1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_XADD 0xc0 /* exclusive add */
25 #define BPF_MOV 0xb0 /* mov reg to reg */
26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28 /* change endianness of a register */
29 #define BPF_END 0xd0 /* flags for endianness conversion: */
30 #define BPF_TO_LE 0x00 /* convert to little-endian */
31 #define BPF_TO_BE 0x08 /* convert to big-endian */
32 #define BPF_FROM_LE BPF_TO_LE
33 #define BPF_FROM_BE BPF_TO_BE
36 #define BPF_JNE 0x50 /* jump != */
37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
43 #define BPF_CALL 0x80 /* function call */
44 #define BPF_EXIT 0x90 /* function return */
46 /* Register numbers */
62 /* BPF has 10 general purpose 64-bit registers and stack frame. */
63 #define MAX_BPF_REG __MAX_BPF_REG
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74 struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data[0]; /* Arbitrary size */
79 struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
84 /* BPF syscall commands, see bpf(2) man-page for details. */
99 BPF_PROG_GET_FD_BY_ID,
100 BPF_MAP_GET_FD_BY_ID,
101 BPF_OBJ_GET_INFO_BY_FD,
103 BPF_RAW_TRACEPOINT_OPEN,
105 BPF_BTF_GET_FD_BY_ID,
107 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
115 BPF_MAP_TYPE_PROG_ARRAY,
116 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
117 BPF_MAP_TYPE_PERCPU_HASH,
118 BPF_MAP_TYPE_PERCPU_ARRAY,
119 BPF_MAP_TYPE_STACK_TRACE,
120 BPF_MAP_TYPE_CGROUP_ARRAY,
121 BPF_MAP_TYPE_LRU_HASH,
122 BPF_MAP_TYPE_LRU_PERCPU_HASH,
123 BPF_MAP_TYPE_LPM_TRIE,
124 BPF_MAP_TYPE_ARRAY_OF_MAPS,
125 BPF_MAP_TYPE_HASH_OF_MAPS,
127 BPF_MAP_TYPE_SOCKMAP,
130 BPF_MAP_TYPE_SOCKHASH,
131 BPF_MAP_TYPE_CGROUP_STORAGE,
132 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
133 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
136 BPF_MAP_TYPE_SK_STORAGE,
139 /* Note that tracing related programs such as
140 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
141 * are not subject to a stable API since kernel internal data
142 * structures can change from release to release and may
143 * therefore break existing tracing BPF programs. Tracing BPF
144 * programs correspond to /a/ specific kernel which is to be
145 * analyzed, and not /a/ specific kernel /and/ all future ones.
148 BPF_PROG_TYPE_UNSPEC,
149 BPF_PROG_TYPE_SOCKET_FILTER,
150 BPF_PROG_TYPE_KPROBE,
151 BPF_PROG_TYPE_SCHED_CLS,
152 BPF_PROG_TYPE_SCHED_ACT,
153 BPF_PROG_TYPE_TRACEPOINT,
155 BPF_PROG_TYPE_PERF_EVENT,
156 BPF_PROG_TYPE_CGROUP_SKB,
157 BPF_PROG_TYPE_CGROUP_SOCK,
158 BPF_PROG_TYPE_LWT_IN,
159 BPF_PROG_TYPE_LWT_OUT,
160 BPF_PROG_TYPE_LWT_XMIT,
161 BPF_PROG_TYPE_SOCK_OPS,
162 BPF_PROG_TYPE_SK_SKB,
163 BPF_PROG_TYPE_CGROUP_DEVICE,
164 BPF_PROG_TYPE_SK_MSG,
165 BPF_PROG_TYPE_RAW_TRACEPOINT,
166 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
167 BPF_PROG_TYPE_LWT_SEG6LOCAL,
168 BPF_PROG_TYPE_LIRC_MODE2,
169 BPF_PROG_TYPE_SK_REUSEPORT,
170 BPF_PROG_TYPE_FLOW_DISSECTOR,
171 BPF_PROG_TYPE_CGROUP_SYSCTL,
172 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
173 BPF_PROG_TYPE_CGROUP_SOCKOPT,
176 enum bpf_attach_type {
177 BPF_CGROUP_INET_INGRESS,
178 BPF_CGROUP_INET_EGRESS,
179 BPF_CGROUP_INET_SOCK_CREATE,
181 BPF_SK_SKB_STREAM_PARSER,
182 BPF_SK_SKB_STREAM_VERDICT,
185 BPF_CGROUP_INET4_BIND,
186 BPF_CGROUP_INET6_BIND,
187 BPF_CGROUP_INET4_CONNECT,
188 BPF_CGROUP_INET6_CONNECT,
189 BPF_CGROUP_INET4_POST_BIND,
190 BPF_CGROUP_INET6_POST_BIND,
191 BPF_CGROUP_UDP4_SENDMSG,
192 BPF_CGROUP_UDP6_SENDMSG,
196 BPF_CGROUP_UDP4_RECVMSG,
197 BPF_CGROUP_UDP6_RECVMSG,
198 BPF_CGROUP_GETSOCKOPT,
199 BPF_CGROUP_SETSOCKOPT,
200 __MAX_BPF_ATTACH_TYPE
203 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
205 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
207 * NONE(default): No further bpf programs allowed in the subtree.
209 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
210 * the program in this cgroup yields to sub-cgroup program.
212 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
213 * that cgroup program gets run in addition to the program in this cgroup.
215 * Only one program is allowed to be attached to a cgroup with
216 * NONE or BPF_F_ALLOW_OVERRIDE flag.
217 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
218 * release old program and attach the new one. Attach flags has to match.
220 * Multiple programs are allowed to be attached to a cgroup with
221 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
222 * (those that were attached first, run first)
223 * The programs of sub-cgroup are executed first, then programs of
224 * this cgroup and then programs of parent cgroup.
225 * When children program makes decision (like picking TCP CA or sock bind)
226 * parent program has a chance to override it.
228 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
229 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
231 * cgrp1 (MULTI progs A, B) ->
232 * cgrp2 (OVERRIDE prog C) ->
233 * cgrp3 (MULTI prog D) ->
234 * cgrp4 (OVERRIDE prog E) ->
235 * cgrp5 (NONE prog F)
236 * the event in cgrp5 triggers execution of F,D,A,B in that order.
237 * if prog F is detached, the execution is E,D,A,B
238 * if prog F and D are detached, the execution is E,A,B
239 * if prog F, E and D are detached, the execution is C,A,B
241 * All eligible programs are executed regardless of return code from
244 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
245 #define BPF_F_ALLOW_MULTI (1U << 1)
247 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
248 * verifier will perform strict alignment checking as if the kernel
249 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
250 * and NET_IP_ALIGN defined to 2.
252 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
254 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
255 * verifier will allow any alignment whatsoever. On platforms
256 * with strict alignment requirements for loads ands stores (such
257 * as sparc and mips) the verifier validates that all loads and
258 * stores provably follow this requirement. This flag turns that
259 * checking and enforcement off.
261 * It is mostly used for testing when we want to validate the
262 * context and memory access aspects of the verifier, but because
263 * of an unaligned access the alignment check would trigger before
264 * the one we are interested in.
266 #define BPF_F_ANY_ALIGNMENT (1U << 1)
268 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
269 * Verifier does sub-register def/use analysis and identifies instructions whose
270 * def only matters for low 32-bit, high 32-bit is never referenced later
271 * through implicit zero extension. Therefore verifier notifies JIT back-ends
272 * that it is safe to ignore clearing high 32-bit for these instructions. This
273 * saves some back-ends a lot of code-gen. However such optimization is not
274 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
275 * hence hasn't used verifier's analysis result. But, we really want to have a
276 * way to be able to verify the correctness of the described optimization on
277 * x86_64 on which testsuites are frequently exercised.
279 * So, this flag is introduced. Once it is set, verifier will randomize high
280 * 32-bit for those instructions who has been identified as safe to ignore them.
281 * Then, if verifier is not doing correct analysis, such randomization will
282 * regress tests to expose bugs.
284 #define BPF_F_TEST_RND_HI32 (1U << 2)
286 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
289 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
290 * insn[0].imm: map fd map fd
291 * insn[1].imm: 0 offset into value
294 * ldimm64 rewrite: address of map address of map[0]+offset
295 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
297 #define BPF_PSEUDO_MAP_FD 1
298 #define BPF_PSEUDO_MAP_VALUE 2
300 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
301 * offset to another bpf function
303 #define BPF_PSEUDO_CALL 1
305 /* flags for BPF_MAP_UPDATE_ELEM command */
306 #define BPF_ANY 0 /* create new element or update existing */
307 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
308 #define BPF_EXIST 2 /* update existing element */
309 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
311 /* flags for BPF_MAP_CREATE command */
312 #define BPF_F_NO_PREALLOC (1U << 0)
313 /* Instead of having one common LRU list in the
314 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
315 * which can scale and perform better.
316 * Note, the LRU nodes (including free nodes) cannot be moved
317 * across different LRU lists.
319 #define BPF_F_NO_COMMON_LRU (1U << 1)
320 /* Specify numa node during map creation */
321 #define BPF_F_NUMA_NODE (1U << 2)
323 #define BPF_OBJ_NAME_LEN 16U
325 /* Flags for accessing BPF object from syscall side. */
326 #define BPF_F_RDONLY (1U << 3)
327 #define BPF_F_WRONLY (1U << 4)
329 /* Flag for stack_map, store build_id+offset instead of pointer */
330 #define BPF_F_STACK_BUILD_ID (1U << 5)
332 /* Zero-initialize hash function seed. This should only be used for testing. */
333 #define BPF_F_ZERO_SEED (1U << 6)
335 /* Flags for accessing BPF object from program side. */
336 #define BPF_F_RDONLY_PROG (1U << 7)
337 #define BPF_F_WRONLY_PROG (1U << 8)
339 /* flags for BPF_PROG_QUERY */
340 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
342 enum bpf_stack_build_id_status {
343 /* user space need an empty entry to identify end of a trace */
344 BPF_STACK_BUILD_ID_EMPTY = 0,
345 /* with valid build_id and offset */
346 BPF_STACK_BUILD_ID_VALID = 1,
347 /* couldn't get build_id, fallback to ip */
348 BPF_STACK_BUILD_ID_IP = 2,
351 #define BPF_BUILD_ID_SIZE 20
352 struct bpf_stack_build_id {
354 unsigned char build_id[BPF_BUILD_ID_SIZE];
362 struct { /* anonymous struct used by BPF_MAP_CREATE command */
363 __u32 map_type; /* one of enum bpf_map_type */
364 __u32 key_size; /* size of key in bytes */
365 __u32 value_size; /* size of value in bytes */
366 __u32 max_entries; /* max number of entries in a map */
367 __u32 map_flags; /* BPF_MAP_CREATE related
368 * flags defined above.
370 __u32 inner_map_fd; /* fd pointing to the inner map */
371 __u32 numa_node; /* numa node (effective only if
372 * BPF_F_NUMA_NODE is set).
374 char map_name[BPF_OBJ_NAME_LEN];
375 __u32 map_ifindex; /* ifindex of netdev to create on */
376 __u32 btf_fd; /* fd pointing to a BTF type data */
377 __u32 btf_key_type_id; /* BTF type_id of the key */
378 __u32 btf_value_type_id; /* BTF type_id of the value */
381 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
386 __aligned_u64 next_key;
391 struct { /* anonymous struct used by BPF_PROG_LOAD command */
392 __u32 prog_type; /* one of enum bpf_prog_type */
395 __aligned_u64 license;
396 __u32 log_level; /* verbosity level of verifier */
397 __u32 log_size; /* size of user buffer */
398 __aligned_u64 log_buf; /* user supplied buffer */
399 __u32 kern_version; /* not used */
401 char prog_name[BPF_OBJ_NAME_LEN];
402 __u32 prog_ifindex; /* ifindex of netdev to prep for */
403 /* For some prog types expected attach type must be known at
404 * load time to verify attach type specific parts of prog
405 * (context accesses, allowed helpers, etc).
407 __u32 expected_attach_type;
408 __u32 prog_btf_fd; /* fd pointing to BTF type data */
409 __u32 func_info_rec_size; /* userspace bpf_func_info size */
410 __aligned_u64 func_info; /* func info */
411 __u32 func_info_cnt; /* number of bpf_func_info records */
412 __u32 line_info_rec_size; /* userspace bpf_line_info size */
413 __aligned_u64 line_info; /* line info */
414 __u32 line_info_cnt; /* number of bpf_line_info records */
417 struct { /* anonymous struct used by BPF_OBJ_* commands */
418 __aligned_u64 pathname;
423 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
424 __u32 target_fd; /* container object to attach to */
425 __u32 attach_bpf_fd; /* eBPF program to attach */
430 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
433 __u32 data_size_in; /* input: len of data_in */
434 __u32 data_size_out; /* input/output: len of data_out
435 * returns ENOSPC if data_out
438 __aligned_u64 data_in;
439 __aligned_u64 data_out;
442 __u32 ctx_size_in; /* input: len of ctx_in */
443 __u32 ctx_size_out; /* input/output: len of ctx_out
444 * returns ENOSPC if ctx_out
447 __aligned_u64 ctx_in;
448 __aligned_u64 ctx_out;
451 struct { /* anonymous struct used by BPF_*_GET_*_ID */
462 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
468 struct { /* anonymous struct used by BPF_PROG_QUERY command */
469 __u32 target_fd; /* container object to query */
473 __aligned_u64 prog_ids;
482 struct { /* anonymous struct for BPF_BTF_LOAD */
484 __aligned_u64 btf_log_buf;
491 __u32 pid; /* input: pid */
492 __u32 fd; /* input: fd */
493 __u32 flags; /* input: flags */
494 __u32 buf_len; /* input/output: buf len */
495 __aligned_u64 buf; /* input/output:
496 * tp_name for tracepoint
498 * filename for uprobe
500 __u32 prog_id; /* output: prod_id */
501 __u32 fd_type; /* output: BPF_FD_TYPE_* */
502 __u64 probe_offset; /* output: probe_offset */
503 __u64 probe_addr; /* output: probe_addr */
505 } __attribute__((aligned(8)));
507 /* The description below is an attempt at providing documentation to eBPF
508 * developers about the multiple available eBPF helper functions. It can be
509 * parsed and used to produce a manual page. The workflow is the following,
510 * and requires the rst2man utility:
512 * $ ./scripts/bpf_helpers_doc.py \
513 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
514 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
515 * $ man /tmp/bpf-helpers.7
517 * Note that in order to produce this external documentation, some RST
518 * formatting is used in the descriptions to get "bold" and "italics" in
519 * manual pages. Also note that the few trailing white spaces are
520 * intentional, removing them would break paragraphs for rst2man.
522 * Start of BPF helper function descriptions:
524 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
526 * Perform a lookup in *map* for an entry associated to *key*.
528 * Map value associated to *key*, or **NULL** if no entry was
531 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
533 * Add or update the value of the entry associated to *key* in
534 * *map* with *value*. *flags* is one of:
537 * The entry for *key* must not exist in the map.
539 * The entry for *key* must already exist in the map.
541 * No condition on the existence of the entry for *key*.
543 * Flag value **BPF_NOEXIST** cannot be used for maps of types
544 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
545 * elements always exist), the helper would return an error.
547 * 0 on success, or a negative error in case of failure.
549 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
551 * Delete entry with *key* from *map*.
553 * 0 on success, or a negative error in case of failure.
555 * int bpf_probe_read(void *dst, u32 size, const void *src)
557 * For tracing programs, safely attempt to read *size* bytes from
558 * address *src* and store the data in *dst*.
560 * 0 on success, or a negative error in case of failure.
562 * u64 bpf_ktime_get_ns(void)
564 * Return the time elapsed since system boot, in nanoseconds.
568 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
570 * This helper is a "printk()-like" facility for debugging. It
571 * prints a message defined by format *fmt* (of size *fmt_size*)
572 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
573 * available. It can take up to three additional **u64**
574 * arguments (as an eBPF helpers, the total number of arguments is
577 * Each time the helper is called, it appends a line to the trace.
578 * The format of the trace is customizable, and the exact output
579 * one will get depends on the options set in
580 * *\/sys/kernel/debug/tracing/trace_options* (see also the
581 * *README* file under the same directory). However, it usually
582 * defaults to something like:
586 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
590 * * ``telnet`` is the name of the current task.
591 * * ``470`` is the PID of the current task.
592 * * ``001`` is the CPU number on which the task is
594 * * In ``.N..``, each character refers to a set of
595 * options (whether irqs are enabled, scheduling
596 * options, whether hard/softirqs are running, level of
597 * preempt_disabled respectively). **N** means that
598 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
600 * * ``419421.045894`` is a timestamp.
601 * * ``0x00000001`` is a fake value used by BPF for the
602 * instruction pointer register.
603 * * ``<formatted msg>`` is the message formatted with
606 * The conversion specifiers supported by *fmt* are similar, but
607 * more limited than for printk(). They are **%d**, **%i**,
608 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
609 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
610 * of field, padding with zeroes, etc.) is available, and the
611 * helper will return **-EINVAL** (but print nothing) if it
612 * encounters an unknown specifier.
614 * Also, note that **bpf_trace_printk**\ () is slow, and should
615 * only be used for debugging purposes. For this reason, a notice
616 * bloc (spanning several lines) is printed to kernel logs and
617 * states that the helper should not be used "for production use"
618 * the first time this helper is used (or more precisely, when
619 * **trace_printk**\ () buffers are allocated). For passing values
620 * to user space, perf events should be preferred.
622 * The number of bytes written to the buffer, or a negative error
623 * in case of failure.
625 * u32 bpf_get_prandom_u32(void)
627 * Get a pseudo-random number.
629 * From a security point of view, this helper uses its own
630 * pseudo-random internal state, and cannot be used to infer the
631 * seed of other random functions in the kernel. However, it is
632 * essential to note that the generator used by the helper is not
633 * cryptographically secure.
635 * A random 32-bit unsigned value.
637 * u32 bpf_get_smp_processor_id(void)
639 * Get the SMP (symmetric multiprocessing) processor id. Note that
640 * all programs run with preemption disabled, which means that the
641 * SMP processor id is stable during all the execution of the
644 * The SMP id of the processor running the program.
646 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
648 * Store *len* bytes from address *from* into the packet
649 * associated to *skb*, at *offset*. *flags* are a combination of
650 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
651 * checksum for the packet after storing the bytes) and
652 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
653 * **->swhash** and *skb*\ **->l4hash** to 0).
655 * A call to this helper is susceptible to change the underlying
656 * packet buffer. Therefore, at load time, all checks on pointers
657 * previously done by the verifier are invalidated and must be
658 * performed again, if the helper is used in combination with
659 * direct packet access.
661 * 0 on success, or a negative error in case of failure.
663 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
665 * Recompute the layer 3 (e.g. IP) checksum for the packet
666 * associated to *skb*. Computation is incremental, so the helper
667 * must know the former value of the header field that was
668 * modified (*from*), the new value of this field (*to*), and the
669 * number of bytes (2 or 4) for this field, stored in *size*.
670 * Alternatively, it is possible to store the difference between
671 * the previous and the new values of the header field in *to*, by
672 * setting *from* and *size* to 0. For both methods, *offset*
673 * indicates the location of the IP checksum within the packet.
675 * This helper works in combination with **bpf_csum_diff**\ (),
676 * which does not update the checksum in-place, but offers more
677 * flexibility and can handle sizes larger than 2 or 4 for the
678 * checksum to update.
680 * A call to this helper is susceptible to change the underlying
681 * packet buffer. Therefore, at load time, all checks on pointers
682 * previously done by the verifier are invalidated and must be
683 * performed again, if the helper is used in combination with
684 * direct packet access.
686 * 0 on success, or a negative error in case of failure.
688 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
690 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
691 * packet associated to *skb*. Computation is incremental, so the
692 * helper must know the former value of the header field that was
693 * modified (*from*), the new value of this field (*to*), and the
694 * number of bytes (2 or 4) for this field, stored on the lowest
695 * four bits of *flags*. Alternatively, it is possible to store
696 * the difference between the previous and the new values of the
697 * header field in *to*, by setting *from* and the four lowest
698 * bits of *flags* to 0. For both methods, *offset* indicates the
699 * location of the IP checksum within the packet. In addition to
700 * the size of the field, *flags* can be added (bitwise OR) actual
701 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
702 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
703 * for updates resulting in a null checksum the value is set to
704 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
705 * the checksum is to be computed against a pseudo-header.
707 * This helper works in combination with **bpf_csum_diff**\ (),
708 * which does not update the checksum in-place, but offers more
709 * flexibility and can handle sizes larger than 2 or 4 for the
710 * checksum to update.
712 * A call to this helper is susceptible to change the underlying
713 * packet buffer. Therefore, at load time, all checks on pointers
714 * previously done by the verifier are invalidated and must be
715 * performed again, if the helper is used in combination with
716 * direct packet access.
718 * 0 on success, or a negative error in case of failure.
720 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
722 * This special helper is used to trigger a "tail call", or in
723 * other words, to jump into another eBPF program. The same stack
724 * frame is used (but values on stack and in registers for the
725 * caller are not accessible to the callee). This mechanism allows
726 * for program chaining, either for raising the maximum number of
727 * available eBPF instructions, or to execute given programs in
728 * conditional blocks. For security reasons, there is an upper
729 * limit to the number of successive tail calls that can be
732 * Upon call of this helper, the program attempts to jump into a
733 * program referenced at index *index* in *prog_array_map*, a
734 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
735 * *ctx*, a pointer to the context.
737 * If the call succeeds, the kernel immediately runs the first
738 * instruction of the new program. This is not a function call,
739 * and it never returns to the previous program. If the call
740 * fails, then the helper has no effect, and the caller continues
741 * to run its subsequent instructions. A call can fail if the
742 * destination program for the jump does not exist (i.e. *index*
743 * is superior to the number of entries in *prog_array_map*), or
744 * if the maximum number of tail calls has been reached for this
745 * chain of programs. This limit is defined in the kernel by the
746 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
747 * which is currently set to 32.
749 * 0 on success, or a negative error in case of failure.
751 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
753 * Clone and redirect the packet associated to *skb* to another
754 * net device of index *ifindex*. Both ingress and egress
755 * interfaces can be used for redirection. The **BPF_F_INGRESS**
756 * value in *flags* is used to make the distinction (ingress path
757 * is selected if the flag is present, egress path otherwise).
758 * This is the only flag supported for now.
760 * In comparison with **bpf_redirect**\ () helper,
761 * **bpf_clone_redirect**\ () has the associated cost of
762 * duplicating the packet buffer, but this can be executed out of
763 * the eBPF program. Conversely, **bpf_redirect**\ () is more
764 * efficient, but it is handled through an action code where the
765 * redirection happens only after the eBPF program has returned.
767 * A call to this helper is susceptible to change the underlying
768 * packet buffer. Therefore, at load time, all checks on pointers
769 * previously done by the verifier are invalidated and must be
770 * performed again, if the helper is used in combination with
771 * direct packet access.
773 * 0 on success, or a negative error in case of failure.
775 * u64 bpf_get_current_pid_tgid(void)
777 * A 64-bit integer containing the current tgid and pid, and
779 * *current_task*\ **->tgid << 32 \|**
780 * *current_task*\ **->pid**.
782 * u64 bpf_get_current_uid_gid(void)
784 * A 64-bit integer containing the current GID and UID, and
785 * created as such: *current_gid* **<< 32 \|** *current_uid*.
787 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
789 * Copy the **comm** attribute of the current task into *buf* of
790 * *size_of_buf*. The **comm** attribute contains the name of
791 * the executable (excluding the path) for the current task. The
792 * *size_of_buf* must be strictly positive. On success, the
793 * helper makes sure that the *buf* is NUL-terminated. On failure,
794 * it is filled with zeroes.
796 * 0 on success, or a negative error in case of failure.
798 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
800 * Retrieve the classid for the current task, i.e. for the net_cls
801 * cgroup to which *skb* belongs.
803 * This helper can be used on TC egress path, but not on ingress.
805 * The net_cls cgroup provides an interface to tag network packets
806 * based on a user-provided identifier for all traffic coming from
807 * the tasks belonging to the related cgroup. See also the related
808 * kernel documentation, available from the Linux sources in file
809 * *Documentation/cgroup-v1/net_cls.txt*.
811 * The Linux kernel has two versions for cgroups: there are
812 * cgroups v1 and cgroups v2. Both are available to users, who can
813 * use a mixture of them, but note that the net_cls cgroup is for
814 * cgroup v1 only. This makes it incompatible with BPF programs
815 * run on cgroups, which is a cgroup-v2-only feature (a socket can
816 * only hold data for one version of cgroups at a time).
818 * This helper is only available is the kernel was compiled with
819 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
820 * "**y**" or to "**m**".
822 * The classid, or 0 for the default unconfigured classid.
824 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
826 * Push a *vlan_tci* (VLAN tag control information) of protocol
827 * *vlan_proto* to the packet associated to *skb*, then update
828 * the checksum. Note that if *vlan_proto* is different from
829 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
830 * be **ETH_P_8021Q**.
832 * A call to this helper is susceptible to change the underlying
833 * packet buffer. Therefore, at load time, all checks on pointers
834 * previously done by the verifier are invalidated and must be
835 * performed again, if the helper is used in combination with
836 * direct packet access.
838 * 0 on success, or a negative error in case of failure.
840 * int bpf_skb_vlan_pop(struct sk_buff *skb)
842 * Pop a VLAN header from the packet associated to *skb*.
844 * A call to this helper is susceptible to change the underlying
845 * packet buffer. Therefore, at load time, all checks on pointers
846 * previously done by the verifier are invalidated and must be
847 * performed again, if the helper is used in combination with
848 * direct packet access.
850 * 0 on success, or a negative error in case of failure.
852 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
854 * Get tunnel metadata. This helper takes a pointer *key* to an
855 * empty **struct bpf_tunnel_key** of **size**, that will be
856 * filled with tunnel metadata for the packet associated to *skb*.
857 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
858 * indicates that the tunnel is based on IPv6 protocol instead of
861 * The **struct bpf_tunnel_key** is an object that generalizes the
862 * principal parameters used by various tunneling protocols into a
863 * single struct. This way, it can be used to easily make a
864 * decision based on the contents of the encapsulation header,
865 * "summarized" in this struct. In particular, it holds the IP
866 * address of the remote end (IPv4 or IPv6, depending on the case)
867 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
868 * this struct exposes the *key*\ **->tunnel_id**, which is
869 * generally mapped to a VNI (Virtual Network Identifier), making
870 * it programmable together with the **bpf_skb_set_tunnel_key**\
873 * Let's imagine that the following code is part of a program
874 * attached to the TC ingress interface, on one end of a GRE
875 * tunnel, and is supposed to filter out all messages coming from
876 * remote ends with IPv4 address other than 10.0.0.1:
881 * struct bpf_tunnel_key key = {};
883 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
885 * return TC_ACT_SHOT; // drop packet
887 * if (key.remote_ipv4 != 0x0a000001)
888 * return TC_ACT_SHOT; // drop packet
890 * return TC_ACT_OK; // accept packet
892 * This interface can also be used with all encapsulation devices
893 * that can operate in "collect metadata" mode: instead of having
894 * one network device per specific configuration, the "collect
895 * metadata" mode only requires a single device where the
896 * configuration can be extracted from this helper.
898 * This can be used together with various tunnels such as VXLan,
899 * Geneve, GRE or IP in IP (IPIP).
901 * 0 on success, or a negative error in case of failure.
903 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
905 * Populate tunnel metadata for packet associated to *skb.* The
906 * tunnel metadata is set to the contents of *key*, of *size*. The
907 * *flags* can be set to a combination of the following values:
909 * **BPF_F_TUNINFO_IPV6**
910 * Indicate that the tunnel is based on IPv6 protocol
912 * **BPF_F_ZERO_CSUM_TX**
913 * For IPv4 packets, add a flag to tunnel metadata
914 * indicating that checksum computation should be skipped
915 * and checksum set to zeroes.
916 * **BPF_F_DONT_FRAGMENT**
917 * Add a flag to tunnel metadata indicating that the
918 * packet should not be fragmented.
919 * **BPF_F_SEQ_NUMBER**
920 * Add a flag to tunnel metadata indicating that a
921 * sequence number should be added to tunnel header before
922 * sending the packet. This flag was added for GRE
923 * encapsulation, but might be used with other protocols
924 * as well in the future.
926 * Here is a typical usage on the transmit path:
930 * struct bpf_tunnel_key key;
932 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
933 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
935 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
936 * helper for additional information.
938 * 0 on success, or a negative error in case of failure.
940 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
942 * Read the value of a perf event counter. This helper relies on a
943 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
944 * the perf event counter is selected when *map* is updated with
945 * perf event file descriptors. The *map* is an array whose size
946 * is the number of available CPUs, and each cell contains a value
947 * relative to one CPU. The value to retrieve is indicated by
948 * *flags*, that contains the index of the CPU to look up, masked
949 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
950 * **BPF_F_CURRENT_CPU** to indicate that the value for the
951 * current CPU should be retrieved.
953 * Note that before Linux 4.13, only hardware perf event can be
956 * Also, be aware that the newer helper
957 * **bpf_perf_event_read_value**\ () is recommended over
958 * **bpf_perf_event_read**\ () in general. The latter has some ABI
959 * quirks where error and counter value are used as a return code
960 * (which is wrong to do since ranges may overlap). This issue is
961 * fixed with **bpf_perf_event_read_value**\ (), which at the same
962 * time provides more features over the **bpf_perf_event_read**\
963 * () interface. Please refer to the description of
964 * **bpf_perf_event_read_value**\ () for details.
966 * The value of the perf event counter read from the map, or a
967 * negative error code in case of failure.
969 * int bpf_redirect(u32 ifindex, u64 flags)
971 * Redirect the packet to another net device of index *ifindex*.
972 * This helper is somewhat similar to **bpf_clone_redirect**\
973 * (), except that the packet is not cloned, which provides
974 * increased performance.
976 * Except for XDP, both ingress and egress interfaces can be used
977 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
978 * to make the distinction (ingress path is selected if the flag
979 * is present, egress path otherwise). Currently, XDP only
980 * supports redirection to the egress interface, and accepts no
983 * The same effect can be attained with the more generic
984 * **bpf_redirect_map**\ (), which requires specific maps to be
985 * used but offers better performance.
987 * For XDP, the helper returns **XDP_REDIRECT** on success or
988 * **XDP_ABORTED** on error. For other program types, the values
989 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
992 * u32 bpf_get_route_realm(struct sk_buff *skb)
994 * Retrieve the realm or the route, that is to say the
995 * **tclassid** field of the destination for the *skb*. The
996 * indentifier retrieved is a user-provided tag, similar to the
997 * one used with the net_cls cgroup (see description for
998 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
999 * held by a route (a destination entry), not by a task.
1001 * Retrieving this identifier works with the clsact TC egress hook
1002 * (see also **tc-bpf(8)**), or alternatively on conventional
1003 * classful egress qdiscs, but not on TC ingress path. In case of
1004 * clsact TC egress hook, this has the advantage that, internally,
1005 * the destination entry has not been dropped yet in the transmit
1006 * path. Therefore, the destination entry does not need to be
1007 * artificially held via **netif_keep_dst**\ () for a classful
1008 * qdisc until the *skb* is freed.
1010 * This helper is available only if the kernel was compiled with
1011 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1013 * The realm of the route for the packet associated to *skb*, or 0
1014 * if none was found.
1016 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1018 * Write raw *data* blob into a special BPF perf event held by
1019 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1020 * event must have the following attributes: **PERF_SAMPLE_RAW**
1021 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1022 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1024 * The *flags* are used to indicate the index in *map* for which
1025 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1026 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1027 * to indicate that the index of the current CPU core should be
1030 * The value to write, of *size*, is passed through eBPF stack and
1031 * pointed by *data*.
1033 * The context of the program *ctx* needs also be passed to the
1036 * On user space, a program willing to read the values needs to
1037 * call **perf_event_open**\ () on the perf event (either for
1038 * one or for all CPUs) and to store the file descriptor into the
1039 * *map*. This must be done before the eBPF program can send data
1040 * into it. An example is available in file
1041 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1042 * tree (the eBPF program counterpart is in
1043 * *samples/bpf/trace_output_kern.c*).
1045 * **bpf_perf_event_output**\ () achieves better performance
1046 * than **bpf_trace_printk**\ () for sharing data with user
1047 * space, and is much better suitable for streaming data from eBPF
1050 * Note that this helper is not restricted to tracing use cases
1051 * and can be used with programs attached to TC or XDP as well,
1052 * where it allows for passing data to user space listeners. Data
1055 * * Only custom structs,
1056 * * Only the packet payload, or
1057 * * A combination of both.
1059 * 0 on success, or a negative error in case of failure.
1061 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1063 * This helper was provided as an easy way to load data from a
1064 * packet. It can be used to load *len* bytes from *offset* from
1065 * the packet associated to *skb*, into the buffer pointed by
1068 * Since Linux 4.7, usage of this helper has mostly been replaced
1069 * by "direct packet access", enabling packet data to be
1070 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1071 * pointing respectively to the first byte of packet data and to
1072 * the byte after the last byte of packet data. However, it
1073 * remains useful if one wishes to read large quantities of data
1074 * at once from a packet into the eBPF stack.
1076 * 0 on success, or a negative error in case of failure.
1078 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1080 * Walk a user or a kernel stack and return its id. To achieve
1081 * this, the helper needs *ctx*, which is a pointer to the context
1082 * on which the tracing program is executed, and a pointer to a
1083 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1085 * The last argument, *flags*, holds the number of stack frames to
1086 * skip (from 0 to 255), masked with
1087 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1088 * a combination of the following flags:
1090 * **BPF_F_USER_STACK**
1091 * Collect a user space stack instead of a kernel stack.
1092 * **BPF_F_FAST_STACK_CMP**
1093 * Compare stacks by hash only.
1094 * **BPF_F_REUSE_STACKID**
1095 * If two different stacks hash into the same *stackid*,
1096 * discard the old one.
1098 * The stack id retrieved is a 32 bit long integer handle which
1099 * can be further combined with other data (including other stack
1100 * ids) and used as a key into maps. This can be useful for
1101 * generating a variety of graphs (such as flame graphs or off-cpu
1104 * For walking a stack, this helper is an improvement over
1105 * **bpf_probe_read**\ (), which can be used with unrolled loops
1106 * but is not efficient and consumes a lot of eBPF instructions.
1107 * Instead, **bpf_get_stackid**\ () can collect up to
1108 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1109 * this limit can be controlled with the **sysctl** program, and
1110 * that it should be manually increased in order to profile long
1111 * user stacks (such as stacks for Java programs). To do so, use:
1115 * # sysctl kernel.perf_event_max_stack=<new value>
1117 * The positive or null stack id on success, or a negative error
1118 * in case of failure.
1120 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1122 * Compute a checksum difference, from the raw buffer pointed by
1123 * *from*, of length *from_size* (that must be a multiple of 4),
1124 * towards the raw buffer pointed by *to*, of size *to_size*
1125 * (same remark). An optional *seed* can be added to the value
1126 * (this can be cascaded, the seed may come from a previous call
1129 * This is flexible enough to be used in several ways:
1131 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1132 * checksum, it can be used when pushing new data.
1133 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1134 * checksum, it can be used when removing data from a packet.
1135 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1136 * can be used to compute a diff. Note that *from_size* and
1137 * *to_size* do not need to be equal.
1139 * This helper can be used in combination with
1140 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1141 * which one can feed in the difference computed with
1142 * **bpf_csum_diff**\ ().
1144 * The checksum result, or a negative error code in case of
1147 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1149 * Retrieve tunnel options metadata for the packet associated to
1150 * *skb*, and store the raw tunnel option data to the buffer *opt*
1153 * This helper can be used with encapsulation devices that can
1154 * operate in "collect metadata" mode (please refer to the related
1155 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1156 * more details). A particular example where this can be used is
1157 * in combination with the Geneve encapsulation protocol, where it
1158 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1159 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1160 * the eBPF program. This allows for full customization of these
1163 * The size of the option data retrieved.
1165 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1167 * Set tunnel options metadata for the packet associated to *skb*
1168 * to the option data contained in the raw buffer *opt* of *size*.
1170 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1171 * helper for additional information.
1173 * 0 on success, or a negative error in case of failure.
1175 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1177 * Change the protocol of the *skb* to *proto*. Currently
1178 * supported are transition from IPv4 to IPv6, and from IPv6 to
1179 * IPv4. The helper takes care of the groundwork for the
1180 * transition, including resizing the socket buffer. The eBPF
1181 * program is expected to fill the new headers, if any, via
1182 * **skb_store_bytes**\ () and to recompute the checksums with
1183 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1184 * (). The main case for this helper is to perform NAT64
1185 * operations out of an eBPF program.
1187 * Internally, the GSO type is marked as dodgy so that headers are
1188 * checked and segments are recalculated by the GSO/GRO engine.
1189 * The size for GSO target is adapted as well.
1191 * All values for *flags* are reserved for future usage, and must
1194 * A call to this helper is susceptible to change the underlying
1195 * packet buffer. Therefore, at load time, all checks on pointers
1196 * previously done by the verifier are invalidated and must be
1197 * performed again, if the helper is used in combination with
1198 * direct packet access.
1200 * 0 on success, or a negative error in case of failure.
1202 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1204 * Change the packet type for the packet associated to *skb*. This
1205 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1206 * the eBPF program does not have a write access to *skb*\
1207 * **->pkt_type** beside this helper. Using a helper here allows
1208 * for graceful handling of errors.
1210 * The major use case is to change incoming *skb*s to
1211 * **PACKET_HOST** in a programmatic way instead of having to
1212 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1215 * Note that *type* only allows certain values. At this time, they
1220 * **PACKET_BROADCAST**
1221 * Send packet to all.
1222 * **PACKET_MULTICAST**
1223 * Send packet to group.
1224 * **PACKET_OTHERHOST**
1225 * Send packet to someone else.
1227 * 0 on success, or a negative error in case of failure.
1229 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1231 * Check whether *skb* is a descendant of the cgroup2 held by
1232 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1234 * The return value depends on the result of the test, and can be:
1236 * * 0, if the *skb* failed the cgroup2 descendant test.
1237 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1238 * * A negative error code, if an error occurred.
1240 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1242 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1243 * not set, in particular if the hash was cleared due to mangling,
1244 * recompute this hash. Later accesses to the hash can be done
1245 * directly with *skb*\ **->hash**.
1247 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1248 * prototype with **bpf_skb_change_proto**\ (), or calling
1249 * **bpf_skb_store_bytes**\ () with the
1250 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1251 * the hash and to trigger a new computation for the next call to
1252 * **bpf_get_hash_recalc**\ ().
1256 * u64 bpf_get_current_task(void)
1258 * A pointer to the current task struct.
1260 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1262 * Attempt in a safe way to write *len* bytes from the buffer
1263 * *src* to *dst* in memory. It only works for threads that are in
1264 * user context, and *dst* must be a valid user space address.
1266 * This helper should not be used to implement any kind of
1267 * security mechanism because of TOC-TOU attacks, but rather to
1268 * debug, divert, and manipulate execution of semi-cooperative
1271 * Keep in mind that this feature is meant for experiments, and it
1272 * has a risk of crashing the system and running programs.
1273 * Therefore, when an eBPF program using this helper is attached,
1274 * a warning including PID and process name is printed to kernel
1277 * 0 on success, or a negative error in case of failure.
1279 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1281 * Check whether the probe is being run is the context of a given
1282 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1283 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1285 * The return value depends on the result of the test, and can be:
1287 * * 0, if the *skb* task belongs to the cgroup2.
1288 * * 1, if the *skb* task does not belong to the cgroup2.
1289 * * A negative error code, if an error occurred.
1291 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1293 * Resize (trim or grow) the packet associated to *skb* to the
1294 * new *len*. The *flags* are reserved for future usage, and must
1297 * The basic idea is that the helper performs the needed work to
1298 * change the size of the packet, then the eBPF program rewrites
1299 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1300 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1301 * and others. This helper is a slow path utility intended for
1302 * replies with control messages. And because it is targeted for
1303 * slow path, the helper itself can afford to be slow: it
1304 * implicitly linearizes, unclones and drops offloads from the
1307 * A call to this helper is susceptible to change the underlying
1308 * packet buffer. Therefore, at load time, all checks on pointers
1309 * previously done by the verifier are invalidated and must be
1310 * performed again, if the helper is used in combination with
1311 * direct packet access.
1313 * 0 on success, or a negative error in case of failure.
1315 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1317 * Pull in non-linear data in case the *skb* is non-linear and not
1318 * all of *len* are part of the linear section. Make *len* bytes
1319 * from *skb* readable and writable. If a zero value is passed for
1320 * *len*, then the whole length of the *skb* is pulled.
1322 * This helper is only needed for reading and writing with direct
1325 * For direct packet access, testing that offsets to access
1326 * are within packet boundaries (test on *skb*\ **->data_end**) is
1327 * susceptible to fail if offsets are invalid, or if the requested
1328 * data is in non-linear parts of the *skb*. On failure the
1329 * program can just bail out, or in the case of a non-linear
1330 * buffer, use a helper to make the data available. The
1331 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1332 * the data. Another one consists in using **bpf_skb_pull_data**
1333 * to pull in once the non-linear parts, then retesting and
1334 * eventually access the data.
1336 * At the same time, this also makes sure the *skb* is uncloned,
1337 * which is a necessary condition for direct write. As this needs
1338 * to be an invariant for the write part only, the verifier
1339 * detects writes and adds a prologue that is calling
1340 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1341 * the very beginning in case it is indeed cloned.
1343 * A call to this helper is susceptible to change the underlying
1344 * packet buffer. Therefore, at load time, all checks on pointers
1345 * previously done by the verifier are invalidated and must be
1346 * performed again, if the helper is used in combination with
1347 * direct packet access.
1349 * 0 on success, or a negative error in case of failure.
1351 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1353 * Add the checksum *csum* into *skb*\ **->csum** in case the
1354 * driver has supplied a checksum for the entire packet into that
1355 * field. Return an error otherwise. This helper is intended to be
1356 * used in combination with **bpf_csum_diff**\ (), in particular
1357 * when the checksum needs to be updated after data has been
1358 * written into the packet through direct packet access.
1360 * The checksum on success, or a negative error code in case of
1363 * void bpf_set_hash_invalid(struct sk_buff *skb)
1365 * Invalidate the current *skb*\ **->hash**. It can be used after
1366 * mangling on headers through direct packet access, in order to
1367 * indicate that the hash is outdated and to trigger a
1368 * recalculation the next time the kernel tries to access this
1369 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1371 * int bpf_get_numa_node_id(void)
1373 * Return the id of the current NUMA node. The primary use case
1374 * for this helper is the selection of sockets for the local NUMA
1375 * node, when the program is attached to sockets using the
1376 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1377 * but the helper is also available to other eBPF program types,
1378 * similarly to **bpf_get_smp_processor_id**\ ().
1380 * The id of current NUMA node.
1382 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1384 * Grows headroom of packet associated to *skb* and adjusts the
1385 * offset of the MAC header accordingly, adding *len* bytes of
1386 * space. It automatically extends and reallocates memory as
1389 * This helper can be used on a layer 3 *skb* to push a MAC header
1390 * for redirection into a layer 2 device.
1392 * All values for *flags* are reserved for future usage, and must
1395 * A call to this helper is susceptible to change the underlying
1396 * packet buffer. Therefore, at load time, all checks on pointers
1397 * previously done by the verifier are invalidated and must be
1398 * performed again, if the helper is used in combination with
1399 * direct packet access.
1401 * 0 on success, or a negative error in case of failure.
1403 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1405 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1406 * it is possible to use a negative value for *delta*. This helper
1407 * can be used to prepare the packet for pushing or popping
1410 * A call to this helper is susceptible to change the underlying
1411 * packet buffer. Therefore, at load time, all checks on pointers
1412 * previously done by the verifier are invalidated and must be
1413 * performed again, if the helper is used in combination with
1414 * direct packet access.
1416 * 0 on success, or a negative error in case of failure.
1418 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1420 * Copy a NUL terminated string from an unsafe address
1421 * *unsafe_ptr* to *dst*. The *size* should include the
1422 * terminating NUL byte. In case the string length is smaller than
1423 * *size*, the target is not padded with further NUL bytes. If the
1424 * string length is larger than *size*, just *size*-1 bytes are
1425 * copied and the last byte is set to NUL.
1427 * On success, the length of the copied string is returned. This
1428 * makes this helper useful in tracing programs for reading
1429 * strings, and more importantly to get its length at runtime. See
1430 * the following snippet:
1434 * SEC("kprobe/sys_open")
1435 * void bpf_sys_open(struct pt_regs *ctx)
1437 * char buf[PATHLEN]; // PATHLEN is defined to 256
1438 * int res = bpf_probe_read_str(buf, sizeof(buf),
1441 * // Consume buf, for example push it to
1442 * // userspace via bpf_perf_event_output(); we
1443 * // can use res (the string length) as event
1444 * // size, after checking its boundaries.
1447 * In comparison, using **bpf_probe_read()** helper here instead
1448 * to read the string would require to estimate the length at
1449 * compile time, and would often result in copying more memory
1452 * Another useful use case is when parsing individual process
1453 * arguments or individual environment variables navigating
1454 * *current*\ **->mm->arg_start** and *current*\
1455 * **->mm->env_start**: using this helper and the return value,
1456 * one can quickly iterate at the right offset of the memory area.
1458 * On success, the strictly positive length of the string,
1459 * including the trailing NUL character. On error, a negative
1462 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1464 * If the **struct sk_buff** pointed by *skb* has a known socket,
1465 * retrieve the cookie (generated by the kernel) of this socket.
1466 * If no cookie has been set yet, generate a new cookie. Once
1467 * generated, the socket cookie remains stable for the life of the
1468 * socket. This helper can be useful for monitoring per socket
1469 * networking traffic statistics as it provides a unique socket
1470 * identifier per namespace.
1472 * A 8-byte long non-decreasing number on success, or 0 if the
1473 * socket field is missing inside *skb*.
1475 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1477 * Equivalent to bpf_get_socket_cookie() helper that accepts
1478 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1480 * A 8-byte long non-decreasing number.
1482 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1484 * Equivalent to bpf_get_socket_cookie() helper that accepts
1485 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1487 * A 8-byte long non-decreasing number.
1489 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1491 * The owner UID of the socket associated to *skb*. If the socket
1492 * is **NULL**, or if it is not a full socket (i.e. if it is a
1493 * time-wait or a request socket instead), **overflowuid** value
1494 * is returned (note that **overflowuid** might also be the actual
1495 * UID value for the socket).
1497 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1499 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1504 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1506 * Emulate a call to **setsockopt()** on the socket associated to
1507 * *bpf_socket*, which must be a full socket. The *level* at
1508 * which the option resides and the name *optname* of the option
1509 * must be specified, see **setsockopt(2)** for more information.
1510 * The option value of length *optlen* is pointed by *optval*.
1512 * This helper actually implements a subset of **setsockopt()**.
1513 * It supports the following *level*\ s:
1515 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1516 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1517 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1518 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1519 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1520 * **TCP_BPF_SNDCWND_CLAMP**.
1521 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1522 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1524 * 0 on success, or a negative error in case of failure.
1526 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1528 * Grow or shrink the room for data in the packet associated to
1529 * *skb* by *len_diff*, and according to the selected *mode*.
1531 * There are two supported modes at this time:
1533 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1534 * (room space is added or removed below the layer 2 header).
1536 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1537 * (room space is added or removed below the layer 3 header).
1539 * The following flags are supported at this time:
1541 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1542 * Adjusting mss in this way is not allowed for datagrams.
1544 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1545 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1546 * Any new space is reserved to hold a tunnel header.
1547 * Configure skb offsets and other fields accordingly.
1549 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1550 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1551 * Use with ENCAP_L3 flags to further specify the tunnel type.
1553 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1554 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1555 * type; *len* is the length of the inner MAC header.
1557 * A call to this helper is susceptible to change the underlying
1558 * packet buffer. Therefore, at load time, all checks on pointers
1559 * previously done by the verifier are invalidated and must be
1560 * performed again, if the helper is used in combination with
1561 * direct packet access.
1563 * 0 on success, or a negative error in case of failure.
1565 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1567 * Redirect the packet to the endpoint referenced by *map* at
1568 * index *key*. Depending on its type, this *map* can contain
1569 * references to net devices (for forwarding packets through other
1570 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1571 * but this is only implemented for native XDP (with driver
1572 * support) as of this writing).
1574 * The lower two bits of *flags* are used as the return code if
1575 * the map lookup fails. This is so that the return value can be
1576 * one of the XDP program return codes up to XDP_TX, as chosen by
1577 * the caller. Any higher bits in the *flags* argument must be
1580 * When used to redirect packets to net devices, this helper
1581 * provides a high performance increase over **bpf_redirect**\ ().
1582 * This is due to various implementation details of the underlying
1583 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1584 * () tries to send packet as a "bulk" to the device.
1586 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1588 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1590 * Redirect the packet to the socket referenced by *map* (of type
1591 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1592 * egress interfaces can be used for redirection. The
1593 * **BPF_F_INGRESS** value in *flags* is used to make the
1594 * distinction (ingress path is selected if the flag is present,
1595 * egress path otherwise). This is the only flag supported for now.
1597 * **SK_PASS** on success, or **SK_DROP** on error.
1599 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1601 * Add an entry to, or update a *map* referencing sockets. The
1602 * *skops* is used as a new value for the entry associated to
1603 * *key*. *flags* is one of:
1606 * The entry for *key* must not exist in the map.
1608 * The entry for *key* must already exist in the map.
1610 * No condition on the existence of the entry for *key*.
1612 * If the *map* has eBPF programs (parser and verdict), those will
1613 * be inherited by the socket being added. If the socket is
1614 * already attached to eBPF programs, this results in an error.
1616 * 0 on success, or a negative error in case of failure.
1618 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1620 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1621 * *delta* (which can be positive or negative). Note that this
1622 * operation modifies the address stored in *xdp_md*\ **->data**,
1623 * so the latter must be loaded only after the helper has been
1626 * The use of *xdp_md*\ **->data_meta** is optional and programs
1627 * are not required to use it. The rationale is that when the
1628 * packet is processed with XDP (e.g. as DoS filter), it is
1629 * possible to push further meta data along with it before passing
1630 * to the stack, and to give the guarantee that an ingress eBPF
1631 * program attached as a TC classifier on the same device can pick
1632 * this up for further post-processing. Since TC works with socket
1633 * buffers, it remains possible to set from XDP the **mark** or
1634 * **priority** pointers, or other pointers for the socket buffer.
1635 * Having this scratch space generic and programmable allows for
1636 * more flexibility as the user is free to store whatever meta
1639 * A call to this helper is susceptible to change the underlying
1640 * packet buffer. Therefore, at load time, all checks on pointers
1641 * previously done by the verifier are invalidated and must be
1642 * performed again, if the helper is used in combination with
1643 * direct packet access.
1645 * 0 on success, or a negative error in case of failure.
1647 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1649 * Read the value of a perf event counter, and store it into *buf*
1650 * of size *buf_size*. This helper relies on a *map* of type
1651 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1652 * counter is selected when *map* is updated with perf event file
1653 * descriptors. The *map* is an array whose size is the number of
1654 * available CPUs, and each cell contains a value relative to one
1655 * CPU. The value to retrieve is indicated by *flags*, that
1656 * contains the index of the CPU to look up, masked with
1657 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1658 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1659 * current CPU should be retrieved.
1661 * This helper behaves in a way close to
1662 * **bpf_perf_event_read**\ () helper, save that instead of
1663 * just returning the value observed, it fills the *buf*
1664 * structure. This allows for additional data to be retrieved: in
1665 * particular, the enabled and running times (in *buf*\
1666 * **->enabled** and *buf*\ **->running**, respectively) are
1667 * copied. In general, **bpf_perf_event_read_value**\ () is
1668 * recommended over **bpf_perf_event_read**\ (), which has some
1669 * ABI issues and provides fewer functionalities.
1671 * These values are interesting, because hardware PMU (Performance
1672 * Monitoring Unit) counters are limited resources. When there are
1673 * more PMU based perf events opened than available counters,
1674 * kernel will multiplex these events so each event gets certain
1675 * percentage (but not all) of the PMU time. In case that
1676 * multiplexing happens, the number of samples or counter value
1677 * will not reflect the case compared to when no multiplexing
1678 * occurs. This makes comparison between different runs difficult.
1679 * Typically, the counter value should be normalized before
1680 * comparing to other experiments. The usual normalization is done
1685 * normalized_counter = counter * t_enabled / t_running
1687 * Where t_enabled is the time enabled for event and t_running is
1688 * the time running for event since last normalization. The
1689 * enabled and running times are accumulated since the perf event
1690 * open. To achieve scaling factor between two invocations of an
1691 * eBPF program, users can can use CPU id as the key (which is
1692 * typical for perf array usage model) to remember the previous
1693 * value and do the calculation inside the eBPF program.
1695 * 0 on success, or a negative error in case of failure.
1697 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1699 * For en eBPF program attached to a perf event, retrieve the
1700 * value of the event counter associated to *ctx* and store it in
1701 * the structure pointed by *buf* and of size *buf_size*. Enabled
1702 * and running times are also stored in the structure (see
1703 * description of helper **bpf_perf_event_read_value**\ () for
1706 * 0 on success, or a negative error in case of failure.
1708 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1710 * Emulate a call to **getsockopt()** on the socket associated to
1711 * *bpf_socket*, which must be a full socket. The *level* at
1712 * which the option resides and the name *optname* of the option
1713 * must be specified, see **getsockopt(2)** for more information.
1714 * The retrieved value is stored in the structure pointed by
1715 * *opval* and of length *optlen*.
1717 * This helper actually implements a subset of **getsockopt()**.
1718 * It supports the following *level*\ s:
1720 * * **IPPROTO_TCP**, which supports *optname*
1721 * **TCP_CONGESTION**.
1722 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1723 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1725 * 0 on success, or a negative error in case of failure.
1727 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1729 * Used for error injection, this helper uses kprobes to override
1730 * the return value of the probed function, and to set it to *rc*.
1731 * The first argument is the context *regs* on which the kprobe
1734 * This helper works by setting setting the PC (program counter)
1735 * to an override function which is run in place of the original
1736 * probed function. This means the probed function is not run at
1737 * all. The replacement function just returns with the required
1740 * This helper has security implications, and thus is subject to
1741 * restrictions. It is only available if the kernel was compiled
1742 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1743 * option, and in this case it only works on functions tagged with
1744 * **ALLOW_ERROR_INJECTION** in the kernel code.
1746 * Also, the helper is only available for the architectures having
1747 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1748 * x86 architecture is the only one to support this feature.
1752 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1754 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1755 * for the full TCP socket associated to *bpf_sock_ops* to
1758 * The primary use of this field is to determine if there should
1759 * be calls to eBPF programs of type
1760 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1761 * code. A program of the same type can change its value, per
1762 * connection and as necessary, when the connection is
1763 * established. This field is directly accessible for reading, but
1764 * this helper must be used for updates in order to return an
1765 * error if an eBPF program tries to set a callback that is not
1766 * supported in the current kernel.
1768 * *argval* is a flag array which can combine these flags:
1770 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1771 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1772 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1774 * Therefore, this function can be used to clear a callback flag by
1775 * setting the appropriate bit to zero. e.g. to disable the RTO
1778 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1779 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1781 * Here are some examples of where one could call such eBPF
1785 * * When a packet is retransmitted.
1786 * * When the connection terminates.
1787 * * When a packet is sent.
1788 * * When a packet is received.
1790 * Code **-EINVAL** if the socket is not a full TCP socket;
1791 * otherwise, a positive number containing the bits that could not
1792 * be set is returned (which comes down to 0 if all bits were set
1795 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1797 * This helper is used in programs implementing policies at the
1798 * socket level. If the message *msg* is allowed to pass (i.e. if
1799 * the verdict eBPF program returns **SK_PASS**), redirect it to
1800 * the socket referenced by *map* (of type
1801 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1802 * egress interfaces can be used for redirection. The
1803 * **BPF_F_INGRESS** value in *flags* is used to make the
1804 * distinction (ingress path is selected if the flag is present,
1805 * egress path otherwise). This is the only flag supported for now.
1807 * **SK_PASS** on success, or **SK_DROP** on error.
1809 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1811 * For socket policies, apply the verdict of the eBPF program to
1812 * the next *bytes* (number of bytes) of message *msg*.
1814 * For example, this helper can be used in the following cases:
1816 * * A single **sendmsg**\ () or **sendfile**\ () system call
1817 * contains multiple logical messages that the eBPF program is
1818 * supposed to read and for which it should apply a verdict.
1819 * * An eBPF program only cares to read the first *bytes* of a
1820 * *msg*. If the message has a large payload, then setting up
1821 * and calling the eBPF program repeatedly for all bytes, even
1822 * though the verdict is already known, would create unnecessary
1825 * When called from within an eBPF program, the helper sets a
1826 * counter internal to the BPF infrastructure, that is used to
1827 * apply the last verdict to the next *bytes*. If *bytes* is
1828 * smaller than the current data being processed from a
1829 * **sendmsg**\ () or **sendfile**\ () system call, the first
1830 * *bytes* will be sent and the eBPF program will be re-run with
1831 * the pointer for start of data pointing to byte number *bytes*
1832 * **+ 1**. If *bytes* is larger than the current data being
1833 * processed, then the eBPF verdict will be applied to multiple
1834 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1837 * Note that if a socket closes with the internal counter holding
1838 * a non-zero value, this is not a problem because data is not
1839 * being buffered for *bytes* and is sent as it is received.
1843 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1845 * For socket policies, prevent the execution of the verdict eBPF
1846 * program for message *msg* until *bytes* (byte number) have been
1849 * This can be used when one needs a specific number of bytes
1850 * before a verdict can be assigned, even if the data spans
1851 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1852 * case would be a user calling **sendmsg**\ () repeatedly with
1853 * 1-byte long message segments. Obviously, this is bad for
1854 * performance, but it is still valid. If the eBPF program needs
1855 * *bytes* bytes to validate a header, this helper can be used to
1856 * prevent the eBPF program to be called again until *bytes* have
1861 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1863 * For socket policies, pull in non-linear data from user space
1864 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1865 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1868 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1869 * *msg* it can only parse data that the (**data**, **data_end**)
1870 * pointers have already consumed. For **sendmsg**\ () hooks this
1871 * is likely the first scatterlist element. But for calls relying
1872 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1873 * be the range (**0**, **0**) because the data is shared with
1874 * user space and by default the objective is to avoid allowing
1875 * user space to modify data while (or after) eBPF verdict is
1876 * being decided. This helper can be used to pull in data and to
1877 * set the start and end pointer to given values. Data will be
1878 * copied if necessary (i.e. if data was not linear and if start
1879 * and end pointers do not point to the same chunk).
1881 * A call to this helper is susceptible to change the underlying
1882 * packet buffer. Therefore, at load time, all checks on pointers
1883 * previously done by the verifier are invalidated and must be
1884 * performed again, if the helper is used in combination with
1885 * direct packet access.
1887 * All values for *flags* are reserved for future usage, and must
1890 * 0 on success, or a negative error in case of failure.
1892 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1894 * Bind the socket associated to *ctx* to the address pointed by
1895 * *addr*, of length *addr_len*. This allows for making outgoing
1896 * connection from the desired IP address, which can be useful for
1897 * example when all processes inside a cgroup should use one
1898 * single IP address on a host that has multiple IP configured.
1900 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1901 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1902 * **AF_INET6**). Looking for a free port to bind to can be
1903 * expensive, therefore binding to port is not permitted by the
1904 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1905 * must be set to zero.
1907 * 0 on success, or a negative error in case of failure.
1909 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1911 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1912 * only possible to shrink the packet as of this writing,
1913 * therefore *delta* must be a negative integer.
1915 * A call to this helper is susceptible to change the underlying
1916 * packet buffer. Therefore, at load time, all checks on pointers
1917 * previously done by the verifier are invalidated and must be
1918 * performed again, if the helper is used in combination with
1919 * direct packet access.
1921 * 0 on success, or a negative error in case of failure.
1923 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1925 * Retrieve the XFRM state (IP transform framework, see also
1926 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1928 * The retrieved value is stored in the **struct bpf_xfrm_state**
1929 * pointed by *xfrm_state* and of length *size*.
1931 * All values for *flags* are reserved for future usage, and must
1934 * This helper is available only if the kernel was compiled with
1935 * **CONFIG_XFRM** configuration option.
1937 * 0 on success, or a negative error in case of failure.
1939 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1941 * Return a user or a kernel stack in bpf program provided buffer.
1942 * To achieve this, the helper needs *ctx*, which is a pointer
1943 * to the context on which the tracing program is executed.
1944 * To store the stacktrace, the bpf program provides *buf* with
1945 * a nonnegative *size*.
1947 * The last argument, *flags*, holds the number of stack frames to
1948 * skip (from 0 to 255), masked with
1949 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1950 * the following flags:
1952 * **BPF_F_USER_STACK**
1953 * Collect a user space stack instead of a kernel stack.
1954 * **BPF_F_USER_BUILD_ID**
1955 * Collect buildid+offset instead of ips for user stack,
1956 * only valid if **BPF_F_USER_STACK** is also specified.
1958 * **bpf_get_stack**\ () can collect up to
1959 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1960 * to sufficient large buffer size. Note that
1961 * this limit can be controlled with the **sysctl** program, and
1962 * that it should be manually increased in order to profile long
1963 * user stacks (such as stacks for Java programs). To do so, use:
1967 * # sysctl kernel.perf_event_max_stack=<new value>
1969 * A non-negative value equal to or less than *size* on success,
1970 * or a negative error in case of failure.
1972 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1974 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1975 * it provides an easy way to load *len* bytes from *offset*
1976 * from the packet associated to *skb*, into the buffer pointed
1977 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1978 * a fifth argument *start_header* exists in order to select a
1979 * base offset to start from. *start_header* can be one of:
1981 * **BPF_HDR_START_MAC**
1982 * Base offset to load data from is *skb*'s mac header.
1983 * **BPF_HDR_START_NET**
1984 * Base offset to load data from is *skb*'s network header.
1986 * In general, "direct packet access" is the preferred method to
1987 * access packet data, however, this helper is in particular useful
1988 * in socket filters where *skb*\ **->data** does not always point
1989 * to the start of the mac header and where "direct packet access"
1992 * 0 on success, or a negative error in case of failure.
1994 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1996 * Do FIB lookup in kernel tables using parameters in *params*.
1997 * If lookup is successful and result shows packet is to be
1998 * forwarded, the neighbor tables are searched for the nexthop.
1999 * If successful (ie., FIB lookup shows forwarding and nexthop
2000 * is resolved), the nexthop address is returned in ipv4_dst
2001 * or ipv6_dst based on family, smac is set to mac address of
2002 * egress device, dmac is set to nexthop mac address, rt_metric
2003 * is set to metric from route (IPv4/IPv6 only), and ifindex
2004 * is set to the device index of the nexthop from the FIB lookup.
2006 * *plen* argument is the size of the passed in struct.
2007 * *flags* argument can be a combination of one or more of the
2010 * **BPF_FIB_LOOKUP_DIRECT**
2011 * Do a direct table lookup vs full lookup using FIB
2013 * **BPF_FIB_LOOKUP_OUTPUT**
2014 * Perform lookup from an egress perspective (default is
2017 * *ctx* is either **struct xdp_md** for XDP programs or
2018 * **struct sk_buff** tc cls_act programs.
2020 * * < 0 if any input argument is invalid
2021 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2022 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2023 * packet is not forwarded or needs assist from full stack
2025 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
2027 * Add an entry to, or update a sockhash *map* referencing sockets.
2028 * The *skops* is used as a new value for the entry associated to
2029 * *key*. *flags* is one of:
2032 * The entry for *key* must not exist in the map.
2034 * The entry for *key* must already exist in the map.
2036 * No condition on the existence of the entry for *key*.
2038 * If the *map* has eBPF programs (parser and verdict), those will
2039 * be inherited by the socket being added. If the socket is
2040 * already attached to eBPF programs, this results in an error.
2042 * 0 on success, or a negative error in case of failure.
2044 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2046 * This helper is used in programs implementing policies at the
2047 * socket level. If the message *msg* is allowed to pass (i.e. if
2048 * the verdict eBPF program returns **SK_PASS**), redirect it to
2049 * the socket referenced by *map* (of type
2050 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2051 * egress interfaces can be used for redirection. The
2052 * **BPF_F_INGRESS** value in *flags* is used to make the
2053 * distinction (ingress path is selected if the flag is present,
2054 * egress path otherwise). This is the only flag supported for now.
2056 * **SK_PASS** on success, or **SK_DROP** on error.
2058 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2060 * This helper is used in programs implementing policies at the
2061 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2062 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2063 * to the socket referenced by *map* (of type
2064 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2065 * egress interfaces can be used for redirection. The
2066 * **BPF_F_INGRESS** value in *flags* is used to make the
2067 * distinction (ingress path is selected if the flag is present,
2068 * egress otherwise). This is the only flag supported for now.
2070 * **SK_PASS** on success, or **SK_DROP** on error.
2072 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2074 * Encapsulate the packet associated to *skb* within a Layer 3
2075 * protocol header. This header is provided in the buffer at
2076 * address *hdr*, with *len* its size in bytes. *type* indicates
2077 * the protocol of the header and can be one of:
2079 * **BPF_LWT_ENCAP_SEG6**
2080 * IPv6 encapsulation with Segment Routing Header
2081 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2082 * the IPv6 header is computed by the kernel.
2083 * **BPF_LWT_ENCAP_SEG6_INLINE**
2084 * Only works if *skb* contains an IPv6 packet. Insert a
2085 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2087 * **BPF_LWT_ENCAP_IP**
2088 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2089 * must be IPv4 or IPv6, followed by zero or more
2090 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2091 * total bytes in all prepended headers. Please note that
2092 * if **skb_is_gso**\ (*skb*) is true, no more than two
2093 * headers can be prepended, and the inner header, if
2094 * present, should be either GRE or UDP/GUE.
2096 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2097 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2098 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2099 * **BPF_PROG_TYPE_LWT_XMIT**.
2101 * A call to this helper is susceptible to change the underlying
2102 * packet buffer. Therefore, at load time, all checks on pointers
2103 * previously done by the verifier are invalidated and must be
2104 * performed again, if the helper is used in combination with
2105 * direct packet access.
2107 * 0 on success, or a negative error in case of failure.
2109 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2111 * Store *len* bytes from address *from* into the packet
2112 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2113 * inside the outermost IPv6 Segment Routing Header can be
2114 * modified through this helper.
2116 * A call to this helper is susceptible to change the underlying
2117 * packet buffer. Therefore, at load time, all checks on pointers
2118 * previously done by the verifier are invalidated and must be
2119 * performed again, if the helper is used in combination with
2120 * direct packet access.
2122 * 0 on success, or a negative error in case of failure.
2124 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2126 * Adjust the size allocated to TLVs in the outermost IPv6
2127 * Segment Routing Header contained in the packet associated to
2128 * *skb*, at position *offset* by *delta* bytes. Only offsets
2129 * after the segments are accepted. *delta* can be as well
2130 * positive (growing) as negative (shrinking).
2132 * A call to this helper is susceptible to change the underlying
2133 * packet buffer. Therefore, at load time, all checks on pointers
2134 * previously done by the verifier are invalidated and must be
2135 * performed again, if the helper is used in combination with
2136 * direct packet access.
2138 * 0 on success, or a negative error in case of failure.
2140 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2142 * Apply an IPv6 Segment Routing action of type *action* to the
2143 * packet associated to *skb*. Each action takes a parameter
2144 * contained at address *param*, and of length *param_len* bytes.
2145 * *action* can be one of:
2147 * **SEG6_LOCAL_ACTION_END_X**
2148 * End.X action: Endpoint with Layer-3 cross-connect.
2149 * Type of *param*: **struct in6_addr**.
2150 * **SEG6_LOCAL_ACTION_END_T**
2151 * End.T action: Endpoint with specific IPv6 table lookup.
2152 * Type of *param*: **int**.
2153 * **SEG6_LOCAL_ACTION_END_B6**
2154 * End.B6 action: Endpoint bound to an SRv6 policy.
2155 * Type of *param*: **struct ipv6_sr_hdr**.
2156 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2157 * End.B6.Encap action: Endpoint bound to an SRv6
2158 * encapsulation policy.
2159 * Type of *param*: **struct ipv6_sr_hdr**.
2161 * A call to this helper is susceptible to change the underlying
2162 * packet buffer. Therefore, at load time, all checks on pointers
2163 * previously done by the verifier are invalidated and must be
2164 * performed again, if the helper is used in combination with
2165 * direct packet access.
2167 * 0 on success, or a negative error in case of failure.
2169 * int bpf_rc_repeat(void *ctx)
2171 * This helper is used in programs implementing IR decoding, to
2172 * report a successfully decoded repeat key message. This delays
2173 * the generation of a key up event for previously generated
2176 * Some IR protocols like NEC have a special IR message for
2177 * repeating last button, for when a button is held down.
2179 * The *ctx* should point to the lirc sample as passed into
2182 * This helper is only available is the kernel was compiled with
2183 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2188 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2190 * This helper is used in programs implementing IR decoding, to
2191 * report a successfully decoded key press with *scancode*,
2192 * *toggle* value in the given *protocol*. The scancode will be
2193 * translated to a keycode using the rc keymap, and reported as
2194 * an input key down event. After a period a key up event is
2195 * generated. This period can be extended by calling either
2196 * **bpf_rc_keydown**\ () again with the same values, or calling
2197 * **bpf_rc_repeat**\ ().
2199 * Some protocols include a toggle bit, in case the button was
2200 * released and pressed again between consecutive scancodes.
2202 * The *ctx* should point to the lirc sample as passed into
2205 * The *protocol* is the decoded protocol number (see
2206 * **enum rc_proto** for some predefined values).
2208 * This helper is only available is the kernel was compiled with
2209 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2214 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2216 * Return the cgroup v2 id of the socket associated with the *skb*.
2217 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2218 * helper for cgroup v1 by providing a tag resp. identifier that
2219 * can be matched on or used for map lookups e.g. to implement
2220 * policy. The cgroup v2 id of a given path in the hierarchy is
2221 * exposed in user space through the f_handle API in order to get
2222 * to the same 64-bit id.
2224 * This helper can be used on TC egress path, but not on ingress,
2225 * and is available only if the kernel was compiled with the
2226 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2228 * The id is returned or 0 in case the id could not be retrieved.
2230 * u64 bpf_get_current_cgroup_id(void)
2232 * A 64-bit integer containing the current cgroup id based
2233 * on the cgroup within which the current task is running.
2235 * void *bpf_get_local_storage(void *map, u64 flags)
2237 * Get the pointer to the local storage area.
2238 * The type and the size of the local storage is defined
2239 * by the *map* argument.
2240 * The *flags* meaning is specific for each map type,
2241 * and has to be 0 for cgroup local storage.
2243 * Depending on the BPF program type, a local storage area
2244 * can be shared between multiple instances of the BPF program,
2245 * running simultaneously.
2247 * A user should care about the synchronization by himself.
2248 * For example, by using the **BPF_STX_XADD** instruction to alter
2251 * A pointer to the local storage area.
2253 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2255 * Select a **SO_REUSEPORT** socket from a
2256 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2257 * It checks the selected socket is matching the incoming
2258 * request in the socket buffer.
2260 * 0 on success, or a negative error in case of failure.
2262 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2264 * Return id of cgroup v2 that is ancestor of cgroup associated
2265 * with the *skb* at the *ancestor_level*. The root cgroup is at
2266 * *ancestor_level* zero and each step down the hierarchy
2267 * increments the level. If *ancestor_level* == level of cgroup
2268 * associated with *skb*, then return value will be same as that
2269 * of **bpf_skb_cgroup_id**\ ().
2271 * The helper is useful to implement policies based on cgroups
2272 * that are upper in hierarchy than immediate cgroup associated
2275 * The format of returned id and helper limitations are same as in
2276 * **bpf_skb_cgroup_id**\ ().
2278 * The id is returned or 0 in case the id could not be retrieved.
2280 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2282 * Look for TCP socket matching *tuple*, optionally in a child
2283 * network namespace *netns*. The return value must be checked,
2284 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2286 * The *ctx* should point to the context of the program, such as
2287 * the skb or socket (depending on the hook in use). This is used
2288 * to determine the base network namespace for the lookup.
2290 * *tuple_size* must be one of:
2292 * **sizeof**\ (*tuple*\ **->ipv4**)
2293 * Look for an IPv4 socket.
2294 * **sizeof**\ (*tuple*\ **->ipv6**)
2295 * Look for an IPv6 socket.
2297 * If the *netns* is a negative signed 32-bit integer, then the
2298 * socket lookup table in the netns associated with the *ctx* will
2299 * will be used. For the TC hooks, this is the netns of the device
2300 * in the skb. For socket hooks, this is the netns of the socket.
2301 * If *netns* is any other signed 32-bit value greater than or
2302 * equal to zero then it specifies the ID of the netns relative to
2303 * the netns associated with the *ctx*. *netns* values beyond the
2304 * range of 32-bit integers are reserved for future use.
2306 * All values for *flags* are reserved for future usage, and must
2309 * This helper is available only if the kernel was compiled with
2310 * **CONFIG_NET** configuration option.
2312 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2313 * For sockets with reuseport option, the **struct bpf_sock**
2314 * result is from *reuse*\ **->socks**\ [] using the hash of the
2317 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2319 * Look for UDP socket matching *tuple*, optionally in a child
2320 * network namespace *netns*. The return value must be checked,
2321 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2323 * The *ctx* should point to the context of the program, such as
2324 * the skb or socket (depending on the hook in use). This is used
2325 * to determine the base network namespace for the lookup.
2327 * *tuple_size* must be one of:
2329 * **sizeof**\ (*tuple*\ **->ipv4**)
2330 * Look for an IPv4 socket.
2331 * **sizeof**\ (*tuple*\ **->ipv6**)
2332 * Look for an IPv6 socket.
2334 * If the *netns* is a negative signed 32-bit integer, then the
2335 * socket lookup table in the netns associated with the *ctx* will
2336 * will be used. For the TC hooks, this is the netns of the device
2337 * in the skb. For socket hooks, this is the netns of the socket.
2338 * If *netns* is any other signed 32-bit value greater than or
2339 * equal to zero then it specifies the ID of the netns relative to
2340 * the netns associated with the *ctx*. *netns* values beyond the
2341 * range of 32-bit integers are reserved for future use.
2343 * All values for *flags* are reserved for future usage, and must
2346 * This helper is available only if the kernel was compiled with
2347 * **CONFIG_NET** configuration option.
2349 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2350 * For sockets with reuseport option, the **struct bpf_sock**
2351 * result is from *reuse*\ **->socks**\ [] using the hash of the
2354 * int bpf_sk_release(struct bpf_sock *sock)
2356 * Release the reference held by *sock*. *sock* must be a
2357 * non-**NULL** pointer that was returned from
2358 * **bpf_sk_lookup_xxx**\ ().
2360 * 0 on success, or a negative error in case of failure.
2362 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2364 * Push an element *value* in *map*. *flags* is one of:
2367 * If the queue/stack is full, the oldest element is
2368 * removed to make room for this.
2370 * 0 on success, or a negative error in case of failure.
2372 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2374 * Pop an element from *map*.
2376 * 0 on success, or a negative error in case of failure.
2378 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2380 * Get an element from *map* without removing it.
2382 * 0 on success, or a negative error in case of failure.
2384 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2386 * For socket policies, insert *len* bytes into *msg* at offset
2389 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2390 * *msg* it may want to insert metadata or options into the *msg*.
2391 * This can later be read and used by any of the lower layer BPF
2394 * This helper may fail if under memory pressure (a malloc
2395 * fails) in these cases BPF programs will get an appropriate
2396 * error and BPF programs will need to handle them.
2398 * 0 on success, or a negative error in case of failure.
2400 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2402 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2403 * This may result in **ENOMEM** errors under certain situations if
2404 * an allocation and copy are required due to a full ring buffer.
2405 * However, the helper will try to avoid doing the allocation
2406 * if possible. Other errors can occur if input parameters are
2407 * invalid either due to *start* byte not being valid part of *msg*
2408 * payload and/or *pop* value being to large.
2410 * 0 on success, or a negative error in case of failure.
2412 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2414 * This helper is used in programs implementing IR decoding, to
2415 * report a successfully decoded pointer movement.
2417 * The *ctx* should point to the lirc sample as passed into
2420 * This helper is only available is the kernel was compiled with
2421 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2426 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2428 * Acquire a spinlock represented by the pointer *lock*, which is
2429 * stored as part of a value of a map. Taking the lock allows to
2430 * safely update the rest of the fields in that value. The
2431 * spinlock can (and must) later be released with a call to
2432 * **bpf_spin_unlock**\ (\ *lock*\ ).
2434 * Spinlocks in BPF programs come with a number of restrictions
2437 * * **bpf_spin_lock** objects are only allowed inside maps of
2438 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2439 * list could be extended in the future).
2440 * * BTF description of the map is mandatory.
2441 * * The BPF program can take ONE lock at a time, since taking two
2442 * or more could cause dead locks.
2443 * * Only one **struct bpf_spin_lock** is allowed per map element.
2444 * * When the lock is taken, calls (either BPF to BPF or helpers)
2446 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2447 * allowed inside a spinlock-ed region.
2448 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2449 * the lock, on all execution paths, before it returns.
2450 * * The BPF program can access **struct bpf_spin_lock** only via
2451 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2452 * helpers. Loading or storing data into the **struct
2453 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2454 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2455 * of the map value must be a struct and have **struct
2456 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2457 * Nested lock inside another struct is not allowed.
2458 * * The **struct bpf_spin_lock** *lock* field in a map value must
2459 * be aligned on a multiple of 4 bytes in that value.
2460 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2461 * the **bpf_spin_lock** field to user space.
2462 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2463 * a BPF program, do not update the **bpf_spin_lock** field.
2464 * * **bpf_spin_lock** cannot be on the stack or inside a
2465 * networking packet (it can only be inside of a map values).
2466 * * **bpf_spin_lock** is available to root only.
2467 * * Tracing programs and socket filter programs cannot use
2468 * **bpf_spin_lock**\ () due to insufficient preemption checks
2469 * (but this may change in the future).
2470 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2474 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2476 * Release the *lock* previously locked by a call to
2477 * **bpf_spin_lock**\ (\ *lock*\ ).
2481 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2483 * This helper gets a **struct bpf_sock** pointer such
2484 * that all the fields in this **bpf_sock** can be accessed.
2486 * A **struct bpf_sock** pointer on success, or **NULL** in
2489 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2491 * This helper gets a **struct bpf_tcp_sock** pointer from a
2492 * **struct bpf_sock** pointer.
2494 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2497 * int bpf_skb_ecn_set_ce(struct sk_buf *skb)
2499 * Set ECN (Explicit Congestion Notification) field of IP header
2500 * to **CE** (Congestion Encountered) if current value is **ECT**
2501 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2504 * 1 if the **CE** flag is set (either by the current helper call
2505 * or because it was already present), 0 if it is not set.
2507 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2509 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2510 * **bpf_sk_release**\ () is unnecessary and not allowed.
2512 * A **struct bpf_sock** pointer on success, or **NULL** in
2515 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2517 * Look for TCP socket matching *tuple*, optionally in a child
2518 * network namespace *netns*. The return value must be checked,
2519 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2521 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2522 * that it also returns timewait or request sockets. Use
2523 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2526 * This helper is available only if the kernel was compiled with
2527 * **CONFIG_NET** configuration option.
2529 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2530 * For sockets with reuseport option, the **struct bpf_sock**
2531 * result is from *reuse*\ **->socks**\ [] using the hash of the
2534 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2536 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2537 * the listening socket in *sk*.
2539 * *iph* points to the start of the IPv4 or IPv6 header, while
2540 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2541 * **sizeof**\ (**struct ip6hdr**).
2543 * *th* points to the start of the TCP header, while *th_len*
2544 * contains **sizeof**\ (**struct tcphdr**).
2547 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2550 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2552 * Get name of sysctl in /proc/sys/ and copy it into provided by
2553 * program buffer *buf* of size *buf_len*.
2555 * The buffer is always NUL terminated, unless it's zero-sized.
2557 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2558 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2559 * only (e.g. "tcp_mem").
2561 * Number of character copied (not including the trailing NUL).
2563 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2564 * truncated name in this case).
2566 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2568 * Get current value of sysctl as it is presented in /proc/sys
2569 * (incl. newline, etc), and copy it as a string into provided
2570 * by program buffer *buf* of size *buf_len*.
2572 * The whole value is copied, no matter what file position user
2573 * space issued e.g. sys_read at.
2575 * The buffer is always NUL terminated, unless it's zero-sized.
2577 * Number of character copied (not including the trailing NUL).
2579 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2580 * truncated name in this case).
2582 * **-EINVAL** if current value was unavailable, e.g. because
2583 * sysctl is uninitialized and read returns -EIO for it.
2585 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2587 * Get new value being written by user space to sysctl (before
2588 * the actual write happens) and copy it as a string into
2589 * provided by program buffer *buf* of size *buf_len*.
2591 * User space may write new value at file position > 0.
2593 * The buffer is always NUL terminated, unless it's zero-sized.
2595 * Number of character copied (not including the trailing NUL).
2597 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2598 * truncated name in this case).
2600 * **-EINVAL** if sysctl is being read.
2602 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2604 * Override new value being written by user space to sysctl with
2605 * value provided by program in buffer *buf* of size *buf_len*.
2607 * *buf* should contain a string in same form as provided by user
2608 * space on sysctl write.
2610 * User space may write new value at file position > 0. To override
2611 * the whole sysctl value file position should be set to zero.
2615 * **-E2BIG** if the *buf_len* is too big.
2617 * **-EINVAL** if sysctl is being read.
2619 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2621 * Convert the initial part of the string from buffer *buf* of
2622 * size *buf_len* to a long integer according to the given base
2623 * and save the result in *res*.
2625 * The string may begin with an arbitrary amount of white space
2626 * (as determined by **isspace**\ (3)) followed by a single
2627 * optional '**-**' sign.
2629 * Five least significant bits of *flags* encode base, other bits
2630 * are currently unused.
2632 * Base must be either 8, 10, 16 or 0 to detect it automatically
2633 * similar to user space **strtol**\ (3).
2635 * Number of characters consumed on success. Must be positive but
2636 * no more than *buf_len*.
2638 * **-EINVAL** if no valid digits were found or unsupported base
2641 * **-ERANGE** if resulting value was out of range.
2643 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2645 * Convert the initial part of the string from buffer *buf* of
2646 * size *buf_len* to an unsigned long integer according to the
2647 * given base and save the result in *res*.
2649 * The string may begin with an arbitrary amount of white space
2650 * (as determined by **isspace**\ (3)).
2652 * Five least significant bits of *flags* encode base, other bits
2653 * are currently unused.
2655 * Base must be either 8, 10, 16 or 0 to detect it automatically
2656 * similar to user space **strtoul**\ (3).
2658 * Number of characters consumed on success. Must be positive but
2659 * no more than *buf_len*.
2661 * **-EINVAL** if no valid digits were found or unsupported base
2664 * **-ERANGE** if resulting value was out of range.
2666 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2668 * Get a bpf-local-storage from a *sk*.
2670 * Logically, it could be thought of getting the value from
2671 * a *map* with *sk* as the **key**. From this
2672 * perspective, the usage is not much different from
2673 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2674 * helper enforces the key must be a full socket and the map must
2675 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2677 * Underneath, the value is stored locally at *sk* instead of
2678 * the *map*. The *map* is used as the bpf-local-storage
2679 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2680 * searched against all bpf-local-storages residing at *sk*.
2682 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2683 * used such that a new bpf-local-storage will be
2684 * created if one does not exist. *value* can be used
2685 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2686 * the initial value of a bpf-local-storage. If *value* is
2687 * **NULL**, the new bpf-local-storage will be zero initialized.
2689 * A bpf-local-storage pointer is returned on success.
2691 * **NULL** if not found or there was an error in adding
2692 * a new bpf-local-storage.
2694 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2696 * Delete a bpf-local-storage from a *sk*.
2700 * **-ENOENT** if the bpf-local-storage cannot be found.
2702 * int bpf_send_signal(u32 sig)
2704 * Send signal *sig* to the current task.
2706 * 0 on success or successfully queued.
2708 * **-EBUSY** if work queue under nmi is full.
2710 * **-EINVAL** if *sig* is invalid.
2712 * **-EPERM** if no permission to send the *sig*.
2714 * **-EAGAIN** if bpf program can try again.
2716 #define __BPF_FUNC_MAPPER(FN) \
2718 FN(map_lookup_elem), \
2719 FN(map_update_elem), \
2720 FN(map_delete_elem), \
2724 FN(get_prandom_u32), \
2725 FN(get_smp_processor_id), \
2726 FN(skb_store_bytes), \
2727 FN(l3_csum_replace), \
2728 FN(l4_csum_replace), \
2730 FN(clone_redirect), \
2731 FN(get_current_pid_tgid), \
2732 FN(get_current_uid_gid), \
2733 FN(get_current_comm), \
2734 FN(get_cgroup_classid), \
2735 FN(skb_vlan_push), \
2737 FN(skb_get_tunnel_key), \
2738 FN(skb_set_tunnel_key), \
2739 FN(perf_event_read), \
2741 FN(get_route_realm), \
2742 FN(perf_event_output), \
2743 FN(skb_load_bytes), \
2746 FN(skb_get_tunnel_opt), \
2747 FN(skb_set_tunnel_opt), \
2748 FN(skb_change_proto), \
2749 FN(skb_change_type), \
2750 FN(skb_under_cgroup), \
2751 FN(get_hash_recalc), \
2752 FN(get_current_task), \
2753 FN(probe_write_user), \
2754 FN(current_task_under_cgroup), \
2755 FN(skb_change_tail), \
2756 FN(skb_pull_data), \
2758 FN(set_hash_invalid), \
2759 FN(get_numa_node_id), \
2760 FN(skb_change_head), \
2761 FN(xdp_adjust_head), \
2762 FN(probe_read_str), \
2763 FN(get_socket_cookie), \
2764 FN(get_socket_uid), \
2767 FN(skb_adjust_room), \
2769 FN(sk_redirect_map), \
2770 FN(sock_map_update), \
2771 FN(xdp_adjust_meta), \
2772 FN(perf_event_read_value), \
2773 FN(perf_prog_read_value), \
2775 FN(override_return), \
2776 FN(sock_ops_cb_flags_set), \
2777 FN(msg_redirect_map), \
2778 FN(msg_apply_bytes), \
2779 FN(msg_cork_bytes), \
2780 FN(msg_pull_data), \
2782 FN(xdp_adjust_tail), \
2783 FN(skb_get_xfrm_state), \
2785 FN(skb_load_bytes_relative), \
2787 FN(sock_hash_update), \
2788 FN(msg_redirect_hash), \
2789 FN(sk_redirect_hash), \
2790 FN(lwt_push_encap), \
2791 FN(lwt_seg6_store_bytes), \
2792 FN(lwt_seg6_adjust_srh), \
2793 FN(lwt_seg6_action), \
2796 FN(skb_cgroup_id), \
2797 FN(get_current_cgroup_id), \
2798 FN(get_local_storage), \
2799 FN(sk_select_reuseport), \
2800 FN(skb_ancestor_cgroup_id), \
2801 FN(sk_lookup_tcp), \
2802 FN(sk_lookup_udp), \
2804 FN(map_push_elem), \
2806 FN(map_peek_elem), \
2807 FN(msg_push_data), \
2809 FN(rc_pointer_rel), \
2814 FN(skb_ecn_set_ce), \
2815 FN(get_listener_sock), \
2816 FN(skc_lookup_tcp), \
2817 FN(tcp_check_syncookie), \
2818 FN(sysctl_get_name), \
2819 FN(sysctl_get_current_value), \
2820 FN(sysctl_get_new_value), \
2821 FN(sysctl_set_new_value), \
2824 FN(sk_storage_get), \
2825 FN(sk_storage_delete), \
2828 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2829 * function eBPF program intends to call
2831 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2833 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2836 #undef __BPF_ENUM_FN
2838 /* All flags used by eBPF helper functions, placed here. */
2840 /* BPF_FUNC_skb_store_bytes flags. */
2841 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2842 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2844 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2845 * First 4 bits are for passing the header field size.
2847 #define BPF_F_HDR_FIELD_MASK 0xfULL
2849 /* BPF_FUNC_l4_csum_replace flags. */
2850 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2851 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2852 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2854 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2855 #define BPF_F_INGRESS (1ULL << 0)
2857 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2858 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2860 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2861 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2862 #define BPF_F_USER_STACK (1ULL << 8)
2863 /* flags used by BPF_FUNC_get_stackid only. */
2864 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2865 #define BPF_F_REUSE_STACKID (1ULL << 10)
2866 /* flags used by BPF_FUNC_get_stack only. */
2867 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2869 /* BPF_FUNC_skb_set_tunnel_key flags. */
2870 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2871 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2872 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2874 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2875 * BPF_FUNC_perf_event_read_value flags.
2877 #define BPF_F_INDEX_MASK 0xffffffffULL
2878 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2879 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2880 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2882 /* Current network namespace */
2883 #define BPF_F_CURRENT_NETNS (-1L)
2885 /* BPF_FUNC_skb_adjust_room flags. */
2886 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
2888 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
2889 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
2891 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
2892 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
2893 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
2894 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
2895 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
2896 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
2897 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
2899 /* BPF_FUNC_sysctl_get_name flags. */
2900 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
2902 /* BPF_FUNC_sk_storage_get flags */
2903 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0)
2905 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2906 enum bpf_adj_room_mode {
2911 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2912 enum bpf_hdr_start_off {
2917 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2918 enum bpf_lwt_encap_mode {
2920 BPF_LWT_ENCAP_SEG6_INLINE,
2924 #define __bpf_md_ptr(type, name) \
2928 } __attribute__((aligned(8)))
2930 /* user accessible mirror of in-kernel sk_buff.
2931 * new fields can only be added to the end of this structure
2937 __u32 queue_mapping;
2943 __u32 ingress_ifindex;
2953 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2955 __u32 remote_ip4; /* Stored in network byte order */
2956 __u32 local_ip4; /* Stored in network byte order */
2957 __u32 remote_ip6[4]; /* Stored in network byte order */
2958 __u32 local_ip6[4]; /* Stored in network byte order */
2959 __u32 remote_port; /* Stored in network byte order */
2960 __u32 local_port; /* stored in host byte order */
2964 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
2968 __bpf_md_ptr(struct bpf_sock *, sk);
2971 struct bpf_tunnel_key {
2975 __u32 remote_ipv6[4];
2979 __u16 tunnel_ext; /* Padding, future use. */
2983 /* user accessible mirror of in-kernel xfrm_state.
2984 * new fields can only be added to the end of this structure
2986 struct bpf_xfrm_state {
2988 __u32 spi; /* Stored in network byte order */
2990 __u16 ext; /* Padding, future use. */
2992 __u32 remote_ipv4; /* Stored in network byte order */
2993 __u32 remote_ipv6[4]; /* Stored in network byte order */
2997 /* Generic BPF return codes which all BPF program types may support.
2998 * The values are binary compatible with their TC_ACT_* counter-part to
2999 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3002 * XDP is handled seprately, see XDP_*.
3010 /* >127 are reserved for prog type specific return codes.
3012 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3013 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3014 * changed and should be routed based on its new L3 header.
3015 * (This is an L3 redirect, as opposed to L2 redirect
3016 * represented by BPF_REDIRECT above).
3018 BPF_LWT_REROUTE = 128,
3028 /* IP address also allows 1 and 2 bytes access */
3031 __u32 src_port; /* host byte order */
3032 __u32 dst_port; /* network byte order */
3038 struct bpf_tcp_sock {
3039 __u32 snd_cwnd; /* Sending congestion window */
3040 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3042 __u32 snd_ssthresh; /* Slow start size threshold */
3043 __u32 rcv_nxt; /* What we want to receive next */
3044 __u32 snd_nxt; /* Next sequence we send */
3045 __u32 snd_una; /* First byte we want an ack for */
3046 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3047 __u32 ecn_flags; /* ECN status bits. */
3048 __u32 rate_delivered; /* saved rate sample: packets delivered */
3049 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3050 __u32 packets_out; /* Packets which are "in flight" */
3051 __u32 retrans_out; /* Retransmitted packets out */
3052 __u32 total_retrans; /* Total retransmits for entire connection */
3053 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3054 * total number of segments in.
3056 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3057 * total number of data segments in.
3059 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3060 * The total number of segments sent.
3062 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3063 * total number of data segments sent.
3065 __u32 lost_out; /* Lost packets */
3066 __u32 sacked_out; /* SACK'd packets */
3067 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3068 * sum(delta(rcv_nxt)), or how many bytes
3071 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3072 * sum(delta(snd_una)), or how many bytes
3077 struct bpf_sock_tuple {
3094 struct bpf_xdp_sock {
3098 #define XDP_PACKET_HEADROOM 256
3100 /* User return codes for XDP prog type.
3101 * A valid XDP program must return one of these defined values. All other
3102 * return codes are reserved for future use. Unknown return codes will
3103 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3113 /* user accessible metadata for XDP packet hook
3114 * new fields must be added to the end of this structure
3120 /* Below access go through struct xdp_rxq_info */
3121 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3122 __u32 rx_queue_index; /* rxq->queue_index */
3130 /* user accessible metadata for SK_MSG packet hook, new fields must
3131 * be added to the end of this structure
3134 __bpf_md_ptr(void *, data);
3135 __bpf_md_ptr(void *, data_end);
3138 __u32 remote_ip4; /* Stored in network byte order */
3139 __u32 local_ip4; /* Stored in network byte order */
3140 __u32 remote_ip6[4]; /* Stored in network byte order */
3141 __u32 local_ip6[4]; /* Stored in network byte order */
3142 __u32 remote_port; /* Stored in network byte order */
3143 __u32 local_port; /* stored in host byte order */
3144 __u32 size; /* Total size of sk_msg */
3147 struct sk_reuseport_md {
3149 * Start of directly accessible data. It begins from
3150 * the tcp/udp header.
3152 __bpf_md_ptr(void *, data);
3153 /* End of directly accessible data */
3154 __bpf_md_ptr(void *, data_end);
3156 * Total length of packet (starting from the tcp/udp header).
3157 * Note that the directly accessible bytes (data_end - data)
3158 * could be less than this "len". Those bytes could be
3159 * indirectly read by a helper "bpf_skb_load_bytes()".
3163 * Eth protocol in the mac header (network byte order). e.g.
3164 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3167 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3168 __u32 bind_inany; /* Is sock bound to an INANY address? */
3169 __u32 hash; /* A hash of the packet 4 tuples */
3172 #define BPF_TAG_SIZE 8
3174 struct bpf_prog_info {
3177 __u8 tag[BPF_TAG_SIZE];
3178 __u32 jited_prog_len;
3179 __u32 xlated_prog_len;
3180 __aligned_u64 jited_prog_insns;
3181 __aligned_u64 xlated_prog_insns;
3182 __u64 load_time; /* ns since boottime */
3183 __u32 created_by_uid;
3185 __aligned_u64 map_ids;
3186 char name[BPF_OBJ_NAME_LEN];
3188 __u32 gpl_compatible:1;
3191 __u32 nr_jited_ksyms;
3192 __u32 nr_jited_func_lens;
3193 __aligned_u64 jited_ksyms;
3194 __aligned_u64 jited_func_lens;
3196 __u32 func_info_rec_size;
3197 __aligned_u64 func_info;
3200 __aligned_u64 line_info;
3201 __aligned_u64 jited_line_info;
3202 __u32 nr_jited_line_info;
3203 __u32 line_info_rec_size;
3204 __u32 jited_line_info_rec_size;
3206 __aligned_u64 prog_tags;
3209 } __attribute__((aligned(8)));
3211 struct bpf_map_info {
3218 char name[BPF_OBJ_NAME_LEN];
3224 __u32 btf_key_type_id;
3225 __u32 btf_value_type_id;
3226 } __attribute__((aligned(8)));
3228 struct bpf_btf_info {
3232 } __attribute__((aligned(8)));
3234 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3235 * by user and intended to be used by socket (e.g. to bind to, depends on
3236 * attach attach type).
3238 struct bpf_sock_addr {
3239 __u32 user_family; /* Allows 4-byte read, but no write. */
3240 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3241 * Stored in network byte order.
3243 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
3244 * Stored in network byte order.
3246 __u32 user_port; /* Allows 4-byte read and write.
3247 * Stored in network byte order
3249 __u32 family; /* Allows 4-byte read, but no write */
3250 __u32 type; /* Allows 4-byte read, but no write */
3251 __u32 protocol; /* Allows 4-byte read, but no write */
3252 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
3253 * Stored in network byte order.
3255 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
3256 * Stored in network byte order.
3258 __bpf_md_ptr(struct bpf_sock *, sk);
3261 /* User bpf_sock_ops struct to access socket values and specify request ops
3262 * and their replies.
3263 * Some of this fields are in network (bigendian) byte order and may need
3264 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3265 * New fields can only be added at the end of this structure
3267 struct bpf_sock_ops {
3270 __u32 args[4]; /* Optionally passed to bpf program */
3271 __u32 reply; /* Returned by bpf program */
3272 __u32 replylong[4]; /* Optionally returned by bpf prog */
3275 __u32 remote_ip4; /* Stored in network byte order */
3276 __u32 local_ip4; /* Stored in network byte order */
3277 __u32 remote_ip6[4]; /* Stored in network byte order */
3278 __u32 local_ip6[4]; /* Stored in network byte order */
3279 __u32 remote_port; /* Stored in network byte order */
3280 __u32 local_port; /* stored in host byte order */
3281 __u32 is_fullsock; /* Some TCP fields are only valid if
3282 * there is a full socket. If not, the
3283 * fields read as zero.
3286 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
3287 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3296 __u32 rate_delivered;
3297 __u32 rate_interval_us;
3300 __u32 total_retrans;
3304 __u32 data_segs_out;
3308 __u64 bytes_received;
3310 __bpf_md_ptr(struct bpf_sock *, sk);
3313 /* Definitions for bpf_sock_ops_cb_flags */
3314 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
3315 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
3316 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3317 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
3318 * supported cb flags
3321 /* List of known BPF sock_ops operators.
3322 * New entries can only be added at the end
3326 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3327 * -1 if default value should be used
3329 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3330 * window (in packets) or -1 if default
3331 * value should be used
3333 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3334 * active connection is initialized
3336 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3337 * active connection is
3340 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3341 * passive connection is
3344 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3347 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3348 * based on the path and may be
3349 * dependent on the congestion control
3350 * algorithm. In general it indicates
3351 * a congestion threshold. RTTs above
3352 * this indicate congestion
3354 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3355 * Arg1: value of icsk_retransmits
3356 * Arg2: value of icsk_rto
3357 * Arg3: whether RTO has expired
3359 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3360 * Arg1: sequence number of 1st byte
3362 * Arg3: return value of
3363 * tcp_transmit_skb (0 => success)
3365 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3369 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3370 * socket transition to LISTEN state.
3374 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3375 * changes between the TCP and BPF versions. Ideally this should never happen.
3376 * If it does, we need to add code to convert them before calling
3377 * the BPF sock_ops function.
3380 BPF_TCP_ESTABLISHED = 1,
3390 BPF_TCP_CLOSING, /* Now a valid state */
3391 BPF_TCP_NEW_SYN_RECV,
3393 BPF_TCP_MAX_STATES /* Leave at the end! */
3396 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3397 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3399 struct bpf_perf_event_value {
3405 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3406 #define BPF_DEVCG_ACC_READ (1ULL << 1)
3407 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3409 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3410 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3412 struct bpf_cgroup_dev_ctx {
3413 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3419 struct bpf_raw_tracepoint_args {
3423 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3424 * OUTPUT: Do lookup from egress perspective; default is ingress
3426 #define BPF_FIB_LOOKUP_DIRECT (1U << 0)
3427 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
3430 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3431 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3432 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3433 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3434 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3435 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3436 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3437 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3438 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3441 struct bpf_fib_lookup {
3442 /* input: network family for lookup (AF_INET, AF_INET6)
3443 * output: network family of egress nexthop
3447 /* set if lookup is to consider L4 data - e.g., FIB rules */
3452 /* total length of packet from network header - used for MTU check */
3455 /* input: L3 device index for lookup
3456 * output: device index from FIB lookup
3461 /* inputs to lookup */
3462 __u8 tos; /* AF_INET */
3463 __be32 flowinfo; /* AF_INET6, flow_label + priority */
3465 /* output: metric of fib result (IPv4/IPv6 only) */
3471 __u32 ipv6_src[4]; /* in6_addr; network order */
3474 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3475 * network header. output: bpf_fib_lookup sets to gateway address
3476 * if FIB lookup returns gateway route
3480 __u32 ipv6_dst[4]; /* in6_addr; network order */
3484 __be16 h_vlan_proto;
3486 __u8 smac[6]; /* ETH_ALEN */
3487 __u8 dmac[6]; /* ETH_ALEN */
3490 enum bpf_task_fd_type {
3491 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3492 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3493 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3494 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3495 BPF_FD_TYPE_UPROBE, /* filename + offset */
3496 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3499 struct bpf_flow_keys {
3502 __u16 addr_proto; /* ETH_P_* of valid addrs */
3516 __u32 ipv6_src[4]; /* in6_addr; network order */
3517 __u32 ipv6_dst[4]; /* in6_addr; network order */
3522 struct bpf_func_info {
3527 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3528 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3530 struct bpf_line_info {
3532 __u32 file_name_off;
3537 struct bpf_spin_lock {
3542 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3543 * Allows 1,2,4-byte read, but no write.
3545 __u32 file_pos; /* Sysctl file position to read from, write to.
3546 * Allows 1,2,4-byte read an 4-byte write.
3550 struct bpf_sockopt {
3551 __bpf_md_ptr(struct bpf_sock *, sk);
3552 __bpf_md_ptr(void *, optval);
3553 __bpf_md_ptr(void *, optval_end);
3561 #endif /* _UAPI__LINUX_BPF_H__ */