unaligned access the alignment check would trigger before
278 * the one we are interested in.
280 #define BPF_F_ANY_ALIGNMENT (1U << 1)
282 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
283 * Verifier does sub-register def/use analysis and identifies instructions whose
284 * def only matters for low 32-bit, high 32-bit is never referenced later
285 * through implicit zero extension. Therefore verifier notifies JIT back-ends
286 * that it is safe to ignore clearing high 32-bit for these instructions. This
287 * saves some back-ends a lot of code-gen. However such optimization is not
288 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
289 * hence hasn't used verifier's analysis result. But, we really want to have a
290 * way to be able to verify the correctness of the described optimization on
291 * x86_64 on which testsuites are frequently exercised.
293 * So, this flag is introduced. Once it is set, verifier will randomize high
294 * 32-bit for those instructions who has been identified as safe to ignore them.
295 * Then, if verifier is not doing correct analysis, such randomization will
296 * regress tests to expose bugs.
298 #define BPF_F_TEST_RND_HI32 (1U << 2)
300 /* The verifier internal test flag. Behavior is undefined */
301 #define BPF_F_TEST_STATE_FREQ (1U << 3)
303 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
306 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
307 * insn[0].imm: map fd map fd
308 * insn[1].imm: 0 offset into value
311 * ldimm64 rewrite: address of map address of map[0]+offset
312 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
314 #define BPF_PSEUDO_MAP_FD 1
315 #define BPF_PSEUDO_MAP_VALUE 2
317 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
318 * offset to another bpf function
320 #define BPF_PSEUDO_CALL 1
322 /* flags for BPF_MAP_UPDATE_ELEM command */
323 #define BPF_ANY 0 /* create new element or update existing */
324 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
325 #define BPF_EXIST 2 /* update existing element */
326 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
328 /* flags for BPF_MAP_CREATE command */
329 #define BPF_F_NO_PREALLOC (1U << 0)
330 /* Instead of having one common LRU list in the
331 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
332 * which can scale and perform better.
333 * Note, the LRU nodes (including free nodes) cannot be moved
334 * across different LRU lists.
336 #define BPF_F_NO_COMMON_LRU (1U << 1)
337 /* Specify numa node during map creation */
338 #define BPF_F_NUMA_NODE (1U << 2)
340 #define BPF_OBJ_NAME_LEN 16U
342 /* Flags for accessing BPF object from syscall side. */
343 #define BPF_F_RDONLY (1U << 3)
344 #define BPF_F_WRONLY (1U << 4)
346 /* Flag for stack_map, store build_id+offset instead of pointer */
347 #define BPF_F_STACK_BUILD_ID (1U << 5)
349 /* Zero-initialize hash function seed. This should only be used for testing. */
350 #define BPF_F_ZERO_SEED (1U << 6)
352 /* Flags for accessing BPF object from program side. */
353 #define BPF_F_RDONLY_PROG (1U << 7)
354 #define BPF_F_WRONLY_PROG (1U << 8)
356 /* Clone map from listener for newly accepted socket */
357 #define BPF_F_CLONE (1U << 9)
359 /* Enable memory-mapping BPF map */
360 #define BPF_F_MMAPABLE (1U << 10)
362 /* Flags for BPF_PROG_QUERY. */
364 /* Query effective (directly attached + inherited from ancestor cgroups)
365 * programs that will be executed for events within a cgroup.
366 * attach_flags with this flag are returned only for directly attached programs.
368 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
370 enum bpf_stack_build_id_status {
371 /* user space need an empty entry to identify end of a trace */
372 BPF_STACK_BUILD_ID_EMPTY = 0,
373 /* with valid build_id and offset */
374 BPF_STACK_BUILD_ID_VALID = 1,
375 /* couldn't get build_id, fallback to ip */
376 BPF_STACK_BUILD_ID_IP = 2,
379 #define BPF_BUILD_ID_SIZE 20
380 struct bpf_stack_build_id {
382 unsigned char build_id[BPF_BUILD_ID_SIZE];
390 struct { /* anonymous struct used by BPF_MAP_CREATE command */
391 __u32 map_type; /* one of enum bpf_map_type */
392 __u32 key_size; /* size of key in bytes */
393 __u32 value_size; /* size of value in bytes */
394 __u32 max_entries; /* max number of entries in a map */
395 __u32 map_flags; /* BPF_MAP_CREATE related
396 * flags defined above.
398 __u32 inner_map_fd; /* fd pointing to the inner map */
399 __u32 numa_node; /* numa node (effective only if
400 * BPF_F_NUMA_NODE is set).
402 char map_name[BPF_OBJ_NAME_LEN];
403 __u32 map_ifindex; /* ifindex of netdev to create on */
404 __u32 btf_fd; /* fd pointing to a BTF type data */
405 __u32 btf_key_type_id; /* BTF type_id of the key */
406 __u32 btf_value_type_id; /* BTF type_id of the value */
407 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
408 * struct stored as the
413 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
418 __aligned_u64 next_key;
423 struct { /* anonymous struct used by BPF_PROG_LOAD command */
424 __u32 prog_type; /* one of enum bpf_prog_type */
427 __aligned_u64 license;
428 __u32 log_level; /* verbosity level of verifier */
429 __u32 log_size; /* size of user buffer */
430 __aligned_u64 log_buf; /* user supplied buffer */
431 __u32 kern_version; /* not used */
433 char prog_name[BPF_OBJ_NAME_LEN];
434 __u32 prog_ifindex; /* ifindex of netdev to prep for */
435 /* For some prog types expected attach type must be known at
436 * load time to verify attach type specific parts of prog
437 * (context accesses, allowed helpers, etc).
439 __u32 expected_attach_type;
440 __u32 prog_btf_fd; /* fd pointing to BTF type data */
441 __u32 func_info_rec_size; /* userspace bpf_func_info size */
442 __aligned_u64 func_info; /* func info */
443 __u32 func_info_cnt; /* number of bpf_func_info records */
444 __u32 line_info_rec_size; /* userspace bpf_line_info size */
445 __aligned_u64 line_info; /* line info */
446 __u32 line_info_cnt; /* number of bpf_line_info records */
447 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
448 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
451 struct { /* anonymous struct used by BPF_OBJ_* commands */
452 __aligned_u64 pathname;
457 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
458 __u32 target_fd; /* container object to attach to */
459 __u32 attach_bpf_fd; /* eBPF program to attach */
462 __u32 replace_bpf_fd; /* previously attached eBPF
463 * program to replace if
464 * BPF_F_REPLACE is used
468 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
471 __u32 data_size_in; /* input: len of data_in */
472 __u32 data_size_out; /* input/output: len of data_out
473 * returns ENOSPC if data_out
476 __aligned_u64 data_in;
477 __aligned_u64 data_out;
480 __u32 ctx_size_in; /* input: len of ctx_in */
481 __u32 ctx_size_out; /* input/output: len of ctx_out
482 * returns ENOSPC if ctx_out
485 __aligned_u64 ctx_in;
486 __aligned_u64 ctx_out;
489 struct { /* anonymous struct used by BPF_*_GET_*_ID */
500 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
506 struct { /* anonymous struct used by BPF_PROG_QUERY command */
507 __u32 target_fd; /* container object to query */
511 __aligned_u64 prog_ids;
520 struct { /* anonymous struct for BPF_BTF_LOAD */
522 __aligned_u64 btf_log_buf;
529 __u32 pid; /* input: pid */
530 __u32 fd; /* input: fd */
531 __u32 flags; /* input: flags */
532 __u32 buf_len; /* input/output: buf len */
533 __aligned_u64 buf; /* input/output:
534 * tp_name for tracepoint
536 * filename for uprobe
538 __u32 prog_id; /* output: prod_id */
539 __u32 fd_type; /* output: BPF_FD_TYPE_* */
540 __u64 probe_offset; /* output: probe_offset */
541 __u64 probe_addr; /* output: probe_addr */
543 } __attribute__((aligned(8)));
545 /* The description below is an attempt at providing documentation to eBPF
546 * developers about the multiple available eBPF helper functions. It can be
547 * parsed and used to produce a manual page. The workflow is the following,
548 * and requires the rst2man utility:
550 * $ ./scripts/bpf_helpers_doc.py \
551 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
552 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
553 * $ man /tmp/bpf-helpers.7
555 * Note that in order to produce this external documentation, some RST
556 * formatting is used in the descriptions to get "bold" and "italics" in
557 * manual pages. Also note that the few trailing white spaces are
558 * intentional, removing them would break paragraphs for rst2man.
560 * Start of BPF helper function descriptions:
562 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
564 * Perform a lookup in *map* for an entry associated to *key*.
566 * Map value associated to *key*, or **NULL** if no entry was
569 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
571 * Add or update the value of the entry associated to *key* in
572 * *map* with *value*. *flags* is one of:
575 * The entry for *key* must not exist in the map.
577 * The entry for *key* must already exist in the map.
579 * No condition on the existence of the entry for *key*.
581 * Flag value **BPF_NOEXIST** cannot be used for maps of types
582 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
583 * elements always exist), the helper would return an error.
585 * 0 on success, or a negative error in case of failure.
587 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
589 * Delete entry with *key* from *map*.
591 * 0 on success, or a negative error in case of failure.
593 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
595 * For tracing programs, safely attempt to read *size* bytes from
596 * kernel space address *unsafe_ptr* and store the data in *dst*.
598 * Generally, use bpf_probe_read_user() or bpf_probe_read_kernel()
601 * 0 on success, or a negative error in case of failure.
603 * u64 bpf_ktime_get_ns(void)
605 * Return the time elapsed since system boot, in nanoseconds.
609 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
611 * This helper is a "printk()-like" facility for debugging. It
612 * prints a message defined by format *fmt* (of size *fmt_size*)
613 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
614 * available. It can take up to three additional **u64**
615 * arguments (as an eBPF helpers, the total number of arguments is
618 * Each time the helper is called, it appends a line to the trace.
619 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
620 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
621 * The format of the trace is customizable, and the exact output
622 * one will get depends on the options set in
623 * *\/sys/kernel/debug/tracing/trace_options* (see also the
624 * *README* file under the same directory). However, it usually
625 * defaults to something like:
629 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
633 * * ``telnet`` is the name of the current task.
634 * * ``470`` is the PID of the current task.
635 * * ``001`` is the CPU number on which the task is
637 * * In ``.N..``, each character refers to a set of
638 * options (whether irqs are enabled, scheduling
639 * options, whether hard/softirqs are running, level of
640 * preempt_disabled respectively). **N** means that
641 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
643 * * ``419421.045894`` is a timestamp.
644 * * ``0x00000001`` is a fake value used by BPF for the
645 * instruction pointer register.
646 * * ``<formatted msg>`` is the message formatted with
649 * The conversion specifiers supported by *fmt* are similar, but
650 * more limited than for printk(). They are **%d**, **%i**,
651 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
652 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
653 * of field, padding with zeroes, etc.) is available, and the
654 * helper will return **-EINVAL** (but print nothing) if it
655 * encounters an unknown specifier.
657 * Also, note that **bpf_trace_printk**\ () is slow, and should
658 * only be used for debugging purposes. For this reason, a notice
659 * bloc (spanning several lines) is printed to kernel logs and
660 * states that the helper should not be used "for production use"
661 * the first time this helper is used (or more precisely, when
662 * **trace_printk**\ () buffers are allocated). For passing values
663 * to user space, perf events should be preferred.
665 * The number of bytes written to the buffer, or a negative error
666 * in case of failure.
668 * u32 bpf_get_prandom_u32(void)
670 * Get a pseudo-random number.
672 * From a security point of view, this helper uses its own
673 * pseudo-random internal state, and cannot be used to infer the
674 * seed of other random functions in the kernel. However, it is
675 * essential to note that the generator used by the helper is not
676 * cryptographically secure.
678 * A random 32-bit unsigned value.
680 * u32 bpf_get_smp_processor_id(void)
682 * Get the SMP (symmetric multiprocessing) processor id. Note that
683 * all programs run with preemption disabled, which means that the
684 * SMP processor id is stable during all the execution of the
687 * The SMP id of the processor running the program.
689 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
691 * Store *len* bytes from address *from* into the packet
692 * associated to *skb*, at *offset*. *flags* are a combination of
693 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
694 * checksum for the packet after storing the bytes) and
695 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
696 * **->swhash** and *skb*\ **->l4hash** to 0).
698 * A call to this helper is susceptible to change the underlying
699 * packet buffer. Therefore, at load time, all checks on pointers
700 * previously done by the verifier are invalidated and must be
701 * performed again, if the helper is used in combination with
702 * direct packet access.
704 * 0 on success, or a negative error in case of failure.
706 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
708 * Recompute the layer 3 (e.g. IP) checksum for the packet
709 * associated to *skb*. Computation is incremental, so the helper
710 * must know the former value of the header field that was
711 * modified (*from*), the new value of this field (*to*), and the
712 * number of bytes (2 or 4) for this field, stored in *size*.
713 * Alternatively, it is possible to store the difference between
714 * the previous and the new values of the header field in *to*, by
715 * setting *from* and *size* to 0. For both methods, *offset*
716 * indicates the location of the IP checksum within the packet.
718 * This helper works in combination with **bpf_csum_diff**\ (),
719 * which does not update the checksum in-place, but offers more
720 * flexibility and can handle sizes larger than 2 or 4 for the
721 * checksum to update.
723 * A call to this helper is susceptible to change the underlying
724 * packet buffer. Therefore, at load time, all checks on pointers
725 * previously done by the verifier are invalidated and must be
726 * performed again, if the helper is used in combination with
727 * direct packet access.
729 * 0 on success, or a negative error in case of failure.
731 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
733 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
734 * packet associated to *skb*. Computation is incremental, so the
735 * helper must know the former value of the header field that was
736 * modified (*from*), the new value of this field (*to*), and the
737 * number of bytes (2 or 4) for this field, stored on the lowest
738 * four bits of *flags*. Alternatively, it is possible to store
739 * the difference between the previous and the new values of the
740 * header field in *to*, by setting *from* and the four lowest
741 * bits of *flags* to 0. For both methods, *offset* indicates the
742 * location of the IP checksum within the packet. In addition to
743 * the size of the field, *flags* can be added (bitwise OR) actual
744 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
745 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
746 * for updates resulting in a null checksum the value is set to
747 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
748 * the checksum is to be computed against a pseudo-header.
750 * This helper works in combination with **bpf_csum_diff**\ (),
751 * which does not update the checksum in-place, but offers more
752 * flexibility and can handle sizes larger than 2 or 4 for the
753 * checksum to update.
755 * A call to this helper is susceptible to change the underlying
756 * packet buffer. Therefore, at load time, all checks on pointers
757 * previously done by the verifier are invalidated and must be
758 * performed again, if the helper is used in combination with
759 * direct packet access.
761 * 0 on success, or a negative error in case of failure.
763 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
765 * This special helper is used to trigger a "tail call", or in
766 * other words, to jump into another eBPF program. The same stack
767 * frame is used (but values on stack and in registers for the
768 * caller are not accessible to the callee). This mechanism allows
769 * for program chaining, either for raising the maximum number of
770 * available eBPF instructions, or to execute given programs in
771 * conditional blocks. For security reasons, there is an upper
772 * limit to the number of successive tail calls that can be
775 * Upon call of this helper, the program attempts to jump into a
776 * program referenced at index *index* in *prog_array_map*, a
777 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
778 * *ctx*, a pointer to the context.
780 * If the call succeeds, the kernel immediately runs the first
781 * instruction of the new program. This is not a function call,
782 * and it never returns to the previous program. If the call
783 * fails, then the helper has no effect, and the caller continues
784 * to run its subsequent instructions. A call can fail if the
785 * destination program for the jump does not exist (i.e. *index*
786 * is superior to the number of entries in *prog_array_map*), or
787 * if the maximum number of tail calls has been reached for this
788 * chain of programs. This limit is defined in the kernel by the
789 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
790 * which is currently set to 32.
792 * 0 on success, or a negative error in case of failure.
794 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
796 * Clone and redirect the packet associated to *skb* to another
797 * net device of index *ifindex*. Both ingress and egress
798 * interfaces can be used for redirection. The **BPF_F_INGRESS**
799 * value in *flags* is used to make the distinction (ingress path
800 * is selected if the flag is present, egress path otherwise).
801 * This is the only flag supported for now.
803 * In comparison with **bpf_redirect**\ () helper,
804 * **bpf_clone_redirect**\ () has the associated cost of
805 * duplicating the packet buffer, but this can be executed out of
806 * the eBPF program. Conversely, **bpf_redirect**\ () is more
807 * efficient, but it is handled through an action code where the
808 * redirection happens only after the eBPF program has returned.
810 * A call to this helper is susceptible to change the underlying
811 * packet buffer. Therefore, at load time, all checks on pointers
812 * previously done by the verifier are invalidated and must be
813 * performed again, if the helper is used in combination with
814 * direct packet access.
816 * 0 on success, or a negative error in case of failure.
818 * u64 bpf_get_current_pid_tgid(void)
820 * A 64-bit integer containing the current tgid and pid, and
822 * *current_task*\ **->tgid << 32 \|**
823 * *current_task*\ **->pid**.
825 * u64 bpf_get_current_uid_gid(void)
827 * A 64-bit integer containing the current GID and UID, and
828 * created as such: *current_gid* **<< 32 \|** *current_uid*.
830 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
832 * Copy the **comm** attribute of the current task into *buf* of
833 * *size_of_buf*. The **comm** attribute contains the name of
834 * the executable (excluding the path) for the current task. The
835 * *size_of_buf* must be strictly positive. On success, the
836 * helper makes sure that the *buf* is NUL-terminated. On failure,
837 * it is filled with zeroes.
839 * 0 on success, or a negative error in case of failure.
841 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
843 * Retrieve the classid for the current task, i.e. for the net_cls
844 * cgroup to which *skb* belongs.
846 * This helper can be used on TC egress path, but not on ingress.
848 * The net_cls cgroup provides an interface to tag network packets
849 * based on a user-provided identifier for all traffic coming from
850 * the tasks belonging to the related cgroup. See also the related
851 * kernel documentation, available from the Linux sources in file
852 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
854 * The Linux kernel has two versions for cgroups: there are
855 * cgroups v1 and cgroups v2. Both are available to users, who can
856 * use a mixture of them, but note that the net_cls cgroup is for
857 * cgroup v1 only. This makes it incompatible with BPF programs
858 * run on cgroups, which is a cgroup-v2-only feature (a socket can
859 * only hold data for one version of cgroups at a time).
861 * This helper is only available is the kernel was compiled with
862 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
863 * "**y**" or to "**m**".
865 * The classid, or 0 for the default unconfigured classid.
867 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
869 * Push a *vlan_tci* (VLAN tag control information) of protocol
870 * *vlan_proto* to the packet associated to *skb*, then update
871 * the checksum. Note that if *vlan_proto* is different from
872 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
873 * be **ETH_P_8021Q**.
875 * A call to this helper is susceptible to change the underlying
876 * packet buffer. Therefore, at load time, all checks on pointers
877 * previously done by the verifier are invalidated and must be
878 * performed again, if the helper is used in combination with
879 * direct packet access.
881 * 0 on success, or a negative error in case of failure.
883 * int bpf_skb_vlan_pop(struct sk_buff *skb)
885 * Pop a VLAN header from the packet associated to *skb*.
887 * A call to this helper is susceptible to change the underlying
888 * packet buffer. Therefore, at load time, all checks on pointers
889 * previously done by the verifier are invalidated and must be
890 * performed again, if the helper is used in combination with
891 * direct packet access.
893 * 0 on success, or a negative error in case of failure.
895 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
897 * Get tunnel metadata. This helper takes a pointer *key* to an
898 * empty **struct bpf_tunnel_key** of **size**, that will be
899 * filled with tunnel metadata for the packet associated to *skb*.
900 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
901 * indicates that the tunnel is based on IPv6 protocol instead of
904 * The **struct bpf_tunnel_key** is an object that generalizes the
905 * principal parameters used by various tunneling protocols into a
906 * single struct. This way, it can be used to easily make a
907 * decision based on the contents of the encapsulation header,
908 * "summarized" in this struct. In particular, it holds the IP
909 * address of the remote end (IPv4 or IPv6, depending on the case)
910 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
911 * this struct exposes the *key*\ **->tunnel_id**, which is
912 * generally mapped to a VNI (Virtual Network Identifier), making
913 * it programmable together with the **bpf_skb_set_tunnel_key**\
916 * Let's imagine that the following code is part of a program
917 * attached to the TC ingress interface, on one end of a GRE
918 * tunnel, and is supposed to filter out all messages coming from
919 * remote ends with IPv4 address other than 10.0.0.1:
924 * struct bpf_tunnel_key key = {};
926 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
928 * return TC_ACT_SHOT; // drop packet
930 * if (key.remote_ipv4 != 0x0a000001)
931 * return TC_ACT_SHOT; // drop packet
933 * return TC_ACT_OK; // accept packet
935 * This interface can also be used with all encapsulation devices
936 * that can operate in "collect metadata" mode: instead of having
937 * one network device per specific configuration, the "collect
938 * metadata" mode only requires a single device where the
939 * configuration can be extracted from this helper.
941 * This can be used together with various tunnels such as VXLan,
942 * Geneve, GRE or IP in IP (IPIP).
944 * 0 on success, or a negative error in case of failure.
946 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
948 * Populate tunnel metadata for packet associated to *skb.* The
949 * tunnel metadata is set to the contents of *key*, of *size*. The
950 * *flags* can be set to a combination of the following values:
952 * **BPF_F_TUNINFO_IPV6**
953 * Indicate that the tunnel is based on IPv6 protocol
955 * **BPF_F_ZERO_CSUM_TX**
956 * For IPv4 packets, add a flag to tunnel metadata
957 * indicating that checksum computation should be skipped
958 * and checksum set to zeroes.
959 * **BPF_F_DONT_FRAGMENT**
960 * Add a flag to tunnel metadata indicating that the
961 * packet should not be fragmented.
962 * **BPF_F_SEQ_NUMBER**
963 * Add a flag to tunnel metadata indicating that a
964 * sequence number should be added to tunnel header before
965 * sending the packet. This flag was added for GRE
966 * encapsulation, but might be used with other protocols
967 * as well in the future.
969 * Here is a typical usage on the transmit path:
973 * struct bpf_tunnel_key key;
975 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
976 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
978 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
979 * helper for additional information.
981 * 0 on success, or a negative error in case of failure.
983 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
985 * Read the value of a perf event counter. This helper relies on a
986 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
987 * the perf event counter is selected when *map* is updated with
988 * perf event file descriptors. The *map* is an array whose size
989 * is the number of available CPUs, and each cell contains a value
990 * relative to one CPU. The value to retrieve is indicated by
991 * *flags*, that contains the index of the CPU to look up, masked
992 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
993 * **BPF_F_CURRENT_CPU** to indicate that the value for the
994 * current CPU should be retrieved.
996 * Note that before Linux 4.13, only hardware perf event can be
999 * Also, be aware that the newer helper
1000 * **bpf_perf_event_read_value**\ () is recommended over
1001 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1002 * quirks where error and counter value are used as a return code
1003 * (which is wrong to do since ranges may overlap). This issue is
1004 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1005 * time provides more features over the **bpf_perf_event_read**\
1006 * () interface. Please refer to the description of
1007 * **bpf_perf_event_read_value**\ () for details.
1009 * The value of the perf event counter read from the map, or a
1010 * negative error code in case of failure.
1012 * int bpf_redirect(u32 ifindex, u64 flags)
1014 * Redirect the packet to another net device of index *ifindex*.
1015 * This helper is somewhat similar to **bpf_clone_redirect**\
1016 * (), except that the packet is not cloned, which provides
1017 * increased performance.
1019 * Except for XDP, both ingress and egress interfaces can be used
1020 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1021 * to make the distinction (ingress path is selected if the flag
1022 * is present, egress path otherwise). Currently, XDP only
1023 * supports redirection to the egress interface, and accepts no
1026 * The same effect can be attained with the more generic
1027 * **bpf_redirect_map**\ (), which requires specific maps to be
1028 * used but offers better performance.
1030 * For XDP, the helper returns **XDP_REDIRECT** on success or
1031 * **XDP_ABORTED** on error. For other program types, the values
1032 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1035 * u32 bpf_get_route_realm(struct sk_buff *skb)
1037 * Retrieve the realm or the route, that is to say the
1038 * **tclassid** field of the destination for the *skb*. The
1039 * indentifier retrieved is a user-provided tag, similar to the
1040 * one used with the net_cls cgroup (see description for
1041 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1042 * held by a route (a destination entry), not by a task.
1044 * Retrieving this identifier works with the clsact TC egress hook
1045 * (see also **tc-bpf(8)**), or alternatively on conventional
1046 * classful egress qdiscs, but not on TC ingress path. In case of
1047 * clsact TC egress hook, this has the advantage that, internally,
1048 * the destination entry has not been dropped yet in the transmit
1049 * path. Therefore, the destination entry does not need to be
1050 * artificially held via **netif_keep_dst**\ () for a classful
1051 * qdisc until the *skb* is freed.
1053 * This helper is available only if the kernel was compiled with
1054 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1056 * The realm of the route for the packet associated to *skb*, or 0
1057 * if none was found.
1059 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1061 * Write raw *data* blob into a special BPF perf event held by
1062 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1063 * event must have the following attributes: **PERF_SAMPLE_RAW**
1064 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1065 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1067 * The *flags* are used to indicate the index in *map* for which
1068 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1069 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1070 * to indicate that the index of the current CPU core should be
1073 * The value to write, of *size*, is passed through eBPF stack and
1074 * pointed by *data*.
1076 * The context of the program *ctx* needs also be passed to the
1079 * On user space, a program willing to read the values needs to
1080 * call **perf_event_open**\ () on the perf event (either for
1081 * one or for all CPUs) and to store the file descriptor into the
1082 * *map*. This must be done before the eBPF program can send data
1083 * into it. An example is available in file
1084 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1085 * tree (the eBPF program counterpart is in
1086 * *samples/bpf/trace_output_kern.c*).
1088 * **bpf_perf_event_output**\ () achieves better performance
1089 * than **bpf_trace_printk**\ () for sharing data with user
1090 * space, and is much better suitable for streaming data from eBPF
1093 * Note that this helper is not restricted to tracing use cases
1094 * and can be used with programs attached to TC or XDP as well,
1095 * where it allows for passing data to user space listeners. Data
1098 * * Only custom structs,
1099 * * Only the packet payload, or
1100 * * A combination of both.
1102 * 0 on success, or a negative error in case of failure.
1104 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1106 * This helper was provided as an easy way to load data from a
1107 * packet. It can be used to load *len* bytes from *offset* from
1108 * the packet associated to *skb*, into the buffer pointed by
1111 * Since Linux 4.7, usage of this helper has mostly been replaced
1112 * by "direct packet access", enabling packet data to be
1113 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1114 * pointing respectively to the first byte of packet data and to
1115 * the byte after the last byte of packet data. However, it
1116 * remains useful if one wishes to read large quantities of data
1117 * at once from a packet into the eBPF stack.
1119 * 0 on success, or a negative error in case of failure.
1121 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1123 * Walk a user or a kernel stack and return its id. To achieve
1124 * this, the helper needs *ctx*, which is a pointer to the context
1125 * on which the tracing program is executed, and a pointer to a
1126 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1128 * The last argument, *flags*, holds the number of stack frames to
1129 * skip (from 0 to 255), masked with
1130 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1131 * a combination of the following flags:
1133 * **BPF_F_USER_STACK**
1134 * Collect a user space stack instead of a kernel stack.
1135 * **BPF_F_FAST_STACK_CMP**
1136 * Compare stacks by hash only.
1137 * **BPF_F_REUSE_STACKID**
1138 * If two different stacks hash into the same *stackid*,
1139 * discard the old one.
1141 * The stack id retrieved is a 32 bit long integer handle which
1142 * can be further combined with other data (including other stack
1143 * ids) and used as a key into maps. This can be useful for
1144 * generating a variety of graphs (such as flame graphs or off-cpu
1147 * For walking a stack, this helper is an improvement over
1148 * **bpf_probe_read**\ (), which can be used with unrolled loops
1149 * but is not efficient and consumes a lot of eBPF instructions.
1150 * Instead, **bpf_get_stackid**\ () can collect up to
1151 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1152 * this limit can be controlled with the **sysctl** program, and
1153 * that it should be manually increased in order to profile long
1154 * user stacks (such as stacks for Java programs). To do so, use:
1158 * # sysctl kernel.perf_event_max_stack=<new value>
1160 * The positive or null stack id on success, or a negative error
1161 * in case of failure.
1163 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1165 * Compute a checksum difference, from the raw buffer pointed by
1166 * *from*, of length *from_size* (that must be a multiple of 4),
1167 * towards the raw buffer pointed by *to*, of size *to_size*
1168 * (same remark). An optional *seed* can be added to the value
1169 * (this can be cascaded, the seed may come from a previous call
1172 * This is flexible enough to be used in several ways:
1174 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1175 * checksum, it can be used when pushing new data.
1176 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1177 * checksum, it can be used when removing data from a packet.
1178 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1179 * can be used to compute a diff. Note that *from_size* and
1180 * *to_size* do not need to be equal.
1182 * This helper can be used in combination with
1183 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1184 * which one can feed in the difference computed with
1185 * **bpf_csum_diff**\ ().
1187 * The checksum result, or a negative error code in case of
1190 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1192 * Retrieve tunnel options metadata for the packet associated to
1193 * *skb*, and store the raw tunnel option data to the buffer *opt*
1196 * This helper can be used with encapsulation devices that can
1197 * operate in "collect metadata" mode (please refer to the related
1198 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1199 * more details). A particular example where this can be used is
1200 * in combination with the Geneve encapsulation protocol, where it
1201 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1202 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1203 * the eBPF program. This allows for full customization of these
1206 * The size of the option data retrieved.
1208 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1210 * Set tunnel options metadata for the packet associated to *skb*
1211 * to the option data contained in the raw buffer *opt* of *size*.
1213 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1214 * helper for additional information.
1216 * 0 on success, or a negative error in case of failure.
1218 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1220 * Change the protocol of the *skb* to *proto*. Currently
1221 * supported are transition from IPv4 to IPv6, and from IPv6 to
1222 * IPv4. The helper takes care of the groundwork for the
1223 * transition, including resizing the socket buffer. The eBPF
1224 * program is expected to fill the new headers, if any, via
1225 * **skb_store_bytes**\ () and to recompute the checksums with
1226 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1227 * (). The main case for this helper is to perform NAT64
1228 * operations out of an eBPF program.
1230 * Internally, the GSO type is marked as dodgy so that headers are
1231 * checked and segments are recalculated by the GSO/GRO engine.
1232 * The size for GSO target is adapted as well.
1234 * All values for *flags* are reserved for future usage, and must
1237 * A call to this helper is susceptible to change the underlying
1238 * packet buffer. Therefore, at load time, all checks on pointers
1239 * previously done by the verifier are invalidated and must be
1240 * performed again, if the helper is used in combination with
1241 * direct packet access.
1243 * 0 on success, or a negative error in case of failure.
1245 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1247 * Change the packet type for the packet associated to *skb*. This
1248 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1249 * the eBPF program does not have a write access to *skb*\
1250 * **->pkt_type** beside this helper. Using a helper here allows
1251 * for graceful handling of errors.
1253 * The major use case is to change incoming *skb*s to
1254 * **PACKET_HOST** in a programmatic way instead of having to
1255 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1258 * Note that *type* only allows certain values. At this time, they
1263 * **PACKET_BROADCAST**
1264 * Send packet to all.
1265 * **PACKET_MULTICAST**
1266 * Send packet to group.
1267 * **PACKET_OTHERHOST**
1268 * Send packet to someone else.
1270 * 0 on success, or a negative error in case of failure.
1272 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1274 * Check whether *skb* is a descendant of the cgroup2 held by
1275 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1277 * The return value depends on the result of the test, and can be:
1279 * * 0, if the *skb* failed the cgroup2 descendant test.
1280 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1281 * * A negative error code, if an error occurred.
1283 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1285 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1286 * not set, in particular if the hash was cleared due to mangling,
1287 * recompute this hash. Later accesses to the hash can be done
1288 * directly with *skb*\ **->hash**.
1290 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1291 * prototype with **bpf_skb_change_proto**\ (), or calling
1292 * **bpf_skb_store_bytes**\ () with the
1293 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1294 * the hash and to trigger a new computation for the next call to
1295 * **bpf_get_hash_recalc**\ ().
1299 * u64 bpf_get_current_task(void)
1301 * A pointer to the current task struct.
1303 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1305 * Attempt in a safe way to write *len* bytes from the buffer
1306 * *src* to *dst* in memory. It only works for threads that are in
1307 * user context, and *dst* must be a valid user space address.
1309 * This helper should not be used to implement any kind of
1310 * security mechanism because of TOC-TOU attacks, but rather to
1311 * debug, divert, and manipulate execution of semi-cooperative
1314 * Keep in mind that this feature is meant for experiments, and it
1315 * has a risk of crashing the system and running programs.
1316 * Therefore, when an eBPF program using this helper is attached,
1317 * a warning including PID and process name is printed to kernel
1320 * 0 on success, or a negative error in case of failure.
1322 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1324 * Check whether the probe is being run is the context of a given
1325 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1326 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1328 * The return value depends on the result of the test, and can be:
1330 * * 0, if the *skb* task belongs to the cgroup2.
1331 * * 1, if the *skb* task does not belong to the cgroup2.
1332 * * A negative error code, if an error occurred.
1334 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1336 * Resize (trim or grow) the packet associated to *skb* to the
1337 * new *len*. The *flags* are reserved for future usage, and must
1340 * The basic idea is that the helper performs the needed work to
1341 * change the size of the packet, then the eBPF program rewrites
1342 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1343 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1344 * and others. This helper is a slow path utility intended for
1345 * replies with control messages. And because it is targeted for
1346 * slow path, the helper itself can afford to be slow: it
1347 * implicitly linearizes, unclones and drops offloads from the
1350 * A call to this helper is susceptible to change the underlying
1351 * packet buffer. Therefore, at load time, all checks on pointers
1352 * previously done by the verifier are invalidated and must be
1353 * performed again, if the helper is used in combination with
1354 * direct packet access.
1356 * 0 on success, or a negative error in case of failure.
1358 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1360 * Pull in non-linear data in case the *skb* is non-linear and not
1361 * all of *len* are part of the linear section. Make *len* bytes
1362 * from *skb* readable and writable. If a zero value is passed for
1363 * *len*, then the whole length of the *skb* is pulled.
1365 * This helper is only needed for reading and writing with direct
1368 * For direct packet access, testing that offsets to access
1369 * are within packet boundaries (test on *skb*\ **->data_end**) is
1370 * susceptible to fail if offsets are invalid, or if the requested
1371 * data is in non-linear parts of the *skb*. On failure the
1372 * program can just bail out, or in the case of a non-linear
1373 * buffer, use a helper to make the data available. The
1374 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1375 * the data. Another one consists in using **bpf_skb_pull_data**
1376 * to pull in once the non-linear parts, then retesting and
1377 * eventually access the data.
1379 * At the same time, this also makes sure the *skb* is uncloned,
1380 * which is a necessary condition for direct write. As this needs
1381 * to be an invariant for the write part only, the verifier
1382 * detects writes and adds a prologue that is calling
1383 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1384 * the very beginning in case it is indeed cloned.
1386 * A call to this helper is susceptible to change the underlying
1387 * packet buffer. Therefore, at load time, all checks on pointers
1388 * previously done by the verifier are invalidated and must be
1389 * performed again, if the helper is used in combination with
1390 * direct packet access.
1392 * 0 on success, or a negative error in case of failure.
1394 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1396 * Add the checksum *csum* into *skb*\ **->csum** in case the
1397 * driver has supplied a checksum for the entire packet into that
1398 * field. Return an error otherwise. This helper is intended to be
1399 * used in combination with **bpf_csum_diff**\ (), in particular
1400 * when the checksum needs to be updated after data has been
1401 * written into the packet through direct packet access.
1403 * The checksum on success, or a negative error code in case of
1406 * void bpf_set_hash_invalid(struct sk_buff *skb)
1408 * Invalidate the current *skb*\ **->hash**. It can be used after
1409 * mangling on headers through direct packet access, in order to
1410 * indicate that the hash is outdated and to trigger a
1411 * recalculation the next time the kernel tries to access this
1412 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1414 * int bpf_get_numa_node_id(void)
1416 * Return the id of the current NUMA node. The primary use case
1417 * for this helper is the selection of sockets for the local NUMA
1418 * node, when the program is attached to sockets using the
1419 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1420 * but the helper is also available to other eBPF program types,
1421 * similarly to **bpf_get_smp_processor_id**\ ().
1423 * The id of current NUMA node.
1425 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1427 * Grows headroom of packet associated to *skb* and adjusts the
1428 * offset of the MAC header accordingly, adding *len* bytes of
1429 * space. It automatically extends and reallocates memory as
1432 * This helper can be used on a layer 3 *skb* to push a MAC header
1433 * for redirection into a layer 2 device.
1435 * All values for *flags* are reserved for future usage, and must
1438 * A call to this helper is susceptible to change the underlying
1439 * packet buffer. Therefore, at load time, all checks on pointers
1440 * previously done by the verifier are invalidated and must be
1441 * performed again, if the helper is used in combination with
1442 * direct packet access.
1444 * 0 on success, or a negative error in case of failure.
1446 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1448 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1449 * it is possible to use a negative value for *delta*. This helper
1450 * can be used to prepare the packet for pushing or popping
1453 * A call to this helper is susceptible to change the underlying
1454 * packet buffer. Therefore, at load time, all checks on pointers
1455 * previously done by the verifier are invalidated and must be
1456 * performed again, if the helper is used in combination with
1457 * direct packet access.
1459 * 0 on success, or a negative error in case of failure.
1461 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1463 * Copy a NUL terminated string from an unsafe kernel address
1464 * *unsafe_ptr* to *dst*. See bpf_probe_read_kernel_str() for
1467 * Generally, use bpf_probe_read_user_str() or bpf_probe_read_kernel_str()
1470 * On success, the strictly positive length of the string,
1471 * including the trailing NUL character. On error, a negative
1474 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1476 * If the **struct sk_buff** pointed by *skb* has a known socket,
1477 * retrieve the cookie (generated by the kernel) of this socket.
1478 * If no cookie has been set yet, generate a new cookie. Once
1479 * generated, the socket cookie remains stable for the life of the
1480 * socket. This helper can be useful for monitoring per socket
1481 * networking traffic statistics as it provides a global socket
1482 * identifier that can be assumed unique.
1484 * A 8-byte long non-decreasing number on success, or 0 if the
1485 * socket field is missing inside *skb*.
1487 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1489 * Equivalent to bpf_get_socket_cookie() helper that accepts
1490 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1492 * A 8-byte long non-decreasing number.
1494 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1496 * Equivalent to bpf_get_socket_cookie() helper that accepts
1497 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1499 * A 8-byte long non-decreasing number.
1501 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1503 * The owner UID of the socket associated to *skb*. If the socket
1504 * is **NULL**, or if it is not a full socket (i.e. if it is a
1505 * time-wait or a request socket instead), **overflowuid** value
1506 * is returned (note that **overflowuid** might also be the actual
1507 * UID value for the socket).
1509 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1511 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1516 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1518 * Emulate a call to **setsockopt()** on the socket associated to
1519 * *bpf_socket*, which must be a full socket. The *level* at
1520 * which the option resides and the name *optname* of the option
1521 * must be specified, see **setsockopt(2)** for more information.
1522 * The option value of length *optlen* is pointed by *optval*.
1524 * This helper actually implements a subset of **setsockopt()**.
1525 * It supports the following *level*\ s:
1527 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1528 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1529 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1530 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1531 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1532 * **TCP_BPF_SNDCWND_CLAMP**.
1533 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1534 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1536 * 0 on success, or a negative error in case of failure.
1538 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1540 * Grow or shrink the room for data in the packet associated to
1541 * *skb* by *len_diff*, and according to the selected *mode*.
1543 * There are two supported modes at this time:
1545 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1546 * (room space is added or removed below the layer 2 header).
1548 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1549 * (room space is added or removed below the layer 3 header).
1551 * The following flags are supported at this time:
1553 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1554 * Adjusting mss in this way is not allowed for datagrams.
1556 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1557 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1558 * Any new space is reserved to hold a tunnel header.
1559 * Configure skb offsets and other fields accordingly.
1561 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1562 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1563 * Use with ENCAP_L3 flags to further specify the tunnel type.
1565 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1566 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1567 * type; *len* is the length of the inner MAC header.
1569 * A call to this helper is susceptible to change the underlying
1570 * packet buffer. Therefore, at load time, all checks on pointers
1571 * previously done by the verifier are invalidated and must be
1572 * performed again, if the helper is used in combination with
1573 * direct packet access.
1575 * 0 on success, or a negative error in case of failure.
1577 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1579 * Redirect the packet to the endpoint referenced by *map* at
1580 * index *key*. Depending on its type, this *map* can contain
1581 * references to net devices (for forwarding packets through other
1582 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1583 * but this is only implemented for native XDP (with driver
1584 * support) as of this writing).
1586 * The lower two bits of *flags* are used as the return code if
1587 * the map lookup fails. This is so that the return value can be
1588 * one of the XDP program return codes up to XDP_TX, as chosen by
1589 * the caller. Any higher bits in the *flags* argument must be
1592 * When used to redirect packets to net devices, this helper
1593 * provides a high performance increase over **bpf_redirect**\ ().
1594 * This is due to various implementation details of the underlying
1595 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1596 * () tries to send packet as a "bulk" to the device.
1598 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1600 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1602 * Redirect the packet to the socket referenced by *map* (of type
1603 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1604 * egress interfaces can be used for redirection. The
1605 * **BPF_F_INGRESS** value in *flags* is used to make the
1606 * distinction (ingress path is selected if the flag is present,
1607 * egress path otherwise). This is the only flag supported for now.
1609 * **SK_PASS** on success, or **SK_DROP** on error.
1611 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1613 * Add an entry to, or update a *map* referencing sockets. The
1614 * *skops* is used as a new value for the entry associated to
1615 * *key*. *flags* is one of:
1618 * The entry for *key* must not exist in the map.
1620 * The entry for *key* must already exist in the map.
1622 * No condition on the existence of the entry for *key*.
1624 * If the *map* has eBPF programs (parser and verdict), those will
1625 * be inherited by the socket being added. If the socket is
1626 * already attached to eBPF programs, this results in an error.
1628 * 0 on success, or a negative error in case of failure.
1630 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1632 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1633 * *delta* (which can be positive or negative). Note that this
1634 * operation modifies the address stored in *xdp_md*\ **->data**,
1635 * so the latter must be loaded only after the helper has been
1638 * The use of *xdp_md*\ **->data_meta** is optional and programs
1639 * are not required to use it. The rationale is that when the
1640 * packet is processed with XDP (e.g. as DoS filter), it is
1641 * possible to push further meta data along with it before passing
1642 * to the stack, and to give the guarantee that an ingress eBPF
1643 * program attached as a TC classifier on the same device can pick
1644 * this up for further post-processing. Since TC works with socket
1645 * buffers, it remains possible to set from XDP the **mark** or
1646 * **priority** pointers, or other pointers for the socket buffer.
1647 * Having this scratch space generic and programmable allows for
1648 * more flexibility as the user is free to store whatever meta
1651 * A call to this helper is susceptible to change the underlying
1652 * packet buffer. Therefore, at load time, all checks on pointers
1653 * previously done by the verifier are invalidated and must be
1654 * performed again, if the helper is used in combination with
1655 * direct packet access.
1657 * 0 on success, or a negative error in case of failure.
1659 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1661 * Read the value of a perf event counter, and store it into *buf*
1662 * of size *buf_size*. This helper relies on a *map* of type
1663 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1664 * counter is selected when *map* is updated with perf event file
1665 * descriptors. The *map* is an array whose size is the number of
1666 * available CPUs, and each cell contains a value relative to one
1667 * CPU. The value to retrieve is indicated by *flags*, that
1668 * contains the index of the CPU to look up, masked with
1669 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1670 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1671 * current CPU should be retrieved.
1673 * This helper behaves in a way close to
1674 * **bpf_perf_event_read**\ () helper, save that instead of
1675 * just returning the value observed, it fills the *buf*
1676 * structure. This allows for additional data to be retrieved: in
1677 * particular, the enabled and running times (in *buf*\
1678 * **->enabled** and *buf*\ **->running**, respectively) are
1679 * copied. In general, **bpf_perf_event_read_value**\ () is
1680 * recommended over **bpf_perf_event_read**\ (), which has some
1681 * ABI issues and provides fewer functionalities.
1683 * These values are interesting, because hardware PMU (Performance
1684 * Monitoring Unit) counters are limited resources. When there are
1685 * more PMU based perf events opened than available counters,
1686 * kernel will multiplex these events so each event gets certain
1687 * percentage (but not all) of the PMU time. In case that
1688 * multiplexing happens, the number of samples or counter value
1689 * will not reflect the case compared to when no multiplexing
1690 * occurs. This makes comparison between different runs difficult.
1691 * Typically, the counter value should be normalized before
1692 * comparing to other experiments. The usual normalization is done
1697 * normalized_counter = counter * t_enabled / t_running
1699 * Where t_enabled is the time enabled for event and t_running is
1700 * the time running for event since last normalization. The
1701 * enabled and running times are accumulated since the perf event
1702 * open. To achieve scaling factor between two invocations of an
1703 * eBPF program, users can can use CPU id as the key (which is
1704 * typical for perf array usage model) to remember the previous
1705 * value and do the calculation inside the eBPF program.
1707 * 0 on success, or a negative error in case of failure.
1709 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1711 * For en eBPF program attached to a perf event, retrieve the
1712 * value of the event counter associated to *ctx* and store it in
1713 * the structure pointed by *buf* and of size *buf_size*. Enabled
1714 * and running times are also stored in the structure (see
1715 * description of helper **bpf_perf_event_read_value**\ () for
1718 * 0 on success, or a negative error in case of failure.
1720 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1722 * Emulate a call to **getsockopt()** on the socket associated to
1723 * *bpf_socket*, which must be a full socket. The *level* at
1724 * which the option resides and the name *optname* of the option
1725 * must be specified, see **getsockopt(2)** for more information.
1726 * The retrieved value is stored in the structure pointed by
1727 * *opval* and of length *optlen*.
1729 * This helper actually implements a subset of **getsockopt()**.
1730 * It supports the following *level*\ s:
1732 * * **IPPROTO_TCP**, which supports *optname*
1733 * **TCP_CONGESTION**.
1734 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1735 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1737 * 0 on success, or a negative error in case of failure.
1739 * int bpf_override_return(struct pt_regs *regs, u64 rc)
1741 * Used for error injection, this helper uses kprobes to override
1742 * the return value of the probed function, and to set it to *rc*.
1743 * The first argument is the context *regs* on which the kprobe
1746 * This helper works by setting setting the PC (program counter)
1747 * to an override function which is run in place of the original
1748 * probed function. This means the probed function is not run at
1749 * all. The replacement function just returns with the required
1752 * This helper has security implications, and thus is subject to
1753 * restrictions. It is only available if the kernel was compiled
1754 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1755 * option, and in this case it only works on functions tagged with
1756 * **ALLOW_ERROR_INJECTION** in the kernel code.
1758 * Also, the helper is only available for the architectures having
1759 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1760 * x86 architecture is the only one to support this feature.
1764 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1766 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1767 * for the full TCP socket associated to *bpf_sock_ops* to
1770 * The primary use of this field is to determine if there should
1771 * be calls to eBPF programs of type
1772 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1773 * code. A program of the same type can change its value, per
1774 * connection and as necessary, when the connection is
1775 * established. This field is directly accessible for reading, but
1776 * this helper must be used for updates in order to return an
1777 * error if an eBPF program tries to set a callback that is not
1778 * supported in the current kernel.
1780 * *argval* is a flag array which can combine these flags:
1782 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1783 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1784 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1785 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1787 * Therefore, this function can be used to clear a callback flag by
1788 * setting the appropriate bit to zero. e.g. to disable the RTO
1791 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1792 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1794 * Here are some examples of where one could call such eBPF
1798 * * When a packet is retransmitted.
1799 * * When the connection terminates.
1800 * * When a packet is sent.
1801 * * When a packet is received.
1803 * Code **-EINVAL** if the socket is not a full TCP socket;
1804 * otherwise, a positive number containing the bits that could not
1805 * be set is returned (which comes down to 0 if all bits were set
1808 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1810 * This helper is used in programs implementing policies at the
1811 * socket level. If the message *msg* is allowed to pass (i.e. if
1812 * the verdict eBPF program returns **SK_PASS**), redirect it to
1813 * the socket referenced by *map* (of type
1814 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1815 * egress interfaces can be used for redirection. The
1816 * **BPF_F_INGRESS** value in *flags* is used to make the
1817 * distinction (ingress path is selected if the flag is present,
1818 * egress path otherwise). This is the only flag supported for now.
1820 * **SK_PASS** on success, or **SK_DROP** on error.
1822 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1824 * For socket policies, apply the verdict of the eBPF program to
1825 * the next *bytes* (number of bytes) of message *msg*.
1827 * For example, this helper can be used in the following cases:
1829 * * A single **sendmsg**\ () or **sendfile**\ () system call
1830 * contains multiple logical messages that the eBPF program is
1831 * supposed to read and for which it should apply a verdict.
1832 * * An eBPF program only cares to read the first *bytes* of a
1833 * *msg*. If the message has a large payload, then setting up
1834 * and calling the eBPF program repeatedly for all bytes, even
1835 * though the verdict is already known, would create unnecessary
1838 * When called from within an eBPF program, the helper sets a
1839 * counter internal to the BPF infrastructure, that is used to
1840 * apply the last verdict to the next *bytes*. If *bytes* is
1841 * smaller than the current data being processed from a
1842 * **sendmsg**\ () or **sendfile**\ () system call, the first
1843 * *bytes* will be sent and the eBPF program will be re-run with
1844 * the pointer for start of data pointing to byte number *bytes*
1845 * **+ 1**. If *bytes* is larger than the current data being
1846 * processed, then the eBPF verdict will be applied to multiple
1847 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1850 * Note that if a socket closes with the internal counter holding
1851 * a non-zero value, this is not a problem because data is not
1852 * being buffered for *bytes* and is sent as it is received.
1856 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1858 * For socket policies, prevent the execution of the verdict eBPF
1859 * program for message *msg* until *bytes* (byte number) have been
1862 * This can be used when one needs a specific number of bytes
1863 * before a verdict can be assigned, even if the data spans
1864 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1865 * case would be a user calling **sendmsg**\ () repeatedly with
1866 * 1-byte long message segments. Obviously, this is bad for
1867 * performance, but it is still valid. If the eBPF program needs
1868 * *bytes* bytes to validate a header, this helper can be used to
1869 * prevent the eBPF program to be called again until *bytes* have
1874 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1876 * For socket policies, pull in non-linear data from user space
1877 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1878 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1881 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1882 * *msg* it can only parse data that the (**data**, **data_end**)
1883 * pointers have already consumed. For **sendmsg**\ () hooks this
1884 * is likely the first scatterlist element. But for calls relying
1885 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1886 * be the range (**0**, **0**) because the data is shared with
1887 * user space and by default the objective is to avoid allowing
1888 * user space to modify data while (or after) eBPF verdict is
1889 * being decided. This helper can be used to pull in data and to
1890 * set the start and end pointer to given values. Data will be
1891 * copied if necessary (i.e. if data was not linear and if start
1892 * and end pointers do not point to the same chunk).
1894 * A call to this helper is susceptible to change the underlying
1895 * packet buffer. Therefore, at load time, all checks on pointers
1896 * previously done by the verifier are invalidated and must be
1897 * performed again, if the helper is used in combination with
1898 * direct packet access.
1900 * All values for *flags* are reserved for future usage, and must
1903 * 0 on success, or a negative error in case of failure.
1905 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1907 * Bind the socket associated to *ctx* to the address pointed by
1908 * *addr*, of length *addr_len*. This allows for making outgoing
1909 * connection from the desired IP address, which can be useful for
1910 * example when all processes inside a cgroup should use one
1911 * single IP address on a host that has multiple IP configured.
1913 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1914 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1915 * **AF_INET6**). Looking for a free port to bind to can be
1916 * expensive, therefore binding to port is not permitted by the
1917 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1918 * must be set to zero.
1920 * 0 on success, or a negative error in case of failure.
1922 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1924 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1925 * only possible to shrink the packet as of this writing,
1926 * therefore *delta* must be a negative integer.
1928 * A call to this helper is susceptible to change the underlying
1929 * packet buffer. Therefore, at load time, all checks on pointers
1930 * previously done by the verifier are invalidated and must be
1931 * performed again, if the helper is used in combination with
1932 * direct packet access.
1934 * 0 on success, or a negative error in case of failure.
1936 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1938 * Retrieve the XFRM state (IP transform framework, see also
1939 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1941 * The retrieved value is stored in the **struct bpf_xfrm_state**
1942 * pointed by *xfrm_state* and of length *size*.
1944 * All values for *flags* are reserved for future usage, and must
1947 * This helper is available only if the kernel was compiled with
1948 * **CONFIG_XFRM** configuration option.
1950 * 0 on success, or a negative error in case of failure.
1952 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
1954 * Return a user or a kernel stack in bpf program provided buffer.
1955 * To achieve this, the helper needs *ctx*, which is a pointer
1956 * to the context on which the tracing program is executed.
1957 * To store the stacktrace, the bpf program provides *buf* with
1958 * a nonnegative *size*.
1960 * The last argument, *flags*, holds the number of stack frames to
1961 * skip (from 0 to 255), masked with
1962 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1963 * the following flags:
1965 * **BPF_F_USER_STACK**
1966 * Collect a user space stack instead of a kernel stack.
1967 * **BPF_F_USER_BUILD_ID**
1968 * Collect buildid+offset instead of ips for user stack,
1969 * only valid if **BPF_F_USER_STACK** is also specified.
1971 * **bpf_get_stack**\ () can collect up to
1972 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1973 * to sufficient large buffer size. Note that
1974 * this limit can be controlled with the **sysctl** program, and
1975 * that it should be manually increased in order to profile long
1976 * user stacks (such as stacks for Java programs). To do so, use:
1980 * # sysctl kernel.perf_event_max_stack=<new value>
1982 * A non-negative value equal to or less than *size* on success,
1983 * or a negative error in case of failure.
1985 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
1987 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1988 * it provides an easy way to load *len* bytes from *offset*
1989 * from the packet associated to *skb*, into the buffer pointed
1990 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1991 * a fifth argument *start_header* exists in order to select a
1992 * base offset to start from. *start_header* can be one of:
1994 * **BPF_HDR_START_MAC**
1995 * Base offset to load data from is *skb*'s mac header.
1996 * **BPF_HDR_START_NET**
1997 * Base offset to load data from is *skb*'s network header.
1999 * In general, "direct packet access" is the preferred method to
2000 * access packet data, however, this helper is in particular useful
2001 * in socket filters where *skb*\ **->data** does not always point
2002 * to the start of the mac header and where "direct packet access"
2005 * 0 on success, or a negative error in case of failure.
2007 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2009 * Do FIB lookup in kernel tables using parameters in *params*.
2010 * If lookup is successful and result shows packet is to be
2011 * forwarded, the neighbor tables are searched for the nexthop.
2012 * If successful (ie., FIB lookup shows forwarding and nexthop
2013 * is resolved), the nexthop address is returned in ipv4_dst
2014 * or ipv6_dst based on family, smac is set to mac address of
2015 * egress device, dmac is set to nexthop mac address, rt_metric
2016 * is set to metric from route (IPv4/IPv6 only), and ifindex
2017 * is set to the device index of the nexthop from the FIB lookup.
2019 * *plen* argument is the size of the passed in struct.
2020 * *flags* argument can be a combination of one or more of the
2023 * **BPF_FIB_LOOKUP_DIRECT**
2024 * Do a direct table lookup vs full lookup using FIB
2026 * **BPF_FIB_LOOKUP_OUTPUT**
2027 * Perform lookup from an egress perspective (default is
2030 * *ctx* is either **struct xdp_md** for XDP programs or
2031 * **struct sk_buff** tc cls_act programs.
2033 * * < 0 if any input argument is invalid
2034 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2035 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2036 * packet is not forwarded or needs assist from full stack
2038 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2040 * Add an entry to, or update a sockhash *map* referencing sockets.
2041 * The *skops* is used as a new value for the entry associated to
2042 * *key*. *flags* is one of:
2045 * The entry for *key* must not exist in the map.
2047 * The entry for *key* must already exist in the map.
2049 * No condition on the existence of the entry for *key*.
2051 * If the *map* has eBPF programs (parser and verdict), those will
2052 * be inherited by the socket being added. If the socket is
2053 * already attached to eBPF programs, this results in an error.
2055 * 0 on success, or a negative error in case of failure.
2057 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2059 * This helper is used in programs implementing policies at the
2060 * socket level. If the message *msg* is allowed to pass (i.e. if
2061 * the verdict eBPF program returns **SK_PASS**), redirect it to
2062 * the socket referenced by *map* (of type
2063 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2064 * egress interfaces can be used for redirection. The
2065 * **BPF_F_INGRESS** value in *flags* is used to make the
2066 * distinction (ingress path is selected if the flag is present,
2067 * egress path otherwise). This is the only flag supported for now.
2069 * **SK_PASS** on success, or **SK_DROP** on error.
2071 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2073 * This helper is used in programs implementing policies at the
2074 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2075 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2076 * to the socket referenced by *map* (of type
2077 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2078 * egress interfaces can be used for redirection. The
2079 * **BPF_F_INGRESS** value in *flags* is used to make the
2080 * distinction (ingress path is selected if the flag is present,
2081 * egress otherwise). This is the only flag supported for now.
2083 * **SK_PASS** on success, or **SK_DROP** on error.
2085 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2087 * Encapsulate the packet associated to *skb* within a Layer 3
2088 * protocol header. This header is provided in the buffer at
2089 * address *hdr*, with *len* its size in bytes. *type* indicates
2090 * the protocol of the header and can be one of:
2092 * **BPF_LWT_ENCAP_SEG6**
2093 * IPv6 encapsulation with Segment Routing Header
2094 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2095 * the IPv6 header is computed by the kernel.
2096 * **BPF_LWT_ENCAP_SEG6_INLINE**
2097 * Only works if *skb* contains an IPv6 packet. Insert a
2098 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2100 * **BPF_LWT_ENCAP_IP**
2101 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2102 * must be IPv4 or IPv6, followed by zero or more
2103 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2104 * total bytes in all prepended headers. Please note that
2105 * if **skb_is_gso**\ (*skb*) is true, no more than two
2106 * headers can be prepended, and the inner header, if
2107 * present, should be either GRE or UDP/GUE.
2109 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2110 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2111 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2112 * **BPF_PROG_TYPE_LWT_XMIT**.
2114 * A call to this helper is susceptible to change the underlying
2115 * packet buffer. Therefore, at load time, all checks on pointers
2116 * previously done by the verifier are invalidated and must be
2117 * performed again, if the helper is used in combination with
2118 * direct packet access.
2120 * 0 on success, or a negative error in case of failure.
2122 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2124 * Store *len* bytes from address *from* into the packet
2125 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2126 * inside the outermost IPv6 Segment Routing Header can be
2127 * modified through this helper.
2129 * A call to this helper is susceptible to change the underlying
2130 * packet buffer. Therefore, at load time, all checks on pointers
2131 * previously done by the verifier are invalidated and must be
2132 * performed again, if the helper is used in combination with
2133 * direct packet access.
2135 * 0 on success, or a negative error in case of failure.
2137 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2139 * Adjust the size allocated to TLVs in the outermost IPv6
2140 * Segment Routing Header contained in the packet associated to
2141 * *skb*, at position *offset* by *delta* bytes. Only offsets
2142 * after the segments are accepted. *delta* can be as well
2143 * positive (growing) as negative (shrinking).
2145 * A call to this helper is susceptible to change the underlying
2146 * packet buffer. Therefore, at load time, all checks on pointers
2147 * previously done by the verifier are invalidated and must be
2148 * performed again, if the helper is used in combination with
2149 * direct packet access.
2151 * 0 on success, or a negative error in case of failure.
2153 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2155 * Apply an IPv6 Segment Routing action of type *action* to the
2156 * packet associated to *skb*. Each action takes a parameter
2157 * contained at address *param*, and of length *param_len* bytes.
2158 * *action* can be one of:
2160 * **SEG6_LOCAL_ACTION_END_X**
2161 * End.X action: Endpoint with Layer-3 cross-connect.
2162 * Type of *param*: **struct in6_addr**.
2163 * **SEG6_LOCAL_ACTION_END_T**
2164 * End.T action: Endpoint with specific IPv6 table lookup.
2165 * Type of *param*: **int**.
2166 * **SEG6_LOCAL_ACTION_END_B6**
2167 * End.B6 action: Endpoint bound to an SRv6 policy.
2168 * Type of *param*: **struct ipv6_sr_hdr**.
2169 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2170 * End.B6.Encap action: Endpoint bound to an SRv6
2171 * encapsulation policy.
2172 * Type of *param*: **struct ipv6_sr_hdr**.
2174 * A call to this helper is susceptible to change the underlying
2175 * packet buffer. Therefore, at load time, all checks on pointers
2176 * previously done by the verifier are invalidated and must be
2177 * performed again, if the helper is used in combination with
2178 * direct packet access.
2180 * 0 on success, or a negative error in case of failure.
2182 * int bpf_rc_repeat(void *ctx)
2184 * This helper is used in programs implementing IR decoding, to
2185 * report a successfully decoded repeat key message. This delays
2186 * the generation of a key up event for previously generated
2189 * Some IR protocols like NEC have a special IR message for
2190 * repeating last button, for when a button is held down.
2192 * The *ctx* should point to the lirc sample as passed into
2195 * This helper is only available is the kernel was compiled with
2196 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2201 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2203 * This helper is used in programs implementing IR decoding, to
2204 * report a successfully decoded key press with *scancode*,
2205 * *toggle* value in the given *protocol*. The scancode will be
2206 * translated to a keycode using the rc keymap, and reported as
2207 * an input key down event. After a period a key up event is
2208 * generated. This period can be extended by calling either
2209 * **bpf_rc_keydown**\ () again with the same values, or calling
2210 * **bpf_rc_repeat**\ ().
2212 * Some protocols include a toggle bit, in case the button was
2213 * released and pressed again between consecutive scancodes.
2215 * The *ctx* should point to the lirc sample as passed into
2218 * The *protocol* is the decoded protocol number (see
2219 * **enum rc_proto** for some predefined values).
2221 * This helper is only available is the kernel was compiled with
2222 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2227 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2229 * Return the cgroup v2 id of the socket associated with the *skb*.
2230 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2231 * helper for cgroup v1 by providing a tag resp. identifier that
2232 * can be matched on or used for map lookups e.g. to implement
2233 * policy. The cgroup v2 id of a given path in the hierarchy is
2234 * exposed in user space through the f_handle API in order to get
2235 * to the same 64-bit id.
2237 * This helper can be used on TC egress path, but not on ingress,
2238 * and is available only if the kernel was compiled with the
2239 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2241 * The id is returned or 0 in case the id could not be retrieved.
2243 * u64 bpf_get_current_cgroup_id(void)
2245 * A 64-bit integer containing the current cgroup id based
2246 * on the cgroup within which the current task is running.
2248 * void *bpf_get_local_storage(void *map, u64 flags)
2250 * Get the pointer to the local storage area.
2251 * The type and the size of the local storage is defined
2252 * by the *map* argument.
2253 * The *flags* meaning is specific for each map type,
2254 * and has to be 0 for cgroup local storage.
2256 * Depending on the BPF program type, a local storage area
2257 * can be shared between multiple instances of the BPF program,
2258 * running simultaneously.
2260 * A user should care about the synchronization by himself.
2261 * For example, by using the **BPF_STX_XADD** instruction to alter
2264 * A pointer to the local storage area.
2266 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2268 * Select a **SO_REUSEPORT** socket from a
2269 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2270 * It checks the selected socket is matching the incoming
2271 * request in the socket buffer.
2273 * 0 on success, or a negative error in case of failure.
2275 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2277 * Return id of cgroup v2 that is ancestor of cgroup associated
2278 * with the *skb* at the *ancestor_level*. The root cgroup is at
2279 * *ancestor_level* zero and each step down the hierarchy
2280 * increments the level. If *ancestor_level* == level of cgroup
2281 * associated with *skb*, then return value will be same as that
2282 * of **bpf_skb_cgroup_id**\ ().
2284 * The helper is useful to implement policies based on cgroups
2285 * that are upper in hierarchy than immediate cgroup associated
2288 * The format of returned id and helper limitations are same as in
2289 * **bpf_skb_cgroup_id**\ ().
2291 * The id is returned or 0 in case the id could not be retrieved.
2293 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2295 * Look for TCP socket matching *tuple*, optionally in a child
2296 * network namespace *netns*. The return value must be checked,
2297 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2299 * The *ctx* should point to the context of the program, such as
2300 * the skb or socket (depending on the hook in use). This is used
2301 * to determine the base network namespace for the lookup.
2303 * *tuple_size* must be one of:
2305 * **sizeof**\ (*tuple*\ **->ipv4**)
2306 * Look for an IPv4 socket.
2307 * **sizeof**\ (*tuple*\ **->ipv6**)
2308 * Look for an IPv6 socket.
2310 * If the *netns* is a negative signed 32-bit integer, then the
2311 * socket lookup table in the netns associated with the *ctx* will
2312 * will be used. For the TC hooks, this is the netns of the device
2313 * in the skb. For socket hooks, this is the netns of the socket.
2314 * If *netns* is any other signed 32-bit value greater than or
2315 * equal to zero then it specifies the ID of the netns relative to
2316 * the netns associated with the *ctx*. *netns* values beyond the
2317 * range of 32-bit integers are reserved for future use.
2319 * All values for *flags* are reserved for future usage, and must
2322 * This helper is available only if the kernel was compiled with
2323 * **CONFIG_NET** configuration option.
2325 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2326 * For sockets with reuseport option, the **struct bpf_sock**
2327 * result is from *reuse*\ **->socks**\ [] using the hash of the
2330 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2332 * Look for UDP socket matching *tuple*, optionally in a child
2333 * network namespace *netns*. The return value must be checked,
2334 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2336 * The *ctx* should point to the context of the program, such as
2337 * the skb or socket (depending on the hook in use). This is used
2338 * to determine the base network namespace for the lookup.
2340 * *tuple_size* must be one of:
2342 * **sizeof**\ (*tuple*\ **->ipv4**)
2343 * Look for an IPv4 socket.
2344 * **sizeof**\ (*tuple*\ **->ipv6**)
2345 * Look for an IPv6 socket.
2347 * If the *netns* is a negative signed 32-bit integer, then the
2348 * socket lookup table in the netns associated with the *ctx* will
2349 * will be used. For the TC hooks, this is the netns of the device
2350 * in the skb. For socket hooks, this is the netns of the socket.
2351 * If *netns* is any other signed 32-bit value greater than or
2352 * equal to zero then it specifies the ID of the netns relative to
2353 * the netns associated with the *ctx*. *netns* values beyond the
2354 * range of 32-bit integers are reserved for future use.
2356 * All values for *flags* are reserved for future usage, and must
2359 * This helper is available only if the kernel was compiled with
2360 * **CONFIG_NET** configuration option.
2362 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2363 * For sockets with reuseport option, the **struct bpf_sock**
2364 * result is from *reuse*\ **->socks**\ [] using the hash of the
2367 * int bpf_sk_release(struct bpf_sock *sock)
2369 * Release the reference held by *sock*. *sock* must be a
2370 * non-**NULL** pointer that was returned from
2371 * **bpf_sk_lookup_xxx**\ ().
2373 * 0 on success, or a negative error in case of failure.
2375 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2377 * Push an element *value* in *map*. *flags* is one of:
2380 * If the queue/stack is full, the oldest element is
2381 * removed to make room for this.
2383 * 0 on success, or a negative error in case of failure.
2385 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2387 * Pop an element from *map*.
2389 * 0 on success, or a negative error in case of failure.
2391 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2393 * Get an element from *map* without removing it.
2395 * 0 on success, or a negative error in case of failure.
2397 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2399 * For socket policies, insert *len* bytes into *msg* at offset
2402 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2403 * *msg* it may want to insert metadata or options into the *msg*.
2404 * This can later be read and used by any of the lower layer BPF
2407 * This helper may fail if under memory pressure (a malloc
2408 * fails) in these cases BPF programs will get an appropriate
2409 * error and BPF programs will need to handle them.
2411 * 0 on success, or a negative error in case of failure.
2413 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2415 * Will remove *len* bytes from a *msg* starting at byte *start*.
2416 * This may result in **ENOMEM** errors under certain situations if
2417 * an allocation and copy are required due to a full ring buffer.
2418 * However, the helper will try to avoid doing the allocation
2419 * if possible. Other errors can occur if input parameters are
2420 * invalid either due to *start* byte not being valid part of *msg*
2421 * payload and/or *pop* value being to large.
2423 * 0 on success, or a negative error in case of failure.
2425 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2427 * This helper is used in programs implementing IR decoding, to
2428 * report a successfully decoded pointer movement.
2430 * The *ctx* should point to the lirc sample as passed into
2433 * This helper is only available is the kernel was compiled with
2434 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2439 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2441 * Acquire a spinlock represented by the pointer *lock*, which is
2442 * stored as part of a value of a map. Taking the lock allows to
2443 * safely update the rest of the fields in that value. The
2444 * spinlock can (and must) later be released with a call to
2445 * **bpf_spin_unlock**\ (\ *lock*\ ).
2447 * Spinlocks in BPF programs come with a number of restrictions
2450 * * **bpf_spin_lock** objects are only allowed inside maps of
2451 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2452 * list could be extended in the future).
2453 * * BTF description of the map is mandatory.
2454 * * The BPF program can take ONE lock at a time, since taking two
2455 * or more could cause dead locks.
2456 * * Only one **struct bpf_spin_lock** is allowed per map element.
2457 * * When the lock is taken, calls (either BPF to BPF or helpers)
2459 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2460 * allowed inside a spinlock-ed region.
2461 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2462 * the lock, on all execution paths, before it returns.
2463 * * The BPF program can access **struct bpf_spin_lock** only via
2464 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2465 * helpers. Loading or storing data into the **struct
2466 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2467 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2468 * of the map value must be a struct and have **struct
2469 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2470 * Nested lock inside another struct is not allowed.
2471 * * The **struct bpf_spin_lock** *lock* field in a map value must
2472 * be aligned on a multiple of 4 bytes in that value.
2473 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2474 * the **bpf_spin_lock** field to user space.
2475 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2476 * a BPF program, do not update the **bpf_spin_lock** field.
2477 * * **bpf_spin_lock** cannot be on the stack or inside a
2478 * networking packet (it can only be inside of a map values).
2479 * * **bpf_spin_lock** is available to root only.
2480 * * Tracing programs and socket filter programs cannot use
2481 * **bpf_spin_lock**\ () due to insufficient preemption checks
2482 * (but this may change in the future).
2483 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2487 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2489 * Release the *lock* previously locked by a call to
2490 * **bpf_spin_lock**\ (\ *lock*\ ).
2494 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2496 * This helper gets a **struct bpf_sock** pointer such
2497 * that all the fields in this **bpf_sock** can be accessed.
2499 * A **struct bpf_sock** pointer on success, or **NULL** in
2502 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2504 * This helper gets a **struct bpf_tcp_sock** pointer from a
2505 * **struct bpf_sock** pointer.
2507 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2510 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
2512 * Set ECN (Explicit Congestion Notification) field of IP header
2513 * to **CE** (Congestion Encountered) if current value is **ECT**
2514 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2517 * 1 if the **CE** flag is set (either by the current helper call
2518 * or because it was already present), 0 if it is not set.
2520 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2522 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2523 * **bpf_sk_release**\ () is unnecessary and not allowed.
2525 * A **struct bpf_sock** pointer on success, or **NULL** in
2528 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2530 * Look for TCP socket matching *tuple*, optionally in a child
2531 * network namespace *netns*. The return value must be checked,
2532 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2534 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2535 * that it also returns timewait or request sockets. Use
2536 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2539 * This helper is available only if the kernel was compiled with
2540 * **CONFIG_NET** configuration option.
2542 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2543 * For sockets with reuseport option, the **struct bpf_sock**
2544 * result is from *reuse*\ **->socks**\ [] using the hash of the
2547 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2549 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2550 * the listening socket in *sk*.
2552 * *iph* points to the start of the IPv4 or IPv6 header, while
2553 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2554 * **sizeof**\ (**struct ip6hdr**).
2556 * *th* points to the start of the TCP header, while *th_len*
2557 * contains **sizeof**\ (**struct tcphdr**).
2560 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2563 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2565 * Get name of sysctl in /proc/sys/ and copy it into provided by
2566 * program buffer *buf* of size *buf_len*.
2568 * The buffer is always NUL terminated, unless it's zero-sized.
2570 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2571 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2572 * only (e.g. "tcp_mem").
2574 * Number of character copied (not including the trailing NUL).
2576 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2577 * truncated name in this case).
2579 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2581 * Get current value of sysctl as it is presented in /proc/sys
2582 * (incl. newline, etc), and copy it as a string into provided
2583 * by program buffer *buf* of size *buf_len*.
2585 * The whole value is copied, no matter what file position user
2586 * space issued e.g. sys_read at.
2588 * The buffer is always NUL terminated, unless it's zero-sized.
2590 * Number of character copied (not including the trailing NUL).
2592 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2593 * truncated name in this case).
2595 * **-EINVAL** if current value was unavailable, e.g. because
2596 * sysctl is uninitialized and read returns -EIO for it.
2598 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2600 * Get new value being written by user space to sysctl (before
2601 * the actual write happens) and copy it as a string into
2602 * provided by program buffer *buf* of size *buf_len*.
2604 * User space may write new value at file position > 0.
2606 * The buffer is always NUL terminated, unless it's zero-sized.
2608 * Number of character copied (not including the trailing NUL).
2610 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2611 * truncated name in this case).
2613 * **-EINVAL** if sysctl is being read.
2615 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2617 * Override new value being written by user space to sysctl with
2618 * value provided by program in buffer *buf* of size *buf_len*.
2620 * *buf* should contain a string in same form as provided by user
2621 * space on sysctl write.
2623 * User space may write new value at file position > 0. To override
2624 * the whole sysctl value file position should be set to zero.
2628 * **-E2BIG** if the *buf_len* is too big.
2630 * **-EINVAL** if sysctl is being read.
2632 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2634 * Convert the initial part of the string from buffer *buf* of
2635 * size *buf_len* to a long integer according to the given base
2636 * and save the result in *res*.
2638 * The string may begin with an arbitrary amount of white space
2639 * (as determined by **isspace**\ (3)) followed by a single
2640 * optional '**-**' sign.
2642 * Five least significant bits of *flags* encode base, other bits
2643 * are currently unused.
2645 * Base must be either 8, 10, 16 or 0 to detect it automatically
2646 * similar to user space **strtol**\ (3).
2648 * Number of characters consumed on success. Must be positive but
2649 * no more than *buf_len*.
2651 * **-EINVAL** if no valid digits were found or unsupported base
2654 * **-ERANGE** if resulting value was out of range.
2656 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2658 * Convert the initial part of the string from buffer *buf* of
2659 * size *buf_len* to an unsigned long integer according to the
2660 * given base and save the result in *res*.
2662 * The string may begin with an arbitrary amount of white space
2663 * (as determined by **isspace**\ (3)).
2665 * Five least significant bits of *flags* encode base, other bits
2666 * are currently unused.
2668 * Base must be either 8, 10, 16 or 0 to detect it automatically
2669 * similar to user space **strtoul**\ (3).
2671 * Number of characters consumed on success. Must be positive but
2672 * no more than *buf_len*.
2674 * **-EINVAL** if no valid digits were found or unsupported base
2677 * **-ERANGE** if resulting value was out of range.
2679 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2681 * Get a bpf-local-storage from a *sk*.
2683 * Logically, it could be thought of getting the value from
2684 * a *map* with *sk* as the **key**. From this
2685 * perspective, the usage is not much different from
2686 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2687 * helper enforces the key must be a full socket and the map must
2688 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2690 * Underneath, the value is stored locally at *sk* instead of
2691 * the *map*. The *map* is used as the bpf-local-storage
2692 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2693 * searched against all bpf-local-storages residing at *sk*.
2695 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2696 * used such that a new bpf-local-storage will be
2697 * created if one does not exist. *value* can be used
2698 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2699 * the initial value of a bpf-local-storage. If *value* is
2700 * **NULL**, the new bpf-local-storage will be zero initialized.
2702 * A bpf-local-storage pointer is returned on success.
2704 * **NULL** if not found or there was an error in adding
2705 * a new bpf-local-storage.
2707 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2709 * Delete a bpf-local-storage from a *sk*.
2713 * **-ENOENT** if the bpf-local-storage cannot be found.
2715 * int bpf_send_signal(u32 sig)
2717 * Send signal *sig* to the current task.
2719 * 0 on success or successfully queued.
2721 * **-EBUSY** if work queue under nmi is full.
2723 * **-EINVAL** if *sig* is invalid.
2725 * **-EPERM** if no permission to send the *sig*.
2727 * **-EAGAIN** if bpf program can try again.
2729 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2731 * Try to issue a SYN cookie for the packet with corresponding
2732 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2734 * *iph* points to the start of the IPv4 or IPv6 header, while
2735 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2736 * **sizeof**\ (**struct ip6hdr**).
2738 * *th* points to the start of the TCP header, while *th_len*
2739 * contains the length of the TCP header.
2742 * On success, lower 32 bits hold the generated SYN cookie in
2743 * followed by 16 bits which hold the MSS value for that cookie,
2744 * and the top 16 bits are unused.
2746 * On failure, the returned value is one of the following:
2748 * **-EINVAL** SYN cookie cannot be issued due to error
2750 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2752 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2754 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2756 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2758 * Write raw *data* blob into a special BPF perf event held by
2759 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2760 * event must have the following attributes: **PERF_SAMPLE_RAW**
2761 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2762 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2764 * The *flags* are used to indicate the index in *map* for which
2765 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2766 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2767 * to indicate that the index of the current CPU core should be
2770 * The value to write, of *size*, is passed through eBPF stack and
2771 * pointed by *data*.
2773 * *ctx* is a pointer to in-kernel struct sk_buff.
2775 * This helper is similar to **bpf_perf_event_output**\ () but
2776 * restricted to raw_tracepoint bpf programs.
2778 * 0 on success, or a negative error in case of failure.
2780 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2782 * Safely attempt to read *size* bytes from user space address
2783 * *unsafe_ptr* and store the data in *dst*.
2785 * 0 on success, or a negative error in case of failure.
2787 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2789 * Safely attempt to read *size* bytes from kernel space address
2790 * *unsafe_ptr* and store the data in *dst*.
2792 * 0 on success, or a negative error in case of failure.
2794 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2796 * Copy a NUL terminated string from an unsafe user address
2797 * *unsafe_ptr* to *dst*. The *size* should include the
2798 * terminating NUL byte. In case the string length is smaller than
2799 * *size*, the target is not padded with further NUL bytes. If the
2800 * string length is larger than *size*, just *size*-1 bytes are
2801 * copied and the last byte is set to NUL.
2803 * On success, the length of the copied string is returned. This
2804 * makes this helper useful in tracing programs for reading
2805 * strings, and more importantly to get its length at runtime. See
2806 * the following snippet:
2810 * SEC("kprobe/sys_open")
2811 * void bpf_sys_open(struct pt_regs *ctx)
2813 * char buf[PATHLEN]; // PATHLEN is defined to 256
2814 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2817 * // Consume buf, for example push it to
2818 * // userspace via bpf_perf_event_output(); we
2819 * // can use res (the string length) as event
2820 * // size, after checking its boundaries.
2823 * In comparison, using **bpf_probe_read_user()** helper here
2824 * instead to read the string would require to estimate the length
2825 * at compile time, and would often result in copying more memory
2828 * Another useful use case is when parsing individual process
2829 * arguments or individual environment variables navigating
2830 * *current*\ **->mm->arg_start** and *current*\
2831 * **->mm->env_start**: using this helper and the return value,
2832 * one can quickly iterate at the right offset of the memory area.
2834 * On success, the strictly positive length of the string,
2835 * including the trailing NUL character. On error, a negative
2838 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2840 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
2841 * to *dst*. Same semantics as with bpf_probe_read_user_str() apply.
2843 * On success, the strictly positive length of the string, including
2844 * the trailing NUL character. On error, a negative value.
2846 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2848 * Send out a tcp-ack. *tp* is the in-kernel struct tcp_sock.
2849 * *rcv_nxt* is the ack_seq to be sent out.
2851 * 0 on success, or a negative error in case of failure.
2854 #define __BPF_FUNC_MAPPER(FN) \
2856 FN(map_lookup_elem), \
2857 FN(map_update_elem), \
2858 FN(map_delete_elem), \
2862 FN(get_prandom_u32), \
2863 FN(get_smp_processor_id), \
2864 FN(skb_store_bytes), \
2865 FN(l3_csum_replace), \
2866 FN(l4_csum_replace), \
2868 FN(clone_redirect), \
2869 FN(get_current_pid_tgid), \
2870 FN(get_current_uid_gid), \
2871 FN(get_current_comm), \
2872 FN(get_cgroup_classid), \
2873 FN(skb_vlan_push), \
2875 FN(skb_get_tunnel_key), \
2876 FN(skb_set_tunnel_key), \
2877 FN(perf_event_read), \
2879 FN(get_route_realm), \
2880 FN(perf_event_output), \
2881 FN(skb_load_bytes), \
2884 FN(skb_get_tunnel_opt), \
2885 FN(skb_set_tunnel_opt), \
2886 FN(skb_change_proto), \
2887 FN(skb_change_type), \
2888 FN(skb_under_cgroup), \
2889 FN(get_hash_recalc), \
2890 FN(get_current_task), \
2891 FN(probe_write_user), \
2892 FN(current_task_under_cgroup), \
2893 FN(skb_change_tail), \
2894 FN(skb_pull_data), \
2896 FN(set_hash_invalid), \
2897 FN(get_numa_node_id), \
2898 FN(skb_change_head), \
2899 FN(xdp_adjust_head), \
2900 FN(probe_read_str), \
2901 FN(get_socket_cookie), \
2902 FN(get_socket_uid), \
2905 FN(skb_adjust_room), \
2907 FN(sk_redirect_map), \
2908 FN(sock_map_update), \
2909 FN(xdp_adjust_meta), \
2910 FN(perf_event_read_value), \
2911 FN(perf_prog_read_value), \
2913 FN(override_return), \
2914 FN(sock_ops_cb_flags_set), \
2915 FN(msg_redirect_map), \
2916 FN(msg_apply_bytes), \
2917 FN(msg_cork_bytes), \
2918 FN(msg_pull_data), \
2920 FN(xdp_adjust_tail), \
2921 FN(skb_get_xfrm_state), \
2923 FN(skb_load_bytes_relative), \
2925 FN(sock_hash_update), \
2926 FN(msg_redirect_hash), \
2927 FN(sk_redirect_hash), \
2928 FN(lwt_push_encap), \
2929 FN(lwt_seg6_store_bytes), \
2930 FN(lwt_seg6_adjust_srh), \
2931 FN(lwt_seg6_action), \
2934 FN(skb_cgroup_id), \
2935 FN(get_current_cgroup_id), \
2936 FN(get_local_storage), \
2937 FN(sk_select_reuseport), \
2938 FN(skb_ancestor_cgroup_id), \
2939 FN(sk_lookup_tcp), \
2940 FN(sk_lookup_udp), \
2942 FN(map_push_elem), \
2944 FN(map_peek_elem), \
2945 FN(msg_push_data), \
2947 FN(rc_pointer_rel), \
2952 FN(skb_ecn_set_ce), \
2953 FN(get_listener_sock), \
2954 FN(skc_lookup_tcp), \
2955 FN(tcp_check_syncookie), \
2956 FN(sysctl_get_name), \
2957 FN(sysctl_get_current_value), \
2958 FN(sysctl_get_new_value), \
2959 FN(sysctl_set_new_value), \
2962 FN(sk_storage_get), \
2963 FN(sk_storage_delete), \
2965 FN(tcp_gen_syncookie), \
2967 FN(probe_read_user), \
2968 FN(probe_read_kernel), \
2969 FN(probe_read_user_str), \
2970 FN(probe_read_kernel_str), \
2973 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2974 * function eBPF program intends to call
2976 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2978 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2981 #undef __BPF_ENUM_FN
2983 /* All flags used by eBPF helper functions, placed here. */
2985 /* BPF_FUNC_skb_store_bytes flags. */
2986 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2987 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2989 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2990 * First 4 bits are for passing the header field size.
2992 #define BPF_F_HDR_FIELD_MASK 0xfULL
2994 /* BPF_FUNC_l4_csum_replace flags. */
2995 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2996 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2997 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2999 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
3000 #define BPF_F_INGRESS (1ULL << 0)
3002 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
3003 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
3005 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
3006 #define BPF_F_SKIP_FIELD_MASK 0xffULL
3007 #define BPF_F_USER_STACK (1ULL << 8)
3008 /* flags used by BPF_FUNC_get_stackid only. */
3009 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
3010 #define BPF_F_REUSE_STACKID (1ULL << 10)
3011 /* flags used by BPF_FUNC_get_stack only. */
3012 #define BPF_F_USER_BUILD_ID (1ULL << 11)
3014 /* BPF_FUNC_skb_set_tunnel_key flags. */
3015 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
3016 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
3017 #define BPF_F_SEQ_NUMBER (1ULL << 3)
3019 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3020 * BPF_FUNC_perf_event_read_value flags.
3022 #define BPF_F_INDEX_MASK 0xffffffffULL
3023 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
3024 /* BPF_FUNC_perf_event_output for sk_buff input context. */
3025 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
3027 /* Current network namespace */
3028 #define BPF_F_CURRENT_NETNS (-1L)
3030 /* BPF_FUNC_skb_adjust_room flags. */
3031 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
3033 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
3034 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
3036 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
3037 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
3038 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
3039 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
3040 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
3041 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3042 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
3044 /* BPF_FUNC_sysctl_get_name flags. */
3045 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
3047 /* BPF_FUNC_sk_storage_get flags */
3048 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0)
3050 /* Mode for BPF_FUNC_skb_adjust_room helper. */
3051 enum bpf_adj_room_mode {
3056 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3057 enum bpf_hdr_start_off {
3062 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3063 enum bpf_lwt_encap_mode {
3065 BPF_LWT_ENCAP_SEG6_INLINE,
3069 #define __bpf_md_ptr(type, name) \
3073 } __attribute__((aligned(8)))
3075 /* user accessible mirror of in-kernel sk_buff.
3076 * new fields can only be added to the end of this structure
3082 __u32 queue_mapping;
3088 __u32 ingress_ifindex;
3098 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
3100 __u32 remote_ip4; /* Stored in network byte order */
3101 __u32 local_ip4; /* Stored in network byte order */
3102 __u32 remote_ip6[4]; /* Stored in network byte order */
3103 __u32 local_ip6[4]; /* Stored in network byte order */
3104 __u32 remote_port; /* Stored in network byte order */
3105 __u32 local_port; /* stored in host byte order */
3109 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
3113 __bpf_md_ptr(struct bpf_sock *, sk);
3116 struct bpf_tunnel_key {
3120 __u32 remote_ipv6[4];
3124 __u16 tunnel_ext; /* Padding, future use. */
3128 /* user accessible mirror of in-kernel xfrm_state.
3129 * new fields can only be added to the end of this structure
3131 struct bpf_xfrm_state {
3133 __u32 spi; /* Stored in network byte order */
3135 __u16 ext; /* Padding, future use. */
3137 __u32 remote_ipv4; /* Stored in network byte order */
3138 __u32 remote_ipv6[4]; /* Stored in network byte order */
3142 /* Generic BPF return codes which all BPF program types may support.
3143 * The values are binary compatible with their TC_ACT_* counter-part to
3144 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3147 * XDP is handled seprately, see XDP_*.
3155 /* >127 are reserved for prog type specific return codes.
3157 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3158 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3159 * changed and should be routed based on its new L3 header.
3160 * (This is an L3 redirect, as opposed to L2 redirect
3161 * represented by BPF_REDIRECT above).
3163 BPF_LWT_REROUTE = 128,
3173 /* IP address also allows 1 and 2 bytes access */
3176 __u32 src_port; /* host byte order */
3177 __u32 dst_port; /* network byte order */
3183 struct bpf_tcp_sock {
3184 __u32 snd_cwnd; /* Sending congestion window */
3185 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3187 __u32 snd_ssthresh; /* Slow start size threshold */
3188 __u32 rcv_nxt; /* What we want to receive next */
3189 __u32 snd_nxt; /* Next sequence we send */
3190 __u32 snd_una; /* First byte we want an ack for */
3191 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3192 __u32 ecn_flags; /* ECN status bits. */
3193 __u32 rate_delivered; /* saved rate sample: packets delivered */
3194 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3195 __u32 packets_out; /* Packets which are "in flight" */
3196 __u32 retrans_out; /* Retransmitted packets out */
3197 __u32 total_retrans; /* Total retransmits for entire connection */
3198 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3199 * total number of segments in.
3201 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3202 * total number of data segments in.
3204 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3205 * The total number of segments sent.
3207 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3208 * total number of data segments sent.
3210 __u32 lost_out; /* Lost packets */
3211 __u32 sacked_out; /* SACK'd packets */
3212 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3213 * sum(delta(rcv_nxt)), or how many bytes
3216 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3217 * sum(delta(snd_una)), or how many bytes
3220 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3221 * total number of DSACK blocks received
3223 __u32 delivered; /* Total data packets delivered incl. rexmits */
3224 __u32 delivered_ce; /* Like the above but only ECE marked packets */
3225 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
3228 struct bpf_sock_tuple {
3245 struct bpf_xdp_sock {
3249 #define XDP_PACKET_HEADROOM 256
3251 /* User return codes for XDP prog type.
3252 * A valid XDP program must return one of these defined values. All other
3253 * return codes are reserved for future use. Unknown return codes will
3254 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3264 /* user accessible metadata for XDP packet hook
3265 * new fields must be added to the end of this structure
3271 /* Below access go through struct xdp_rxq_info */
3272 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3273 __u32 rx_queue_index; /* rxq->queue_index */
3281 /* user accessible metadata for SK_MSG packet hook, new fields must
3282 * be added to the end of this structure
3285 __bpf_md_ptr(void *, data);
3286 __bpf_md_ptr(void *, data_end);
3289 __u32 remote_ip4; /* Stored in network byte order */
3290 __u32 local_ip4; /* Stored in network byte order */
3291 __u32 remote_ip6[4]; /* Stored in network byte order */
3292 __u32 local_ip6[4]; /* Stored in network byte order */
3293 __u32 remote_port; /* Stored in network byte order */
3294 __u32 local_port; /* stored in host byte order */
3295 __u32 size; /* Total size of sk_msg */
3298 struct sk_reuseport_md {
3300 * Start of directly accessible data. It begins from
3301 * the tcp/udp header.
3303 __bpf_md_ptr(void *, data);
3304 /* End of directly accessible data */
3305 __bpf_md_ptr(void *, data_end);
3307 * Total length of packet (starting from the tcp/udp header).
3308 * Note that the directly accessible bytes (data_end - data)
3309 * could be less than this "len". Those bytes could be
3310 * indirectly read by a helper "bpf_skb_load_bytes()".
3314 * Eth protocol in the mac header (network byte order). e.g.
3315 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3318 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3319 __u32 bind_inany; /* Is sock bound to an INANY address? */
3320 __u32 hash; /* A hash of the packet 4 tuples */
3323 #define BPF_TAG_SIZE 8
3325 struct bpf_prog_info {
3328 __u8 tag[BPF_TAG_SIZE];
3329 __u32 jited_prog_len;
3330 __u32 xlated_prog_len;
3331 __aligned_u64 jited_prog_insns;
3332 __aligned_u64 xlated_prog_insns;
3333 __u64 load_time; /* ns since boottime */
3334 __u32 created_by_uid;
3336 __aligned_u64 map_ids;
3337 char name[BPF_OBJ_NAME_LEN];
3339 __u32 gpl_compatible:1;
3340 __u32 :31; /* alignment pad */
3343 __u32 nr_jited_ksyms;
3344 __u32 nr_jited_func_lens;
3345 __aligned_u64 jited_ksyms;
3346 __aligned_u64 jited_func_lens;
3348 __u32 func_info_rec_size;
3349 __aligned_u64 func_info;
3352 __aligned_u64 line_info;
3353 __aligned_u64 jited_line_info;
3354 __u32 nr_jited_line_info;
3355 __u32 line_info_rec_size;
3356 __u32 jited_line_info_rec_size;
3358 __aligned_u64 prog_tags;
3361 } __attribute__((aligned(8)));
3363 struct bpf_map_info {
3370 char name[BPF_OBJ_NAME_LEN];
3372 __u32 btf_vmlinux_value_type_id;
3376 __u32 btf_key_type_id;
3377 __u32 btf_value_type_id;
3378 } __attribute__((aligned(8)));
3380 struct bpf_btf_info {
3384 } __attribute__((aligned(8)));
3386 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3387 * by user and intended to be used by socket (e.g. to bind to, depends on
3388 * attach attach type).
3390 struct bpf_sock_addr {
3391 __u32 user_family; /* Allows 4-byte read, but no write. */
3392 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3393 * Stored in network byte order.
3395 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3396 * Stored in network byte order.
3398 __u32 user_port; /* Allows 4-byte read and write.
3399 * Stored in network byte order
3401 __u32 family; /* Allows 4-byte read, but no write */
3402 __u32 type; /* Allows 4-byte read, but no write */
3403 __u32 protocol; /* Allows 4-byte read, but no write */
3404 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3405 * Stored in network byte order.
3407 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3408 * Stored in network byte order.
3410 __bpf_md_ptr(struct bpf_sock *, sk);
3413 /* User bpf_sock_ops struct to access socket values and specify request ops
3414 * and their replies.
3415 * Some of this fields are in network (bigendian) byte order and may need
3416 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3417 * New fields can only be added at the end of this structure
3419 struct bpf_sock_ops {
3422 __u32 args[4]; /* Optionally passed to bpf program */
3423 __u32 reply; /* Returned by bpf program */
3424 __u32 replylong[4]; /* Optionally returned by bpf prog */
3427 __u32 remote_ip4; /* Stored in network byte order */
3428 __u32 local_ip4; /* Stored in network byte order */
3429 __u32 remote_ip6[4]; /* Stored in network byte order */
3430 __u32 local_ip6[4]; /* Stored in network byte order */
3431 __u32 remote_port; /* Stored in network byte order */
3432 __u32 local_port; /* stored in host byte order */
3433 __u32 is_fullsock; /* Some TCP fields are only valid if
3434 * there is a full socket. If not, the
3435 * fields read as zero.
3438 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
3439 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3448 __u32 rate_delivered;
3449 __u32 rate_interval_us;
3452 __u32 total_retrans;
3456 __u32 data_segs_out;
3460 __u64 bytes_received;
3462 __bpf_md_ptr(struct bpf_sock *, sk);
3465 /* Definitions for bpf_sock_ops_cb_flags */
3466 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
3467 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
3468 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3469 #define BPF_SOCK_OPS_RTT_CB_FLAG (1<<3)
3470 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0xF /* Mask of all currently
3471 * supported cb flags
3474 /* List of known BPF sock_ops operators.
3475 * New entries can only be added at the end
3479 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3480 * -1 if default value should be used
3482 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3483 * window (in packets) or -1 if default
3484 * value should be used
3486 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3487 * active connection is initialized
3489 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3490 * active connection is
3493 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3494 * passive connection is
3497 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3500 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3501 * based on the path and may be
3502 * dependent on the congestion control
3503 * algorithm. In general it indicates
3504 * a congestion threshold. RTTs above
3505 * this indicate congestion
3507 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3508 * Arg1: value of icsk_retransmits
3509 * Arg2: value of icsk_rto
3510 * Arg3: whether RTO has expired
3512 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3513 * Arg1: sequence number of 1st byte
3515 * Arg3: return value of
3516 * tcp_transmit_skb (0 => success)
3518 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3522 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3523 * socket transition to LISTEN state.
3525 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3529 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3530 * changes between the TCP and BPF versions. Ideally this should never happen.
3531 * If it does, we need to add code to convert them before calling
3532 * the BPF sock_ops function.
3535 BPF_TCP_ESTABLISHED = 1,
3545 BPF_TCP_CLOSING, /* Now a valid state */
3546 BPF_TCP_NEW_SYN_RECV,
3548 BPF_TCP_MAX_STATES /* Leave at the end! */
3551 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3552 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3554 struct bpf_perf_event_value {
3560 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3561 #define BPF_DEVCG_ACC_READ (1ULL << 1)
3562 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3564 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3565 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3567 struct bpf_cgroup_dev_ctx {
3568 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3574 struct bpf_raw_tracepoint_args {
3578 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3579 * OUTPUT: Do lookup from egress perspective; default is ingress
3581 #define BPF_FIB_LOOKUP_DIRECT (1U << 0)
3582 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
3585 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3586 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3587 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3588 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3589 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3590 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3591 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3592 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3593 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3596 struct bpf_fib_lookup {
3597 /* input: network family for lookup (AF_INET, AF_INET6)
3598 * output: network family of egress nexthop
3602 /* set if lookup is to consider L4 data - e.g., FIB rules */
3607 /* total length of packet from network header - used for MTU check */
3610 /* input: L3 device index for lookup
3611 * output: device index from FIB lookup
3616 /* inputs to lookup */
3617 __u8 tos; /* AF_INET */
3618 __be32 flowinfo; /* AF_INET6, flow_label + priority */
3620 /* output: metric of fib result (IPv4/IPv6 only) */
3626 __u32 ipv6_src[4]; /* in6_addr; network order */
3629 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3630 * network header. output: bpf_fib_lookup sets to gateway address
3631 * if FIB lookup returns gateway route
3635 __u32 ipv6_dst[4]; /* in6_addr; network order */
3639 __be16 h_vlan_proto;
3641 __u8 smac[6]; /* ETH_ALEN */
3642 __u8 dmac[6]; /* ETH_ALEN */
3645 enum bpf_task_fd_type {
3646 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3647 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3648 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3649 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3650 BPF_FD_TYPE_UPROBE, /* filename + offset */
3651 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3654 #define BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG (1U << 0)
3655 #define BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL (1U << 1)
3656 #define BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP (1U << 2)
3658 struct bpf_flow_keys {
3661 __u16 addr_proto; /* ETH_P_* of valid addrs */
3675 __u32 ipv6_src[4]; /* in6_addr; network order */
3676 __u32 ipv6_dst[4]; /* in6_addr; network order */
3683 struct bpf_func_info {
3688 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3689 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3691 struct bpf_line_info {
3693 __u32 file_name_off;
3698 struct bpf_spin_lock {
3703 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3704 * Allows 1,2,4-byte read, but no write.
3706 __u32 file_pos; /* Sysctl file position to read from, write to.
3707 * Allows 1,2,4-byte read an 4-byte write.
3711 struct bpf_sockopt {
3712 __bpf_md_ptr(struct bpf_sock *, sk);
3713 __bpf_md_ptr(void *, optval);
3714 __bpf_md_ptr(void *, optval_end);
3722 #endif /* _UAPI__LINUX_BPF_H__ */