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_ALU64 0x07 /* alu mode in double word width */
20 #define BPF_DW 0x18 /* double word (64-bit) */
21 #define BPF_XADD 0xc0 /* exclusive add */
24 #define BPF_MOV 0xb0 /* mov reg to reg */
25 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
27 /* change endianness of a register */
28 #define BPF_END 0xd0 /* flags for endianness conversion: */
29 #define BPF_TO_LE 0x00 /* convert to little-endian */
30 #define BPF_TO_BE 0x08 /* convert to big-endian */
31 #define BPF_FROM_LE BPF_TO_LE
32 #define BPF_FROM_BE BPF_TO_BE
35 #define BPF_JNE 0x50 /* jump != */
36 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
37 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
38 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
39 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
40 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
41 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
42 #define BPF_CALL 0x80 /* function call */
43 #define BPF_EXIT 0x90 /* function return */
45 /* Register numbers */
61 /* BPF has 10 general purpose 64-bit registers and stack frame. */
62 #define MAX_BPF_REG __MAX_BPF_REG
65 __u8 code; /* opcode */
66 __u8 dst_reg:4; /* dest register */
67 __u8 src_reg:4; /* source register */
68 __s16 off; /* signed offset */
69 __s32 imm; /* signed immediate constant */
72 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
73 struct bpf_lpm_trie_key {
74 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
75 __u8 data[0]; /* Arbitrary size */
78 struct bpf_cgroup_storage_key {
79 __u64 cgroup_inode_id; /* cgroup inode id */
80 __u32 attach_type; /* program attach type */
83 /* BPF syscall commands, see bpf(2) man-page for details. */
98 BPF_PROG_GET_FD_BY_ID,
100 BPF_OBJ_GET_INFO_BY_FD,
102 BPF_RAW_TRACEPOINT_OPEN,
104 BPF_BTF_GET_FD_BY_ID,
106 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
113 BPF_MAP_TYPE_PROG_ARRAY,
114 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
115 BPF_MAP_TYPE_PERCPU_HASH,
116 BPF_MAP_TYPE_PERCPU_ARRAY,
117 BPF_MAP_TYPE_STACK_TRACE,
118 BPF_MAP_TYPE_CGROUP_ARRAY,
119 BPF_MAP_TYPE_LRU_HASH,
120 BPF_MAP_TYPE_LRU_PERCPU_HASH,
121 BPF_MAP_TYPE_LPM_TRIE,
122 BPF_MAP_TYPE_ARRAY_OF_MAPS,
123 BPF_MAP_TYPE_HASH_OF_MAPS,
125 BPF_MAP_TYPE_SOCKMAP,
128 BPF_MAP_TYPE_SOCKHASH,
129 BPF_MAP_TYPE_CGROUP_STORAGE,
130 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
131 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
137 BPF_PROG_TYPE_UNSPEC,
138 BPF_PROG_TYPE_SOCKET_FILTER,
139 BPF_PROG_TYPE_KPROBE,
140 BPF_PROG_TYPE_SCHED_CLS,
141 BPF_PROG_TYPE_SCHED_ACT,
142 BPF_PROG_TYPE_TRACEPOINT,
144 BPF_PROG_TYPE_PERF_EVENT,
145 BPF_PROG_TYPE_CGROUP_SKB,
146 BPF_PROG_TYPE_CGROUP_SOCK,
147 BPF_PROG_TYPE_LWT_IN,
148 BPF_PROG_TYPE_LWT_OUT,
149 BPF_PROG_TYPE_LWT_XMIT,
150 BPF_PROG_TYPE_SOCK_OPS,
151 BPF_PROG_TYPE_SK_SKB,
152 BPF_PROG_TYPE_CGROUP_DEVICE,
153 BPF_PROG_TYPE_SK_MSG,
154 BPF_PROG_TYPE_RAW_TRACEPOINT,
155 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
156 BPF_PROG_TYPE_LWT_SEG6LOCAL,
157 BPF_PROG_TYPE_LIRC_MODE2,
158 BPF_PROG_TYPE_SK_REUSEPORT,
159 BPF_PROG_TYPE_FLOW_DISSECTOR,
162 enum bpf_attach_type {
163 BPF_CGROUP_INET_INGRESS,
164 BPF_CGROUP_INET_EGRESS,
165 BPF_CGROUP_INET_SOCK_CREATE,
167 BPF_SK_SKB_STREAM_PARSER,
168 BPF_SK_SKB_STREAM_VERDICT,
171 BPF_CGROUP_INET4_BIND,
172 BPF_CGROUP_INET6_BIND,
173 BPF_CGROUP_INET4_CONNECT,
174 BPF_CGROUP_INET6_CONNECT,
175 BPF_CGROUP_INET4_POST_BIND,
176 BPF_CGROUP_INET6_POST_BIND,
177 BPF_CGROUP_UDP4_SENDMSG,
178 BPF_CGROUP_UDP6_SENDMSG,
181 __MAX_BPF_ATTACH_TYPE
184 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
186 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
188 * NONE(default): No further bpf programs allowed in the subtree.
190 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
191 * the program in this cgroup yields to sub-cgroup program.
193 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
194 * that cgroup program gets run in addition to the program in this cgroup.
196 * Only one program is allowed to be attached to a cgroup with
197 * NONE or BPF_F_ALLOW_OVERRIDE flag.
198 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
199 * release old program and attach the new one. Attach flags has to match.
201 * Multiple programs are allowed to be attached to a cgroup with
202 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
203 * (those that were attached first, run first)
204 * The programs of sub-cgroup are executed first, then programs of
205 * this cgroup and then programs of parent cgroup.
206 * When children program makes decision (like picking TCP CA or sock bind)
207 * parent program has a chance to override it.
209 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
210 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
212 * cgrp1 (MULTI progs A, B) ->
213 * cgrp2 (OVERRIDE prog C) ->
214 * cgrp3 (MULTI prog D) ->
215 * cgrp4 (OVERRIDE prog E) ->
216 * cgrp5 (NONE prog F)
217 * the event in cgrp5 triggers execution of F,D,A,B in that order.
218 * if prog F is detached, the execution is E,D,A,B
219 * if prog F and D are detached, the execution is E,A,B
220 * if prog F, E and D are detached, the execution is C,A,B
222 * All eligible programs are executed regardless of return code from
225 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
226 #define BPF_F_ALLOW_MULTI (1U << 1)
228 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
229 * verifier will perform strict alignment checking as if the kernel
230 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
231 * and NET_IP_ALIGN defined to 2.
233 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
235 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
236 * verifier will allow any alignment whatsoever. On platforms
237 * with strict alignment requirements for loads ands stores (such
238 * as sparc and mips) the verifier validates that all loads and
239 * stores provably follow this requirement. This flag turns that
240 * checking and enforcement off.
242 * It is mostly used for testing when we want to validate the
243 * context and memory access aspects of the verifier, but because
244 * of an unaligned access the alignment check would trigger before
245 * the one we are interested in.
247 #define BPF_F_ANY_ALIGNMENT (1U << 1)
249 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
250 #define BPF_PSEUDO_MAP_FD 1
252 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
253 * offset to another bpf function
255 #define BPF_PSEUDO_CALL 1
257 /* flags for BPF_MAP_UPDATE_ELEM command */
258 #define BPF_ANY 0 /* create new element or update existing */
259 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
260 #define BPF_EXIST 2 /* update existing element */
262 /* flags for BPF_MAP_CREATE command */
263 #define BPF_F_NO_PREALLOC (1U << 0)
264 /* Instead of having one common LRU list in the
265 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
266 * which can scale and perform better.
267 * Note, the LRU nodes (including free nodes) cannot be moved
268 * across different LRU lists.
270 #define BPF_F_NO_COMMON_LRU (1U << 1)
271 /* Specify numa node during map creation */
272 #define BPF_F_NUMA_NODE (1U << 2)
274 #define BPF_OBJ_NAME_LEN 16U
276 /* Flags for accessing BPF object */
277 #define BPF_F_RDONLY (1U << 3)
278 #define BPF_F_WRONLY (1U << 4)
280 /* Flag for stack_map, store build_id+offset instead of pointer */
281 #define BPF_F_STACK_BUILD_ID (1U << 5)
283 /* Zero-initialize hash function seed. This should only be used for testing. */
284 #define BPF_F_ZERO_SEED (1U << 6)
286 /* flags for BPF_PROG_QUERY */
287 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
289 enum bpf_stack_build_id_status {
290 /* user space need an empty entry to identify end of a trace */
291 BPF_STACK_BUILD_ID_EMPTY = 0,
292 /* with valid build_id and offset */
293 BPF_STACK_BUILD_ID_VALID = 1,
294 /* couldn't get build_id, fallback to ip */
295 BPF_STACK_BUILD_ID_IP = 2,
298 #define BPF_BUILD_ID_SIZE 20
299 struct bpf_stack_build_id {
301 unsigned char build_id[BPF_BUILD_ID_SIZE];
309 struct { /* anonymous struct used by BPF_MAP_CREATE command */
310 __u32 map_type; /* one of enum bpf_map_type */
311 __u32 key_size; /* size of key in bytes */
312 __u32 value_size; /* size of value in bytes */
313 __u32 max_entries; /* max number of entries in a map */
314 __u32 map_flags; /* BPF_MAP_CREATE related
315 * flags defined above.
317 __u32 inner_map_fd; /* fd pointing to the inner map */
318 __u32 numa_node; /* numa node (effective only if
319 * BPF_F_NUMA_NODE is set).
321 char map_name[BPF_OBJ_NAME_LEN];
322 __u32 map_ifindex; /* ifindex of netdev to create on */
323 __u32 btf_fd; /* fd pointing to a BTF type data */
324 __u32 btf_key_type_id; /* BTF type_id of the key */
325 __u32 btf_value_type_id; /* BTF type_id of the value */
328 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
333 __aligned_u64 next_key;
338 struct { /* anonymous struct used by BPF_PROG_LOAD command */
339 __u32 prog_type; /* one of enum bpf_prog_type */
342 __aligned_u64 license;
343 __u32 log_level; /* verbosity level of verifier */
344 __u32 log_size; /* size of user buffer */
345 __aligned_u64 log_buf; /* user supplied buffer */
346 __u32 kern_version; /* checked when prog_type=kprobe */
348 char prog_name[BPF_OBJ_NAME_LEN];
349 __u32 prog_ifindex; /* ifindex of netdev to prep for */
350 /* For some prog types expected attach type must be known at
351 * load time to verify attach type specific parts of prog
352 * (context accesses, allowed helpers, etc).
354 __u32 expected_attach_type;
355 __u32 prog_btf_fd; /* fd pointing to BTF type data */
356 __u32 func_info_rec_size; /* userspace bpf_func_info size */
357 __aligned_u64 func_info; /* func info */
358 __u32 func_info_cnt; /* number of bpf_func_info records */
361 struct { /* anonymous struct used by BPF_OBJ_* commands */
362 __aligned_u64 pathname;
367 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
368 __u32 target_fd; /* container object to attach to */
369 __u32 attach_bpf_fd; /* eBPF program to attach */
374 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
379 __aligned_u64 data_in;
380 __aligned_u64 data_out;
385 struct { /* anonymous struct used by BPF_*_GET_*_ID */
396 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
402 struct { /* anonymous struct used by BPF_PROG_QUERY command */
403 __u32 target_fd; /* container object to query */
407 __aligned_u64 prog_ids;
416 struct { /* anonymous struct for BPF_BTF_LOAD */
418 __aligned_u64 btf_log_buf;
425 __u32 pid; /* input: pid */
426 __u32 fd; /* input: fd */
427 __u32 flags; /* input: flags */
428 __u32 buf_len; /* input/output: buf len */
429 __aligned_u64 buf; /* input/output:
430 * tp_name for tracepoint
432 * filename for uprobe
434 __u32 prog_id; /* output: prod_id */
435 __u32 fd_type; /* output: BPF_FD_TYPE_* */
436 __u64 probe_offset; /* output: probe_offset */
437 __u64 probe_addr; /* output: probe_addr */
439 } __attribute__((aligned(8)));
441 /* The description below is an attempt at providing documentation to eBPF
442 * developers about the multiple available eBPF helper functions. It can be
443 * parsed and used to produce a manual page. The workflow is the following,
444 * and requires the rst2man utility:
446 * $ ./scripts/bpf_helpers_doc.py \
447 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
448 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
449 * $ man /tmp/bpf-helpers.7
451 * Note that in order to produce this external documentation, some RST
452 * formatting is used in the descriptions to get "bold" and "italics" in
453 * manual pages. Also note that the few trailing white spaces are
454 * intentional, removing them would break paragraphs for rst2man.
456 * Start of BPF helper function descriptions:
458 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
460 * Perform a lookup in *map* for an entry associated to *key*.
462 * Map value associated to *key*, or **NULL** if no entry was
465 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
467 * Add or update the value of the entry associated to *key* in
468 * *map* with *value*. *flags* is one of:
471 * The entry for *key* must not exist in the map.
473 * The entry for *key* must already exist in the map.
475 * No condition on the existence of the entry for *key*.
477 * Flag value **BPF_NOEXIST** cannot be used for maps of types
478 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
479 * elements always exist), the helper would return an error.
481 * 0 on success, or a negative error in case of failure.
483 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
485 * Delete entry with *key* from *map*.
487 * 0 on success, or a negative error in case of failure.
489 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
491 * Push an element *value* in *map*. *flags* is one of:
494 * If the queue/stack is full, the oldest element is removed to
495 * make room for this.
497 * 0 on success, or a negative error in case of failure.
499 * int bpf_probe_read(void *dst, u32 size, const void *src)
501 * For tracing programs, safely attempt to read *size* bytes from
502 * address *src* and store the data in *dst*.
504 * 0 on success, or a negative error in case of failure.
506 * u64 bpf_ktime_get_ns(void)
508 * Return the time elapsed since system boot, in nanoseconds.
512 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
514 * This helper is a "printk()-like" facility for debugging. It
515 * prints a message defined by format *fmt* (of size *fmt_size*)
516 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
517 * available. It can take up to three additional **u64**
518 * arguments (as an eBPF helpers, the total number of arguments is
521 * Each time the helper is called, it appends a line to the trace.
522 * The format of the trace is customizable, and the exact output
523 * one will get depends on the options set in
524 * *\/sys/kernel/debug/tracing/trace_options* (see also the
525 * *README* file under the same directory). However, it usually
526 * defaults to something like:
530 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
534 * * ``telnet`` is the name of the current task.
535 * * ``470`` is the PID of the current task.
536 * * ``001`` is the CPU number on which the task is
538 * * In ``.N..``, each character refers to a set of
539 * options (whether irqs are enabled, scheduling
540 * options, whether hard/softirqs are running, level of
541 * preempt_disabled respectively). **N** means that
542 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
544 * * ``419421.045894`` is a timestamp.
545 * * ``0x00000001`` is a fake value used by BPF for the
546 * instruction pointer register.
547 * * ``<formatted msg>`` is the message formatted with
550 * The conversion specifiers supported by *fmt* are similar, but
551 * more limited than for printk(). They are **%d**, **%i**,
552 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
553 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
554 * of field, padding with zeroes, etc.) is available, and the
555 * helper will return **-EINVAL** (but print nothing) if it
556 * encounters an unknown specifier.
558 * Also, note that **bpf_trace_printk**\ () is slow, and should
559 * only be used for debugging purposes. For this reason, a notice
560 * bloc (spanning several lines) is printed to kernel logs and
561 * states that the helper should not be used "for production use"
562 * the first time this helper is used (or more precisely, when
563 * **trace_printk**\ () buffers are allocated). For passing values
564 * to user space, perf events should be preferred.
566 * The number of bytes written to the buffer, or a negative error
567 * in case of failure.
569 * u32 bpf_get_prandom_u32(void)
571 * Get a pseudo-random number.
573 * From a security point of view, this helper uses its own
574 * pseudo-random internal state, and cannot be used to infer the
575 * seed of other random functions in the kernel. However, it is
576 * essential to note that the generator used by the helper is not
577 * cryptographically secure.
579 * A random 32-bit unsigned value.
581 * u32 bpf_get_smp_processor_id(void)
583 * Get the SMP (symmetric multiprocessing) processor id. Note that
584 * all programs run with preemption disabled, which means that the
585 * SMP processor id is stable during all the execution of the
588 * The SMP id of the processor running the program.
590 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
592 * Store *len* bytes from address *from* into the packet
593 * associated to *skb*, at *offset*. *flags* are a combination of
594 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
595 * checksum for the packet after storing the bytes) and
596 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
597 * **->swhash** and *skb*\ **->l4hash** to 0).
599 * A call to this helper is susceptible to change the underlaying
600 * packet buffer. Therefore, at load time, all checks on pointers
601 * previously done by the verifier are invalidated and must be
602 * performed again, if the helper is used in combination with
603 * direct packet access.
605 * 0 on success, or a negative error in case of failure.
607 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
609 * Recompute the layer 3 (e.g. IP) checksum for the packet
610 * associated to *skb*. Computation is incremental, so the helper
611 * must know the former value of the header field that was
612 * modified (*from*), the new value of this field (*to*), and the
613 * number of bytes (2 or 4) for this field, stored in *size*.
614 * Alternatively, it is possible to store the difference between
615 * the previous and the new values of the header field in *to*, by
616 * setting *from* and *size* to 0. For both methods, *offset*
617 * indicates the location of the IP checksum within the packet.
619 * This helper works in combination with **bpf_csum_diff**\ (),
620 * which does not update the checksum in-place, but offers more
621 * flexibility and can handle sizes larger than 2 or 4 for the
622 * checksum to update.
624 * A call to this helper is susceptible to change the underlaying
625 * packet buffer. Therefore, at load time, all checks on pointers
626 * previously done by the verifier are invalidated and must be
627 * performed again, if the helper is used in combination with
628 * direct packet access.
630 * 0 on success, or a negative error in case of failure.
632 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
634 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
635 * packet associated to *skb*. Computation is incremental, so the
636 * helper must know the former value of the header field that was
637 * modified (*from*), the new value of this field (*to*), and the
638 * number of bytes (2 or 4) for this field, stored on the lowest
639 * four bits of *flags*. Alternatively, it is possible to store
640 * the difference between the previous and the new values of the
641 * header field in *to*, by setting *from* and the four lowest
642 * bits of *flags* to 0. For both methods, *offset* indicates the
643 * location of the IP checksum within the packet. In addition to
644 * the size of the field, *flags* can be added (bitwise OR) actual
645 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
646 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
647 * for updates resulting in a null checksum the value is set to
648 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
649 * the checksum is to be computed against a pseudo-header.
651 * This helper works in combination with **bpf_csum_diff**\ (),
652 * which does not update the checksum in-place, but offers more
653 * flexibility and can handle sizes larger than 2 or 4 for the
654 * checksum to update.
656 * A call to this helper is susceptible to change the underlaying
657 * packet buffer. Therefore, at load time, all checks on pointers
658 * previously done by the verifier are invalidated and must be
659 * performed again, if the helper is used in combination with
660 * direct packet access.
662 * 0 on success, or a negative error in case of failure.
664 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
666 * This special helper is used to trigger a "tail call", or in
667 * other words, to jump into another eBPF program. The same stack
668 * frame is used (but values on stack and in registers for the
669 * caller are not accessible to the callee). This mechanism allows
670 * for program chaining, either for raising the maximum number of
671 * available eBPF instructions, or to execute given programs in
672 * conditional blocks. For security reasons, there is an upper
673 * limit to the number of successive tail calls that can be
676 * Upon call of this helper, the program attempts to jump into a
677 * program referenced at index *index* in *prog_array_map*, a
678 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
679 * *ctx*, a pointer to the context.
681 * If the call succeeds, the kernel immediately runs the first
682 * instruction of the new program. This is not a function call,
683 * and it never returns to the previous program. If the call
684 * fails, then the helper has no effect, and the caller continues
685 * to run its subsequent instructions. A call can fail if the
686 * destination program for the jump does not exist (i.e. *index*
687 * is superior to the number of entries in *prog_array_map*), or
688 * if the maximum number of tail calls has been reached for this
689 * chain of programs. This limit is defined in the kernel by the
690 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
691 * which is currently set to 32.
693 * 0 on success, or a negative error in case of failure.
695 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
697 * Clone and redirect the packet associated to *skb* to another
698 * net device of index *ifindex*. Both ingress and egress
699 * interfaces can be used for redirection. The **BPF_F_INGRESS**
700 * value in *flags* is used to make the distinction (ingress path
701 * is selected if the flag is present, egress path otherwise).
702 * This is the only flag supported for now.
704 * In comparison with **bpf_redirect**\ () helper,
705 * **bpf_clone_redirect**\ () has the associated cost of
706 * duplicating the packet buffer, but this can be executed out of
707 * the eBPF program. Conversely, **bpf_redirect**\ () is more
708 * efficient, but it is handled through an action code where the
709 * redirection happens only after the eBPF program has returned.
711 * A call to this helper is susceptible to change the underlaying
712 * packet buffer. Therefore, at load time, all checks on pointers
713 * previously done by the verifier are invalidated and must be
714 * performed again, if the helper is used in combination with
715 * direct packet access.
717 * 0 on success, or a negative error in case of failure.
719 * u64 bpf_get_current_pid_tgid(void)
721 * A 64-bit integer containing the current tgid and pid, and
723 * *current_task*\ **->tgid << 32 \|**
724 * *current_task*\ **->pid**.
726 * u64 bpf_get_current_uid_gid(void)
728 * A 64-bit integer containing the current GID and UID, and
729 * created as such: *current_gid* **<< 32 \|** *current_uid*.
731 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
733 * Copy the **comm** attribute of the current task into *buf* of
734 * *size_of_buf*. The **comm** attribute contains the name of
735 * the executable (excluding the path) for the current task. The
736 * *size_of_buf* must be strictly positive. On success, the
737 * helper makes sure that the *buf* is NUL-terminated. On failure,
738 * it is filled with zeroes.
740 * 0 on success, or a negative error in case of failure.
742 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
744 * Retrieve the classid for the current task, i.e. for the net_cls
745 * cgroup to which *skb* belongs.
747 * This helper can be used on TC egress path, but not on ingress.
749 * The net_cls cgroup provides an interface to tag network packets
750 * based on a user-provided identifier for all traffic coming from
751 * the tasks belonging to the related cgroup. See also the related
752 * kernel documentation, available from the Linux sources in file
753 * *Documentation/cgroup-v1/net_cls.txt*.
755 * The Linux kernel has two versions for cgroups: there are
756 * cgroups v1 and cgroups v2. Both are available to users, who can
757 * use a mixture of them, but note that the net_cls cgroup is for
758 * cgroup v1 only. This makes it incompatible with BPF programs
759 * run on cgroups, which is a cgroup-v2-only feature (a socket can
760 * only hold data for one version of cgroups at a time).
762 * This helper is only available is the kernel was compiled with
763 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
764 * "**y**" or to "**m**".
766 * The classid, or 0 for the default unconfigured classid.
768 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
770 * Push a *vlan_tci* (VLAN tag control information) of protocol
771 * *vlan_proto* to the packet associated to *skb*, then update
772 * the checksum. Note that if *vlan_proto* is different from
773 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
774 * be **ETH_P_8021Q**.
776 * A call to this helper is susceptible to change the underlaying
777 * packet buffer. Therefore, at load time, all checks on pointers
778 * previously done by the verifier are invalidated and must be
779 * performed again, if the helper is used in combination with
780 * direct packet access.
782 * 0 on success, or a negative error in case of failure.
784 * int bpf_skb_vlan_pop(struct sk_buff *skb)
786 * Pop a VLAN header from the packet associated to *skb*.
788 * A call to this helper is susceptible to change the underlaying
789 * packet buffer. Therefore, at load time, all checks on pointers
790 * previously done by the verifier are invalidated and must be
791 * performed again, if the helper is used in combination with
792 * direct packet access.
794 * 0 on success, or a negative error in case of failure.
796 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
798 * Get tunnel metadata. This helper takes a pointer *key* to an
799 * empty **struct bpf_tunnel_key** of **size**, that will be
800 * filled with tunnel metadata for the packet associated to *skb*.
801 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
802 * indicates that the tunnel is based on IPv6 protocol instead of
805 * The **struct bpf_tunnel_key** is an object that generalizes the
806 * principal parameters used by various tunneling protocols into a
807 * single struct. This way, it can be used to easily make a
808 * decision based on the contents of the encapsulation header,
809 * "summarized" in this struct. In particular, it holds the IP
810 * address of the remote end (IPv4 or IPv6, depending on the case)
811 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
812 * this struct exposes the *key*\ **->tunnel_id**, which is
813 * generally mapped to a VNI (Virtual Network Identifier), making
814 * it programmable together with the **bpf_skb_set_tunnel_key**\
817 * Let's imagine that the following code is part of a program
818 * attached to the TC ingress interface, on one end of a GRE
819 * tunnel, and is supposed to filter out all messages coming from
820 * remote ends with IPv4 address other than 10.0.0.1:
825 * struct bpf_tunnel_key key = {};
827 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
829 * return TC_ACT_SHOT; // drop packet
831 * if (key.remote_ipv4 != 0x0a000001)
832 * return TC_ACT_SHOT; // drop packet
834 * return TC_ACT_OK; // accept packet
836 * This interface can also be used with all encapsulation devices
837 * that can operate in "collect metadata" mode: instead of having
838 * one network device per specific configuration, the "collect
839 * metadata" mode only requires a single device where the
840 * configuration can be extracted from this helper.
842 * This can be used together with various tunnels such as VXLan,
843 * Geneve, GRE or IP in IP (IPIP).
845 * 0 on success, or a negative error in case of failure.
847 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
849 * Populate tunnel metadata for packet associated to *skb.* The
850 * tunnel metadata is set to the contents of *key*, of *size*. The
851 * *flags* can be set to a combination of the following values:
853 * **BPF_F_TUNINFO_IPV6**
854 * Indicate that the tunnel is based on IPv6 protocol
856 * **BPF_F_ZERO_CSUM_TX**
857 * For IPv4 packets, add a flag to tunnel metadata
858 * indicating that checksum computation should be skipped
859 * and checksum set to zeroes.
860 * **BPF_F_DONT_FRAGMENT**
861 * Add a flag to tunnel metadata indicating that the
862 * packet should not be fragmented.
863 * **BPF_F_SEQ_NUMBER**
864 * Add a flag to tunnel metadata indicating that a
865 * sequence number should be added to tunnel header before
866 * sending the packet. This flag was added for GRE
867 * encapsulation, but might be used with other protocols
868 * as well in the future.
870 * Here is a typical usage on the transmit path:
874 * struct bpf_tunnel_key key;
876 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
877 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
879 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
880 * helper for additional information.
882 * 0 on success, or a negative error in case of failure.
884 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
886 * Read the value of a perf event counter. This helper relies on a
887 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
888 * the perf event counter is selected when *map* is updated with
889 * perf event file descriptors. The *map* is an array whose size
890 * is the number of available CPUs, and each cell contains a value
891 * relative to one CPU. The value to retrieve is indicated by
892 * *flags*, that contains the index of the CPU to look up, masked
893 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
894 * **BPF_F_CURRENT_CPU** to indicate that the value for the
895 * current CPU should be retrieved.
897 * Note that before Linux 4.13, only hardware perf event can be
900 * Also, be aware that the newer helper
901 * **bpf_perf_event_read_value**\ () is recommended over
902 * **bpf_perf_event_read**\ () in general. The latter has some ABI
903 * quirks where error and counter value are used as a return code
904 * (which is wrong to do since ranges may overlap). This issue is
905 * fixed with **bpf_perf_event_read_value**\ (), which at the same
906 * time provides more features over the **bpf_perf_event_read**\
907 * () interface. Please refer to the description of
908 * **bpf_perf_event_read_value**\ () for details.
910 * The value of the perf event counter read from the map, or a
911 * negative error code in case of failure.
913 * int bpf_redirect(u32 ifindex, u64 flags)
915 * Redirect the packet to another net device of index *ifindex*.
916 * This helper is somewhat similar to **bpf_clone_redirect**\
917 * (), except that the packet is not cloned, which provides
918 * increased performance.
920 * Except for XDP, both ingress and egress interfaces can be used
921 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
922 * to make the distinction (ingress path is selected if the flag
923 * is present, egress path otherwise). Currently, XDP only
924 * supports redirection to the egress interface, and accepts no
927 * The same effect can be attained with the more generic
928 * **bpf_redirect_map**\ (), which requires specific maps to be
929 * used but offers better performance.
931 * For XDP, the helper returns **XDP_REDIRECT** on success or
932 * **XDP_ABORTED** on error. For other program types, the values
933 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
936 * u32 bpf_get_route_realm(struct sk_buff *skb)
938 * Retrieve the realm or the route, that is to say the
939 * **tclassid** field of the destination for the *skb*. The
940 * indentifier retrieved is a user-provided tag, similar to the
941 * one used with the net_cls cgroup (see description for
942 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
943 * held by a route (a destination entry), not by a task.
945 * Retrieving this identifier works with the clsact TC egress hook
946 * (see also **tc-bpf(8)**), or alternatively on conventional
947 * classful egress qdiscs, but not on TC ingress path. In case of
948 * clsact TC egress hook, this has the advantage that, internally,
949 * the destination entry has not been dropped yet in the transmit
950 * path. Therefore, the destination entry does not need to be
951 * artificially held via **netif_keep_dst**\ () for a classful
952 * qdisc until the *skb* is freed.
954 * This helper is available only if the kernel was compiled with
955 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
957 * The realm of the route for the packet associated to *skb*, or 0
960 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
962 * Write raw *data* blob into a special BPF perf event held by
963 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
964 * event must have the following attributes: **PERF_SAMPLE_RAW**
965 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
966 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
968 * The *flags* are used to indicate the index in *map* for which
969 * the value must be put, masked with **BPF_F_INDEX_MASK**.
970 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
971 * to indicate that the index of the current CPU core should be
974 * The value to write, of *size*, is passed through eBPF stack and
977 * The context of the program *ctx* needs also be passed to the
980 * On user space, a program willing to read the values needs to
981 * call **perf_event_open**\ () on the perf event (either for
982 * one or for all CPUs) and to store the file descriptor into the
983 * *map*. This must be done before the eBPF program can send data
984 * into it. An example is available in file
985 * *samples/bpf/trace_output_user.c* in the Linux kernel source
986 * tree (the eBPF program counterpart is in
987 * *samples/bpf/trace_output_kern.c*).
989 * **bpf_perf_event_output**\ () achieves better performance
990 * than **bpf_trace_printk**\ () for sharing data with user
991 * space, and is much better suitable for streaming data from eBPF
994 * Note that this helper is not restricted to tracing use cases
995 * and can be used with programs attached to TC or XDP as well,
996 * where it allows for passing data to user space listeners. Data
999 * * Only custom structs,
1000 * * Only the packet payload, or
1001 * * A combination of both.
1003 * 0 on success, or a negative error in case of failure.
1005 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1007 * This helper was provided as an easy way to load data from a
1008 * packet. It can be used to load *len* bytes from *offset* from
1009 * the packet associated to *skb*, into the buffer pointed by
1012 * Since Linux 4.7, usage of this helper has mostly been replaced
1013 * by "direct packet access", enabling packet data to be
1014 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1015 * pointing respectively to the first byte of packet data and to
1016 * the byte after the last byte of packet data. However, it
1017 * remains useful if one wishes to read large quantities of data
1018 * at once from a packet into the eBPF stack.
1020 * 0 on success, or a negative error in case of failure.
1022 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1024 * Walk a user or a kernel stack and return its id. To achieve
1025 * this, the helper needs *ctx*, which is a pointer to the context
1026 * on which the tracing program is executed, and a pointer to a
1027 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1029 * The last argument, *flags*, holds the number of stack frames to
1030 * skip (from 0 to 255), masked with
1031 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1032 * a combination of the following flags:
1034 * **BPF_F_USER_STACK**
1035 * Collect a user space stack instead of a kernel stack.
1036 * **BPF_F_FAST_STACK_CMP**
1037 * Compare stacks by hash only.
1038 * **BPF_F_REUSE_STACKID**
1039 * If two different stacks hash into the same *stackid*,
1040 * discard the old one.
1042 * The stack id retrieved is a 32 bit long integer handle which
1043 * can be further combined with other data (including other stack
1044 * ids) and used as a key into maps. This can be useful for
1045 * generating a variety of graphs (such as flame graphs or off-cpu
1048 * For walking a stack, this helper is an improvement over
1049 * **bpf_probe_read**\ (), which can be used with unrolled loops
1050 * but is not efficient and consumes a lot of eBPF instructions.
1051 * Instead, **bpf_get_stackid**\ () can collect up to
1052 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1053 * this limit can be controlled with the **sysctl** program, and
1054 * that it should be manually increased in order to profile long
1055 * user stacks (such as stacks for Java programs). To do so, use:
1059 * # sysctl kernel.perf_event_max_stack=<new value>
1061 * The positive or null stack id on success, or a negative error
1062 * in case of failure.
1064 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1066 * Compute a checksum difference, from the raw buffer pointed by
1067 * *from*, of length *from_size* (that must be a multiple of 4),
1068 * towards the raw buffer pointed by *to*, of size *to_size*
1069 * (same remark). An optional *seed* can be added to the value
1070 * (this can be cascaded, the seed may come from a previous call
1073 * This is flexible enough to be used in several ways:
1075 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1076 * checksum, it can be used when pushing new data.
1077 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1078 * checksum, it can be used when removing data from a packet.
1079 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1080 * can be used to compute a diff. Note that *from_size* and
1081 * *to_size* do not need to be equal.
1083 * This helper can be used in combination with
1084 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1085 * which one can feed in the difference computed with
1086 * **bpf_csum_diff**\ ().
1088 * The checksum result, or a negative error code in case of
1091 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1093 * Retrieve tunnel options metadata for the packet associated to
1094 * *skb*, and store the raw tunnel option data to the buffer *opt*
1097 * This helper can be used with encapsulation devices that can
1098 * operate in "collect metadata" mode (please refer to the related
1099 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1100 * more details). A particular example where this can be used is
1101 * in combination with the Geneve encapsulation protocol, where it
1102 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1103 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1104 * the eBPF program. This allows for full customization of these
1107 * The size of the option data retrieved.
1109 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1111 * Set tunnel options metadata for the packet associated to *skb*
1112 * to the option data contained in the raw buffer *opt* of *size*.
1114 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1115 * helper for additional information.
1117 * 0 on success, or a negative error in case of failure.
1119 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1121 * Change the protocol of the *skb* to *proto*. Currently
1122 * supported are transition from IPv4 to IPv6, and from IPv6 to
1123 * IPv4. The helper takes care of the groundwork for the
1124 * transition, including resizing the socket buffer. The eBPF
1125 * program is expected to fill the new headers, if any, via
1126 * **skb_store_bytes**\ () and to recompute the checksums with
1127 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1128 * (). The main case for this helper is to perform NAT64
1129 * operations out of an eBPF program.
1131 * Internally, the GSO type is marked as dodgy so that headers are
1132 * checked and segments are recalculated by the GSO/GRO engine.
1133 * The size for GSO target is adapted as well.
1135 * All values for *flags* are reserved for future usage, and must
1138 * A call to this helper is susceptible to change the underlaying
1139 * packet buffer. Therefore, at load time, all checks on pointers
1140 * previously done by the verifier are invalidated and must be
1141 * performed again, if the helper is used in combination with
1142 * direct packet access.
1144 * 0 on success, or a negative error in case of failure.
1146 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1148 * Change the packet type for the packet associated to *skb*. This
1149 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1150 * the eBPF program does not have a write access to *skb*\
1151 * **->pkt_type** beside this helper. Using a helper here allows
1152 * for graceful handling of errors.
1154 * The major use case is to change incoming *skb*s to
1155 * **PACKET_HOST** in a programmatic way instead of having to
1156 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1159 * Note that *type* only allows certain values. At this time, they
1164 * **PACKET_BROADCAST**
1165 * Send packet to all.
1166 * **PACKET_MULTICAST**
1167 * Send packet to group.
1168 * **PACKET_OTHERHOST**
1169 * Send packet to someone else.
1171 * 0 on success, or a negative error in case of failure.
1173 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1175 * Check whether *skb* is a descendant of the cgroup2 held by
1176 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1178 * The return value depends on the result of the test, and can be:
1180 * * 0, if the *skb* failed the cgroup2 descendant test.
1181 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1182 * * A negative error code, if an error occurred.
1184 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1186 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1187 * not set, in particular if the hash was cleared due to mangling,
1188 * recompute this hash. Later accesses to the hash can be done
1189 * directly with *skb*\ **->hash**.
1191 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1192 * prototype with **bpf_skb_change_proto**\ (), or calling
1193 * **bpf_skb_store_bytes**\ () with the
1194 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1195 * the hash and to trigger a new computation for the next call to
1196 * **bpf_get_hash_recalc**\ ().
1200 * u64 bpf_get_current_task(void)
1202 * A pointer to the current task struct.
1204 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1206 * Attempt in a safe way to write *len* bytes from the buffer
1207 * *src* to *dst* in memory. It only works for threads that are in
1208 * user context, and *dst* must be a valid user space address.
1210 * This helper should not be used to implement any kind of
1211 * security mechanism because of TOC-TOU attacks, but rather to
1212 * debug, divert, and manipulate execution of semi-cooperative
1215 * Keep in mind that this feature is meant for experiments, and it
1216 * has a risk of crashing the system and running programs.
1217 * Therefore, when an eBPF program using this helper is attached,
1218 * a warning including PID and process name is printed to kernel
1221 * 0 on success, or a negative error in case of failure.
1223 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1225 * Check whether the probe is being run is the context of a given
1226 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1227 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1229 * The return value depends on the result of the test, and can be:
1231 * * 0, if the *skb* task belongs to the cgroup2.
1232 * * 1, if the *skb* task does not belong to the cgroup2.
1233 * * A negative error code, if an error occurred.
1235 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1237 * Resize (trim or grow) the packet associated to *skb* to the
1238 * new *len*. The *flags* are reserved for future usage, and must
1241 * The basic idea is that the helper performs the needed work to
1242 * change the size of the packet, then the eBPF program rewrites
1243 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1244 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1245 * and others. This helper is a slow path utility intended for
1246 * replies with control messages. And because it is targeted for
1247 * slow path, the helper itself can afford to be slow: it
1248 * implicitly linearizes, unclones and drops offloads from the
1251 * A call to this helper is susceptible to change the underlaying
1252 * packet buffer. Therefore, at load time, all checks on pointers
1253 * previously done by the verifier are invalidated and must be
1254 * performed again, if the helper is used in combination with
1255 * direct packet access.
1257 * 0 on success, or a negative error in case of failure.
1259 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1261 * Pull in non-linear data in case the *skb* is non-linear and not
1262 * all of *len* are part of the linear section. Make *len* bytes
1263 * from *skb* readable and writable. If a zero value is passed for
1264 * *len*, then the whole length of the *skb* is pulled.
1266 * This helper is only needed for reading and writing with direct
1269 * For direct packet access, testing that offsets to access
1270 * are within packet boundaries (test on *skb*\ **->data_end**) is
1271 * susceptible to fail if offsets are invalid, or if the requested
1272 * data is in non-linear parts of the *skb*. On failure the
1273 * program can just bail out, or in the case of a non-linear
1274 * buffer, use a helper to make the data available. The
1275 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1276 * the data. Another one consists in using **bpf_skb_pull_data**
1277 * to pull in once the non-linear parts, then retesting and
1278 * eventually access the data.
1280 * At the same time, this also makes sure the *skb* is uncloned,
1281 * which is a necessary condition for direct write. As this needs
1282 * to be an invariant for the write part only, the verifier
1283 * detects writes and adds a prologue that is calling
1284 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1285 * the very beginning in case it is indeed cloned.
1287 * A call to this helper is susceptible to change the underlaying
1288 * packet buffer. Therefore, at load time, all checks on pointers
1289 * previously done by the verifier are invalidated and must be
1290 * performed again, if the helper is used in combination with
1291 * direct packet access.
1293 * 0 on success, or a negative error in case of failure.
1295 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1297 * Add the checksum *csum* into *skb*\ **->csum** in case the
1298 * driver has supplied a checksum for the entire packet into that
1299 * field. Return an error otherwise. This helper is intended to be
1300 * used in combination with **bpf_csum_diff**\ (), in particular
1301 * when the checksum needs to be updated after data has been
1302 * written into the packet through direct packet access.
1304 * The checksum on success, or a negative error code in case of
1307 * void bpf_set_hash_invalid(struct sk_buff *skb)
1309 * Invalidate the current *skb*\ **->hash**. It can be used after
1310 * mangling on headers through direct packet access, in order to
1311 * indicate that the hash is outdated and to trigger a
1312 * recalculation the next time the kernel tries to access this
1313 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1315 * int bpf_get_numa_node_id(void)
1317 * Return the id of the current NUMA node. The primary use case
1318 * for this helper is the selection of sockets for the local NUMA
1319 * node, when the program is attached to sockets using the
1320 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1321 * but the helper is also available to other eBPF program types,
1322 * similarly to **bpf_get_smp_processor_id**\ ().
1324 * The id of current NUMA node.
1326 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1328 * Grows headroom of packet associated to *skb* and adjusts the
1329 * offset of the MAC header accordingly, adding *len* bytes of
1330 * space. It automatically extends and reallocates memory as
1333 * This helper can be used on a layer 3 *skb* to push a MAC header
1334 * for redirection into a layer 2 device.
1336 * All values for *flags* are reserved for future usage, and must
1339 * A call to this helper is susceptible to change the underlaying
1340 * packet buffer. Therefore, at load time, all checks on pointers
1341 * previously done by the verifier are invalidated and must be
1342 * performed again, if the helper is used in combination with
1343 * direct packet access.
1345 * 0 on success, or a negative error in case of failure.
1347 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1349 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1350 * it is possible to use a negative value for *delta*. This helper
1351 * can be used to prepare the packet for pushing or popping
1354 * A call to this helper is susceptible to change the underlaying
1355 * packet buffer. Therefore, at load time, all checks on pointers
1356 * previously done by the verifier are invalidated and must be
1357 * performed again, if the helper is used in combination with
1358 * direct packet access.
1360 * 0 on success, or a negative error in case of failure.
1362 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1364 * Copy a NUL terminated string from an unsafe address
1365 * *unsafe_ptr* to *dst*. The *size* should include the
1366 * terminating NUL byte. In case the string length is smaller than
1367 * *size*, the target is not padded with further NUL bytes. If the
1368 * string length is larger than *size*, just *size*-1 bytes are
1369 * copied and the last byte is set to NUL.
1371 * On success, the length of the copied string is returned. This
1372 * makes this helper useful in tracing programs for reading
1373 * strings, and more importantly to get its length at runtime. See
1374 * the following snippet:
1378 * SEC("kprobe/sys_open")
1379 * void bpf_sys_open(struct pt_regs *ctx)
1381 * char buf[PATHLEN]; // PATHLEN is defined to 256
1382 * int res = bpf_probe_read_str(buf, sizeof(buf),
1385 * // Consume buf, for example push it to
1386 * // userspace via bpf_perf_event_output(); we
1387 * // can use res (the string length) as event
1388 * // size, after checking its boundaries.
1391 * In comparison, using **bpf_probe_read()** helper here instead
1392 * to read the string would require to estimate the length at
1393 * compile time, and would often result in copying more memory
1396 * Another useful use case is when parsing individual process
1397 * arguments or individual environment variables navigating
1398 * *current*\ **->mm->arg_start** and *current*\
1399 * **->mm->env_start**: using this helper and the return value,
1400 * one can quickly iterate at the right offset of the memory area.
1402 * On success, the strictly positive length of the string,
1403 * including the trailing NUL character. On error, a negative
1406 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1408 * If the **struct sk_buff** pointed by *skb* has a known socket,
1409 * retrieve the cookie (generated by the kernel) of this socket.
1410 * If no cookie has been set yet, generate a new cookie. Once
1411 * generated, the socket cookie remains stable for the life of the
1412 * socket. This helper can be useful for monitoring per socket
1413 * networking traffic statistics as it provides a unique socket
1414 * identifier per namespace.
1416 * A 8-byte long non-decreasing number on success, or 0 if the
1417 * socket field is missing inside *skb*.
1419 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1421 * Equivalent to bpf_get_socket_cookie() helper that accepts
1422 * *skb*, but gets socket from **struct bpf_sock_addr** contex.
1424 * A 8-byte long non-decreasing number.
1426 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1428 * Equivalent to bpf_get_socket_cookie() helper that accepts
1429 * *skb*, but gets socket from **struct bpf_sock_ops** contex.
1431 * A 8-byte long non-decreasing number.
1433 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1435 * The owner UID of the socket associated to *skb*. If the socket
1436 * is **NULL**, or if it is not a full socket (i.e. if it is a
1437 * time-wait or a request socket instead), **overflowuid** value
1438 * is returned (note that **overflowuid** might also be the actual
1439 * UID value for the socket).
1441 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1443 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1448 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1450 * Emulate a call to **setsockopt()** on the socket associated to
1451 * *bpf_socket*, which must be a full socket. The *level* at
1452 * which the option resides and the name *optname* of the option
1453 * must be specified, see **setsockopt(2)** for more information.
1454 * The option value of length *optlen* is pointed by *optval*.
1456 * This helper actually implements a subset of **setsockopt()**.
1457 * It supports the following *level*\ s:
1459 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1460 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1461 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1462 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1463 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1464 * **TCP_BPF_SNDCWND_CLAMP**.
1465 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1466 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1468 * 0 on success, or a negative error in case of failure.
1470 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1472 * Grow or shrink the room for data in the packet associated to
1473 * *skb* by *len_diff*, and according to the selected *mode*.
1475 * There is a single supported mode at this time:
1477 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1478 * (room space is added or removed below the layer 3 header).
1480 * All values for *flags* are reserved for future usage, and must
1483 * A call to this helper is susceptible to change the underlaying
1484 * packet buffer. Therefore, at load time, all checks on pointers
1485 * previously done by the verifier are invalidated and must be
1486 * performed again, if the helper is used in combination with
1487 * direct packet access.
1489 * 0 on success, or a negative error in case of failure.
1491 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1493 * Redirect the packet to the endpoint referenced by *map* at
1494 * index *key*. Depending on its type, this *map* can contain
1495 * references to net devices (for forwarding packets through other
1496 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1497 * but this is only implemented for native XDP (with driver
1498 * support) as of this writing).
1500 * All values for *flags* are reserved for future usage, and must
1503 * When used to redirect packets to net devices, this helper
1504 * provides a high performance increase over **bpf_redirect**\ ().
1505 * This is due to various implementation details of the underlying
1506 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1507 * () tries to send packet as a "bulk" to the device.
1509 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1511 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1513 * Redirect the packet to the socket referenced by *map* (of type
1514 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1515 * egress interfaces can be used for redirection. The
1516 * **BPF_F_INGRESS** value in *flags* is used to make the
1517 * distinction (ingress path is selected if the flag is present,
1518 * egress path otherwise). This is the only flag supported for now.
1520 * **SK_PASS** on success, or **SK_DROP** on error.
1522 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1524 * Add an entry to, or update a *map* referencing sockets. The
1525 * *skops* is used as a new value for the entry associated to
1526 * *key*. *flags* is one of:
1529 * The entry for *key* must not exist in the map.
1531 * The entry for *key* must already exist in the map.
1533 * No condition on the existence of the entry for *key*.
1535 * If the *map* has eBPF programs (parser and verdict), those will
1536 * be inherited by the socket being added. If the socket is
1537 * already attached to eBPF programs, this results in an error.
1539 * 0 on success, or a negative error in case of failure.
1541 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1543 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1544 * *delta* (which can be positive or negative). Note that this
1545 * operation modifies the address stored in *xdp_md*\ **->data**,
1546 * so the latter must be loaded only after the helper has been
1549 * The use of *xdp_md*\ **->data_meta** is optional and programs
1550 * are not required to use it. The rationale is that when the
1551 * packet is processed with XDP (e.g. as DoS filter), it is
1552 * possible to push further meta data along with it before passing
1553 * to the stack, and to give the guarantee that an ingress eBPF
1554 * program attached as a TC classifier on the same device can pick
1555 * this up for further post-processing. Since TC works with socket
1556 * buffers, it remains possible to set from XDP the **mark** or
1557 * **priority** pointers, or other pointers for the socket buffer.
1558 * Having this scratch space generic and programmable allows for
1559 * more flexibility as the user is free to store whatever meta
1562 * A call to this helper is susceptible to change the underlaying
1563 * packet buffer. Therefore, at load time, all checks on pointers
1564 * previously done by the verifier are invalidated and must be
1565 * performed again, if the helper is used in combination with
1566 * direct packet access.
1568 * 0 on success, or a negative error in case of failure.
1570 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1572 * Read the value of a perf event counter, and store it into *buf*
1573 * of size *buf_size*. This helper relies on a *map* of type
1574 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1575 * counter is selected when *map* is updated with perf event file
1576 * descriptors. The *map* is an array whose size is the number of
1577 * available CPUs, and each cell contains a value relative to one
1578 * CPU. The value to retrieve is indicated by *flags*, that
1579 * contains the index of the CPU to look up, masked with
1580 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1581 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1582 * current CPU should be retrieved.
1584 * This helper behaves in a way close to
1585 * **bpf_perf_event_read**\ () helper, save that instead of
1586 * just returning the value observed, it fills the *buf*
1587 * structure. This allows for additional data to be retrieved: in
1588 * particular, the enabled and running times (in *buf*\
1589 * **->enabled** and *buf*\ **->running**, respectively) are
1590 * copied. In general, **bpf_perf_event_read_value**\ () is
1591 * recommended over **bpf_perf_event_read**\ (), which has some
1592 * ABI issues and provides fewer functionalities.
1594 * These values are interesting, because hardware PMU (Performance
1595 * Monitoring Unit) counters are limited resources. When there are
1596 * more PMU based perf events opened than available counters,
1597 * kernel will multiplex these events so each event gets certain
1598 * percentage (but not all) of the PMU time. In case that
1599 * multiplexing happens, the number of samples or counter value
1600 * will not reflect the case compared to when no multiplexing
1601 * occurs. This makes comparison between different runs difficult.
1602 * Typically, the counter value should be normalized before
1603 * comparing to other experiments. The usual normalization is done
1608 * normalized_counter = counter * t_enabled / t_running
1610 * Where t_enabled is the time enabled for event and t_running is
1611 * the time running for event since last normalization. The
1612 * enabled and running times are accumulated since the perf event
1613 * open. To achieve scaling factor between two invocations of an
1614 * eBPF program, users can can use CPU id as the key (which is
1615 * typical for perf array usage model) to remember the previous
1616 * value and do the calculation inside the eBPF program.
1618 * 0 on success, or a negative error in case of failure.
1620 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1622 * For en eBPF program attached to a perf event, retrieve the
1623 * value of the event counter associated to *ctx* and store it in
1624 * the structure pointed by *buf* and of size *buf_size*. Enabled
1625 * and running times are also stored in the structure (see
1626 * description of helper **bpf_perf_event_read_value**\ () for
1629 * 0 on success, or a negative error in case of failure.
1631 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1633 * Emulate a call to **getsockopt()** on the socket associated to
1634 * *bpf_socket*, which must be a full socket. The *level* at
1635 * which the option resides and the name *optname* of the option
1636 * must be specified, see **getsockopt(2)** for more information.
1637 * The retrieved value is stored in the structure pointed by
1638 * *opval* and of length *optlen*.
1640 * This helper actually implements a subset of **getsockopt()**.
1641 * It supports the following *level*\ s:
1643 * * **IPPROTO_TCP**, which supports *optname*
1644 * **TCP_CONGESTION**.
1645 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1646 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1648 * 0 on success, or a negative error in case of failure.
1650 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1652 * Used for error injection, this helper uses kprobes to override
1653 * the return value of the probed function, and to set it to *rc*.
1654 * The first argument is the context *regs* on which the kprobe
1657 * This helper works by setting setting the PC (program counter)
1658 * to an override function which is run in place of the original
1659 * probed function. This means the probed function is not run at
1660 * all. The replacement function just returns with the required
1663 * This helper has security implications, and thus is subject to
1664 * restrictions. It is only available if the kernel was compiled
1665 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1666 * option, and in this case it only works on functions tagged with
1667 * **ALLOW_ERROR_INJECTION** in the kernel code.
1669 * Also, the helper is only available for the architectures having
1670 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1671 * x86 architecture is the only one to support this feature.
1675 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1677 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1678 * for the full TCP socket associated to *bpf_sock_ops* to
1681 * The primary use of this field is to determine if there should
1682 * be calls to eBPF programs of type
1683 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1684 * code. A program of the same type can change its value, per
1685 * connection and as necessary, when the connection is
1686 * established. This field is directly accessible for reading, but
1687 * this helper must be used for updates in order to return an
1688 * error if an eBPF program tries to set a callback that is not
1689 * supported in the current kernel.
1691 * The supported callback values that *argval* can combine are:
1693 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1694 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1695 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1697 * Here are some examples of where one could call such eBPF
1701 * * When a packet is retransmitted.
1702 * * When the connection terminates.
1703 * * When a packet is sent.
1704 * * When a packet is received.
1706 * Code **-EINVAL** if the socket is not a full TCP socket;
1707 * otherwise, a positive number containing the bits that could not
1708 * be set is returned (which comes down to 0 if all bits were set
1711 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1713 * This helper is used in programs implementing policies at the
1714 * socket level. If the message *msg* is allowed to pass (i.e. if
1715 * the verdict eBPF program returns **SK_PASS**), redirect it to
1716 * the socket referenced by *map* (of type
1717 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1718 * egress interfaces can be used for redirection. The
1719 * **BPF_F_INGRESS** value in *flags* is used to make the
1720 * distinction (ingress path is selected if the flag is present,
1721 * egress path otherwise). This is the only flag supported for now.
1723 * **SK_PASS** on success, or **SK_DROP** on error.
1725 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1727 * For socket policies, apply the verdict of the eBPF program to
1728 * the next *bytes* (number of bytes) of message *msg*.
1730 * For example, this helper can be used in the following cases:
1732 * * A single **sendmsg**\ () or **sendfile**\ () system call
1733 * contains multiple logical messages that the eBPF program is
1734 * supposed to read and for which it should apply a verdict.
1735 * * An eBPF program only cares to read the first *bytes* of a
1736 * *msg*. If the message has a large payload, then setting up
1737 * and calling the eBPF program repeatedly for all bytes, even
1738 * though the verdict is already known, would create unnecessary
1741 * When called from within an eBPF program, the helper sets a
1742 * counter internal to the BPF infrastructure, that is used to
1743 * apply the last verdict to the next *bytes*. If *bytes* is
1744 * smaller than the current data being processed from a
1745 * **sendmsg**\ () or **sendfile**\ () system call, the first
1746 * *bytes* will be sent and the eBPF program will be re-run with
1747 * the pointer for start of data pointing to byte number *bytes*
1748 * **+ 1**. If *bytes* is larger than the current data being
1749 * processed, then the eBPF verdict will be applied to multiple
1750 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1753 * Note that if a socket closes with the internal counter holding
1754 * a non-zero value, this is not a problem because data is not
1755 * being buffered for *bytes* and is sent as it is received.
1759 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1761 * For socket policies, prevent the execution of the verdict eBPF
1762 * program for message *msg* until *bytes* (byte number) have been
1765 * This can be used when one needs a specific number of bytes
1766 * before a verdict can be assigned, even if the data spans
1767 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1768 * case would be a user calling **sendmsg**\ () repeatedly with
1769 * 1-byte long message segments. Obviously, this is bad for
1770 * performance, but it is still valid. If the eBPF program needs
1771 * *bytes* bytes to validate a header, this helper can be used to
1772 * prevent the eBPF program to be called again until *bytes* have
1777 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1779 * For socket policies, pull in non-linear data from user space
1780 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1781 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1784 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1785 * *msg* it can only parse data that the (**data**, **data_end**)
1786 * pointers have already consumed. For **sendmsg**\ () hooks this
1787 * is likely the first scatterlist element. But for calls relying
1788 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1789 * be the range (**0**, **0**) because the data is shared with
1790 * user space and by default the objective is to avoid allowing
1791 * user space to modify data while (or after) eBPF verdict is
1792 * being decided. This helper can be used to pull in data and to
1793 * set the start and end pointer to given values. Data will be
1794 * copied if necessary (i.e. if data was not linear and if start
1795 * and end pointers do not point to the same chunk).
1797 * A call to this helper is susceptible to change the underlaying
1798 * packet buffer. Therefore, at load time, all checks on pointers
1799 * previously done by the verifier are invalidated and must be
1800 * performed again, if the helper is used in combination with
1801 * direct packet access.
1803 * All values for *flags* are reserved for future usage, and must
1806 * 0 on success, or a negative error in case of failure.
1808 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1810 * Bind the socket associated to *ctx* to the address pointed by
1811 * *addr*, of length *addr_len*. This allows for making outgoing
1812 * connection from the desired IP address, which can be useful for
1813 * example when all processes inside a cgroup should use one
1814 * single IP address on a host that has multiple IP configured.
1816 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1817 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1818 * **AF_INET6**). Looking for a free port to bind to can be
1819 * expensive, therefore binding to port is not permitted by the
1820 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1821 * must be set to zero.
1823 * 0 on success, or a negative error in case of failure.
1825 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1827 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1828 * only possible to shrink the packet as of this writing,
1829 * therefore *delta* must be a negative integer.
1831 * A call to this helper is susceptible to change the underlaying
1832 * packet buffer. Therefore, at load time, all checks on pointers
1833 * previously done by the verifier are invalidated and must be
1834 * performed again, if the helper is used in combination with
1835 * direct packet access.
1837 * 0 on success, or a negative error in case of failure.
1839 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1841 * Retrieve the XFRM state (IP transform framework, see also
1842 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1844 * The retrieved value is stored in the **struct bpf_xfrm_state**
1845 * pointed by *xfrm_state* and of length *size*.
1847 * All values for *flags* are reserved for future usage, and must
1850 * This helper is available only if the kernel was compiled with
1851 * **CONFIG_XFRM** configuration option.
1853 * 0 on success, or a negative error in case of failure.
1855 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1857 * Return a user or a kernel stack in bpf program provided buffer.
1858 * To achieve this, the helper needs *ctx*, which is a pointer
1859 * to the context on which the tracing program is executed.
1860 * To store the stacktrace, the bpf program provides *buf* with
1861 * a nonnegative *size*.
1863 * The last argument, *flags*, holds the number of stack frames to
1864 * skip (from 0 to 255), masked with
1865 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1866 * the following flags:
1868 * **BPF_F_USER_STACK**
1869 * Collect a user space stack instead of a kernel stack.
1870 * **BPF_F_USER_BUILD_ID**
1871 * Collect buildid+offset instead of ips for user stack,
1872 * only valid if **BPF_F_USER_STACK** is also specified.
1874 * **bpf_get_stack**\ () can collect up to
1875 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1876 * to sufficient large buffer size. Note that
1877 * this limit can be controlled with the **sysctl** program, and
1878 * that it should be manually increased in order to profile long
1879 * user stacks (such as stacks for Java programs). To do so, use:
1883 * # sysctl kernel.perf_event_max_stack=<new value>
1885 * A non-negative value equal to or less than *size* on success,
1886 * or a negative error in case of failure.
1888 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1890 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1891 * it provides an easy way to load *len* bytes from *offset*
1892 * from the packet associated to *skb*, into the buffer pointed
1893 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1894 * a fifth argument *start_header* exists in order to select a
1895 * base offset to start from. *start_header* can be one of:
1897 * **BPF_HDR_START_MAC**
1898 * Base offset to load data from is *skb*'s mac header.
1899 * **BPF_HDR_START_NET**
1900 * Base offset to load data from is *skb*'s network header.
1902 * In general, "direct packet access" is the preferred method to
1903 * access packet data, however, this helper is in particular useful
1904 * in socket filters where *skb*\ **->data** does not always point
1905 * to the start of the mac header and where "direct packet access"
1908 * 0 on success, or a negative error in case of failure.
1910 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1912 * Do FIB lookup in kernel tables using parameters in *params*.
1913 * If lookup is successful and result shows packet is to be
1914 * forwarded, the neighbor tables are searched for the nexthop.
1915 * If successful (ie., FIB lookup shows forwarding and nexthop
1916 * is resolved), the nexthop address is returned in ipv4_dst
1917 * or ipv6_dst based on family, smac is set to mac address of
1918 * egress device, dmac is set to nexthop mac address, rt_metric
1919 * is set to metric from route (IPv4/IPv6 only), and ifindex
1920 * is set to the device index of the nexthop from the FIB lookup.
1922 * *plen* argument is the size of the passed in struct.
1923 * *flags* argument can be a combination of one or more of the
1926 * **BPF_FIB_LOOKUP_DIRECT**
1927 * Do a direct table lookup vs full lookup using FIB
1929 * **BPF_FIB_LOOKUP_OUTPUT**
1930 * Perform lookup from an egress perspective (default is
1933 * *ctx* is either **struct xdp_md** for XDP programs or
1934 * **struct sk_buff** tc cls_act programs.
1936 * * < 0 if any input argument is invalid
1937 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1938 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1939 * packet is not forwarded or needs assist from full stack
1941 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1943 * Add an entry to, or update a sockhash *map* referencing sockets.
1944 * The *skops* is used as a new value for the entry associated to
1945 * *key*. *flags* is one of:
1948 * The entry for *key* must not exist in the map.
1950 * The entry for *key* must already exist in the map.
1952 * No condition on the existence of the entry for *key*.
1954 * If the *map* has eBPF programs (parser and verdict), those will
1955 * be inherited by the socket being added. If the socket is
1956 * already attached to eBPF programs, this results in an error.
1958 * 0 on success, or a negative error in case of failure.
1960 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1962 * This helper is used in programs implementing policies at the
1963 * socket level. If the message *msg* is allowed to pass (i.e. if
1964 * the verdict eBPF program returns **SK_PASS**), redirect it to
1965 * the socket referenced by *map* (of type
1966 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1967 * egress interfaces can be used for redirection. The
1968 * **BPF_F_INGRESS** value in *flags* is used to make the
1969 * distinction (ingress path is selected if the flag is present,
1970 * egress path otherwise). This is the only flag supported for now.
1972 * **SK_PASS** on success, or **SK_DROP** on error.
1974 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1976 * This helper is used in programs implementing policies at the
1977 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1978 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1979 * to the socket referenced by *map* (of type
1980 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1981 * egress interfaces can be used for redirection. The
1982 * **BPF_F_INGRESS** value in *flags* is used to make the
1983 * distinction (ingress path is selected if the flag is present,
1984 * egress otherwise). This is the only flag supported for now.
1986 * **SK_PASS** on success, or **SK_DROP** on error.
1988 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
1990 * Encapsulate the packet associated to *skb* within a Layer 3
1991 * protocol header. This header is provided in the buffer at
1992 * address *hdr*, with *len* its size in bytes. *type* indicates
1993 * the protocol of the header and can be one of:
1995 * **BPF_LWT_ENCAP_SEG6**
1996 * IPv6 encapsulation with Segment Routing Header
1997 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
1998 * the IPv6 header is computed by the kernel.
1999 * **BPF_LWT_ENCAP_SEG6_INLINE**
2000 * Only works if *skb* contains an IPv6 packet. Insert a
2001 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2004 * A call to this helper is susceptible to change the underlaying
2005 * packet buffer. Therefore, at load time, all checks on pointers
2006 * previously done by the verifier are invalidated and must be
2007 * performed again, if the helper is used in combination with
2008 * direct packet access.
2010 * 0 on success, or a negative error in case of failure.
2012 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2014 * Store *len* bytes from address *from* into the packet
2015 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2016 * inside the outermost IPv6 Segment Routing Header can be
2017 * modified through this helper.
2019 * A call to this helper is susceptible to change the underlaying
2020 * packet buffer. Therefore, at load time, all checks on pointers
2021 * previously done by the verifier are invalidated and must be
2022 * performed again, if the helper is used in combination with
2023 * direct packet access.
2025 * 0 on success, or a negative error in case of failure.
2027 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2029 * Adjust the size allocated to TLVs in the outermost IPv6
2030 * Segment Routing Header contained in the packet associated to
2031 * *skb*, at position *offset* by *delta* bytes. Only offsets
2032 * after the segments are accepted. *delta* can be as well
2033 * positive (growing) as negative (shrinking).
2035 * A call to this helper is susceptible to change the underlaying
2036 * packet buffer. Therefore, at load time, all checks on pointers
2037 * previously done by the verifier are invalidated and must be
2038 * performed again, if the helper is used in combination with
2039 * direct packet access.
2041 * 0 on success, or a negative error in case of failure.
2043 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2045 * Apply an IPv6 Segment Routing action of type *action* to the
2046 * packet associated to *skb*. Each action takes a parameter
2047 * contained at address *param*, and of length *param_len* bytes.
2048 * *action* can be one of:
2050 * **SEG6_LOCAL_ACTION_END_X**
2051 * End.X action: Endpoint with Layer-3 cross-connect.
2052 * Type of *param*: **struct in6_addr**.
2053 * **SEG6_LOCAL_ACTION_END_T**
2054 * End.T action: Endpoint with specific IPv6 table lookup.
2055 * Type of *param*: **int**.
2056 * **SEG6_LOCAL_ACTION_END_B6**
2057 * End.B6 action: Endpoint bound to an SRv6 policy.
2058 * Type of param: **struct ipv6_sr_hdr**.
2059 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2060 * End.B6.Encap action: Endpoint bound to an SRv6
2061 * encapsulation policy.
2062 * Type of param: **struct ipv6_sr_hdr**.
2064 * A call to this helper is susceptible to change the underlaying
2065 * packet buffer. Therefore, at load time, all checks on pointers
2066 * previously done by the verifier are invalidated and must be
2067 * performed again, if the helper is used in combination with
2068 * direct packet access.
2070 * 0 on success, or a negative error in case of failure.
2072 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2074 * This helper is used in programs implementing IR decoding, to
2075 * report a successfully decoded key press with *scancode*,
2076 * *toggle* value in the given *protocol*. The scancode will be
2077 * translated to a keycode using the rc keymap, and reported as
2078 * an input key down event. After a period a key up event is
2079 * generated. This period can be extended by calling either
2080 * **bpf_rc_keydown**\ () again with the same values, or calling
2081 * **bpf_rc_repeat**\ ().
2083 * Some protocols include a toggle bit, in case the button was
2084 * released and pressed again between consecutive scancodes.
2086 * The *ctx* should point to the lirc sample as passed into
2089 * The *protocol* is the decoded protocol number (see
2090 * **enum rc_proto** for some predefined values).
2092 * This helper is only available is the kernel was compiled with
2093 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2098 * int bpf_rc_repeat(void *ctx)
2100 * This helper is used in programs implementing IR decoding, to
2101 * report a successfully decoded repeat key message. This delays
2102 * the generation of a key up event for previously generated
2105 * Some IR protocols like NEC have a special IR message for
2106 * repeating last button, for when a button is held down.
2108 * The *ctx* should point to the lirc sample as passed into
2111 * This helper is only available is the kernel was compiled with
2112 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2117 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2119 * Return the cgroup v2 id of the socket associated with the *skb*.
2120 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2121 * helper for cgroup v1 by providing a tag resp. identifier that
2122 * can be matched on or used for map lookups e.g. to implement
2123 * policy. The cgroup v2 id of a given path in the hierarchy is
2124 * exposed in user space through the f_handle API in order to get
2125 * to the same 64-bit id.
2127 * This helper can be used on TC egress path, but not on ingress,
2128 * and is available only if the kernel was compiled with the
2129 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2131 * The id is returned or 0 in case the id could not be retrieved.
2133 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2135 * Return id of cgroup v2 that is ancestor of cgroup associated
2136 * with the *skb* at the *ancestor_level*. The root cgroup is at
2137 * *ancestor_level* zero and each step down the hierarchy
2138 * increments the level. If *ancestor_level* == level of cgroup
2139 * associated with *skb*, then return value will be same as that
2140 * of **bpf_skb_cgroup_id**\ ().
2142 * The helper is useful to implement policies based on cgroups
2143 * that are upper in hierarchy than immediate cgroup associated
2146 * The format of returned id and helper limitations are same as in
2147 * **bpf_skb_cgroup_id**\ ().
2149 * The id is returned or 0 in case the id could not be retrieved.
2151 * u64 bpf_get_current_cgroup_id(void)
2153 * A 64-bit integer containing the current cgroup id based
2154 * on the cgroup within which the current task is running.
2156 * void* get_local_storage(void *map, u64 flags)
2158 * Get the pointer to the local storage area.
2159 * The type and the size of the local storage is defined
2160 * by the *map* argument.
2161 * The *flags* meaning is specific for each map type,
2162 * and has to be 0 for cgroup local storage.
2164 * Depending on the BPF program type, a local storage area
2165 * can be shared between multiple instances of the BPF program,
2166 * running simultaneously.
2168 * A user should care about the synchronization by himself.
2169 * For example, by using the **BPF_STX_XADD** instruction to alter
2172 * A pointer to the local storage area.
2174 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2176 * Select a **SO_REUSEPORT** socket from a
2177 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2178 * It checks the selected socket is matching the incoming
2179 * request in the socket buffer.
2181 * 0 on success, or a negative error in case of failure.
2183 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2185 * Look for TCP socket matching *tuple*, optionally in a child
2186 * network namespace *netns*. The return value must be checked,
2187 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2189 * The *ctx* should point to the context of the program, such as
2190 * the skb or socket (depending on the hook in use). This is used
2191 * to determine the base network namespace for the lookup.
2193 * *tuple_size* must be one of:
2195 * **sizeof**\ (*tuple*\ **->ipv4**)
2196 * Look for an IPv4 socket.
2197 * **sizeof**\ (*tuple*\ **->ipv6**)
2198 * Look for an IPv6 socket.
2200 * If the *netns* is zero, then the socket lookup table in the
2201 * netns associated with the *ctx* will be used. For the TC hooks,
2202 * this in the netns of the device in the skb. For socket hooks,
2203 * this in the netns of the socket. If *netns* is non-zero, then
2204 * it specifies the ID of the netns relative to the netns
2205 * associated with the *ctx*.
2207 * All values for *flags* are reserved for future usage, and must
2210 * This helper is available only if the kernel was compiled with
2211 * **CONFIG_NET** configuration option.
2213 * A pointer to *struct bpf_sock*, or **NULL** in case of failure.
2214 * For sockets with reuseport option, **struct bpf_sock**
2215 * return is from **reuse->socks**\ [] using hash of the packet.
2217 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2219 * Look for UDP socket matching *tuple*, optionally in a child
2220 * network namespace *netns*. The return value must be checked,
2221 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2223 * The *ctx* should point to the context of the program, such as
2224 * the skb or socket (depending on the hook in use). This is used
2225 * to determine the base network namespace for the lookup.
2227 * *tuple_size* must be one of:
2229 * **sizeof**\ (*tuple*\ **->ipv4**)
2230 * Look for an IPv4 socket.
2231 * **sizeof**\ (*tuple*\ **->ipv6**)
2232 * Look for an IPv6 socket.
2234 * If the *netns* is zero, then the socket lookup table in the
2235 * netns associated with the *ctx* will be used. For the TC hooks,
2236 * this in the netns of the device in the skb. For socket hooks,
2237 * this in the netns of the socket. If *netns* is non-zero, then
2238 * it specifies the ID of the netns relative to the netns
2239 * associated with the *ctx*.
2241 * All values for *flags* are reserved for future usage, and must
2244 * This helper is available only if the kernel was compiled with
2245 * **CONFIG_NET** configuration option.
2247 * A pointer to **struct bpf_sock**, or **NULL** in case of
2248 * failure. For sockets with reuseport option, **struct bpf_sock**
2249 * return is from **reuse->socks**\ [] using hash of the packet.
2251 * int bpf_sk_release(struct bpf_sock *sock)
2253 * Release the reference held by *sock*. *sock* must be a
2254 * non-**NULL** pointer that was returned from
2255 * **bpf_sk_lookup_xxx**\ ().
2257 * 0 on success, or a negative error in case of failure.
2259 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2261 * Pop an element from *map*.
2263 * 0 on success, or a negative error in case of failure.
2265 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2267 * Get an element from *map* without removing it.
2269 * 0 on success, or a negative error in case of failure.
2271 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2273 * For socket policies, insert *len* bytes into *msg* at offset
2276 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2277 * *msg* it may want to insert metadata or options into the *msg*.
2278 * This can later be read and used by any of the lower layer BPF
2281 * This helper may fail if under memory pressure (a malloc
2282 * fails) in these cases BPF programs will get an appropriate
2283 * error and BPF programs will need to handle them.
2285 * 0 on success, or a negative error in case of failure.
2287 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2289 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2290 * This may result in **ENOMEM** errors under certain situations if
2291 * an allocation and copy are required due to a full ring buffer.
2292 * However, the helper will try to avoid doing the allocation
2293 * if possible. Other errors can occur if input parameters are
2294 * invalid either due to *start* byte not being valid part of *msg*
2295 * payload and/or *pop* value being to large.
2297 * 0 on success, or a negative error in case of failure.
2299 #define __BPF_FUNC_MAPPER(FN) \
2301 FN(map_lookup_elem), \
2302 FN(map_update_elem), \
2303 FN(map_delete_elem), \
2307 FN(get_prandom_u32), \
2308 FN(get_smp_processor_id), \
2309 FN(skb_store_bytes), \
2310 FN(l3_csum_replace), \
2311 FN(l4_csum_replace), \
2313 FN(clone_redirect), \
2314 FN(get_current_pid_tgid), \
2315 FN(get_current_uid_gid), \
2316 FN(get_current_comm), \
2317 FN(get_cgroup_classid), \
2318 FN(skb_vlan_push), \
2320 FN(skb_get_tunnel_key), \
2321 FN(skb_set_tunnel_key), \
2322 FN(perf_event_read), \
2324 FN(get_route_realm), \
2325 FN(perf_event_output), \
2326 FN(skb_load_bytes), \
2329 FN(skb_get_tunnel_opt), \
2330 FN(skb_set_tunnel_opt), \
2331 FN(skb_change_proto), \
2332 FN(skb_change_type), \
2333 FN(skb_under_cgroup), \
2334 FN(get_hash_recalc), \
2335 FN(get_current_task), \
2336 FN(probe_write_user), \
2337 FN(current_task_under_cgroup), \
2338 FN(skb_change_tail), \
2339 FN(skb_pull_data), \
2341 FN(set_hash_invalid), \
2342 FN(get_numa_node_id), \
2343 FN(skb_change_head), \
2344 FN(xdp_adjust_head), \
2345 FN(probe_read_str), \
2346 FN(get_socket_cookie), \
2347 FN(get_socket_uid), \
2350 FN(skb_adjust_room), \
2352 FN(sk_redirect_map), \
2353 FN(sock_map_update), \
2354 FN(xdp_adjust_meta), \
2355 FN(perf_event_read_value), \
2356 FN(perf_prog_read_value), \
2358 FN(override_return), \
2359 FN(sock_ops_cb_flags_set), \
2360 FN(msg_redirect_map), \
2361 FN(msg_apply_bytes), \
2362 FN(msg_cork_bytes), \
2363 FN(msg_pull_data), \
2365 FN(xdp_adjust_tail), \
2366 FN(skb_get_xfrm_state), \
2368 FN(skb_load_bytes_relative), \
2370 FN(sock_hash_update), \
2371 FN(msg_redirect_hash), \
2372 FN(sk_redirect_hash), \
2373 FN(lwt_push_encap), \
2374 FN(lwt_seg6_store_bytes), \
2375 FN(lwt_seg6_adjust_srh), \
2376 FN(lwt_seg6_action), \
2379 FN(skb_cgroup_id), \
2380 FN(get_current_cgroup_id), \
2381 FN(get_local_storage), \
2382 FN(sk_select_reuseport), \
2383 FN(skb_ancestor_cgroup_id), \
2384 FN(sk_lookup_tcp), \
2385 FN(sk_lookup_udp), \
2387 FN(map_push_elem), \
2389 FN(map_peek_elem), \
2390 FN(msg_push_data), \
2393 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2394 * function eBPF program intends to call
2396 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2398 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2401 #undef __BPF_ENUM_FN
2403 /* All flags used by eBPF helper functions, placed here. */
2405 /* BPF_FUNC_skb_store_bytes flags. */
2406 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2407 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2409 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2410 * First 4 bits are for passing the header field size.
2412 #define BPF_F_HDR_FIELD_MASK 0xfULL
2414 /* BPF_FUNC_l4_csum_replace flags. */
2415 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2416 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2417 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2419 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2420 #define BPF_F_INGRESS (1ULL << 0)
2422 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2423 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2425 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2426 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2427 #define BPF_F_USER_STACK (1ULL << 8)
2428 /* flags used by BPF_FUNC_get_stackid only. */
2429 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2430 #define BPF_F_REUSE_STACKID (1ULL << 10)
2431 /* flags used by BPF_FUNC_get_stack only. */
2432 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2434 /* BPF_FUNC_skb_set_tunnel_key flags. */
2435 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2436 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2437 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2439 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2440 * BPF_FUNC_perf_event_read_value flags.
2442 #define BPF_F_INDEX_MASK 0xffffffffULL
2443 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2444 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2445 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2447 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2448 enum bpf_adj_room_mode {
2452 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2453 enum bpf_hdr_start_off {
2458 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2459 enum bpf_lwt_encap_mode {
2461 BPF_LWT_ENCAP_SEG6_INLINE
2464 /* user accessible mirror of in-kernel sk_buff.
2465 * new fields can only be added to the end of this structure
2471 __u32 queue_mapping;
2477 __u32 ingress_ifindex;
2487 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2489 __u32 remote_ip4; /* Stored in network byte order */
2490 __u32 local_ip4; /* Stored in network byte order */
2491 __u32 remote_ip6[4]; /* Stored in network byte order */
2492 __u32 local_ip6[4]; /* Stored in network byte order */
2493 __u32 remote_port; /* Stored in network byte order */
2494 __u32 local_port; /* stored in host byte order */
2498 struct bpf_flow_keys *flow_keys;
2503 struct bpf_tunnel_key {
2507 __u32 remote_ipv6[4];
2511 __u16 tunnel_ext; /* Padding, future use. */
2515 /* user accessible mirror of in-kernel xfrm_state.
2516 * new fields can only be added to the end of this structure
2518 struct bpf_xfrm_state {
2520 __u32 spi; /* Stored in network byte order */
2522 __u16 ext; /* Padding, future use. */
2524 __u32 remote_ipv4; /* Stored in network byte order */
2525 __u32 remote_ipv6[4]; /* Stored in network byte order */
2529 /* Generic BPF return codes which all BPF program types may support.
2530 * The values are binary compatible with their TC_ACT_* counter-part to
2531 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2534 * XDP is handled seprately, see XDP_*.
2542 /* >127 are reserved for prog type specific return codes */
2552 __u32 src_ip4; /* Allows 1,2,4-byte read.
2553 * Stored in network byte order.
2555 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2556 * Stored in network byte order.
2558 __u32 src_port; /* Allows 4-byte read.
2559 * Stored in host byte order
2563 struct bpf_sock_tuple {
2580 #define XDP_PACKET_HEADROOM 256
2582 /* User return codes for XDP prog type.
2583 * A valid XDP program must return one of these defined values. All other
2584 * return codes are reserved for future use. Unknown return codes will
2585 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2595 /* user accessible metadata for XDP packet hook
2596 * new fields must be added to the end of this structure
2602 /* Below access go through struct xdp_rxq_info */
2603 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2604 __u32 rx_queue_index; /* rxq->queue_index */
2612 /* user accessible metadata for SK_MSG packet hook, new fields must
2613 * be added to the end of this structure
2620 __u32 remote_ip4; /* Stored in network byte order */
2621 __u32 local_ip4; /* Stored in network byte order */
2622 __u32 remote_ip6[4]; /* Stored in network byte order */
2623 __u32 local_ip6[4]; /* Stored in network byte order */
2624 __u32 remote_port; /* Stored in network byte order */
2625 __u32 local_port; /* stored in host byte order */
2628 struct sk_reuseport_md {
2630 * Start of directly accessible data. It begins from
2631 * the tcp/udp header.
2634 void *data_end; /* End of directly accessible data */
2636 * Total length of packet (starting from the tcp/udp header).
2637 * Note that the directly accessible bytes (data_end - data)
2638 * could be less than this "len". Those bytes could be
2639 * indirectly read by a helper "bpf_skb_load_bytes()".
2643 * Eth protocol in the mac header (network byte order). e.g.
2644 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2647 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2648 __u32 bind_inany; /* Is sock bound to an INANY address? */
2649 __u32 hash; /* A hash of the packet 4 tuples */
2652 #define BPF_TAG_SIZE 8
2654 struct bpf_prog_info {
2657 __u8 tag[BPF_TAG_SIZE];
2658 __u32 jited_prog_len;
2659 __u32 xlated_prog_len;
2660 __aligned_u64 jited_prog_insns;
2661 __aligned_u64 xlated_prog_insns;
2662 __u64 load_time; /* ns since boottime */
2663 __u32 created_by_uid;
2665 __aligned_u64 map_ids;
2666 char name[BPF_OBJ_NAME_LEN];
2668 __u32 gpl_compatible:1;
2671 __u32 nr_jited_ksyms;
2672 __u32 nr_jited_func_lens;
2673 __aligned_u64 jited_ksyms;
2674 __aligned_u64 jited_func_lens;
2676 __u32 func_info_rec_size;
2677 __aligned_u64 func_info;
2678 __u32 func_info_cnt;
2679 } __attribute__((aligned(8)));
2681 struct bpf_map_info {
2688 char name[BPF_OBJ_NAME_LEN];
2694 __u32 btf_key_type_id;
2695 __u32 btf_value_type_id;
2696 } __attribute__((aligned(8)));
2698 struct bpf_btf_info {
2702 } __attribute__((aligned(8)));
2704 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2705 * by user and intended to be used by socket (e.g. to bind to, depends on
2706 * attach attach type).
2708 struct bpf_sock_addr {
2709 __u32 user_family; /* Allows 4-byte read, but no write. */
2710 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2711 * Stored in network byte order.
2713 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2714 * Stored in network byte order.
2716 __u32 user_port; /* Allows 4-byte read and write.
2717 * Stored in network byte order
2719 __u32 family; /* Allows 4-byte read, but no write */
2720 __u32 type; /* Allows 4-byte read, but no write */
2721 __u32 protocol; /* Allows 4-byte read, but no write */
2722 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2723 * Stored in network byte order.
2725 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2726 * Stored in network byte order.
2730 /* User bpf_sock_ops struct to access socket values and specify request ops
2731 * and their replies.
2732 * Some of this fields are in network (bigendian) byte order and may need
2733 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2734 * New fields can only be added at the end of this structure
2736 struct bpf_sock_ops {
2739 __u32 args[4]; /* Optionally passed to bpf program */
2740 __u32 reply; /* Returned by bpf program */
2741 __u32 replylong[4]; /* Optionally returned by bpf prog */
2744 __u32 remote_ip4; /* Stored in network byte order */
2745 __u32 local_ip4; /* Stored in network byte order */
2746 __u32 remote_ip6[4]; /* Stored in network byte order */
2747 __u32 local_ip6[4]; /* Stored in network byte order */
2748 __u32 remote_port; /* Stored in network byte order */
2749 __u32 local_port; /* stored in host byte order */
2750 __u32 is_fullsock; /* Some TCP fields are only valid if
2751 * there is a full socket. If not, the
2752 * fields read as zero.
2755 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2756 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2765 __u32 rate_delivered;
2766 __u32 rate_interval_us;
2769 __u32 total_retrans;
2773 __u32 data_segs_out;
2777 __u64 bytes_received;
2781 /* Definitions for bpf_sock_ops_cb_flags */
2782 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2783 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2784 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2785 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2786 * supported cb flags
2789 /* List of known BPF sock_ops operators.
2790 * New entries can only be added at the end
2794 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2795 * -1 if default value should be used
2797 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2798 * window (in packets) or -1 if default
2799 * value should be used
2801 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2802 * active connection is initialized
2804 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2805 * active connection is
2808 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2809 * passive connection is
2812 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2815 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2816 * based on the path and may be
2817 * dependent on the congestion control
2818 * algorithm. In general it indicates
2819 * a congestion threshold. RTTs above
2820 * this indicate congestion
2822 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2823 * Arg1: value of icsk_retransmits
2824 * Arg2: value of icsk_rto
2825 * Arg3: whether RTO has expired
2827 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2828 * Arg1: sequence number of 1st byte
2830 * Arg3: return value of
2831 * tcp_transmit_skb (0 => success)
2833 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2837 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
2838 * socket transition to LISTEN state.
2842 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2843 * changes between the TCP and BPF versions. Ideally this should never happen.
2844 * If it does, we need to add code to convert them before calling
2845 * the BPF sock_ops function.
2848 BPF_TCP_ESTABLISHED = 1,
2858 BPF_TCP_CLOSING, /* Now a valid state */
2859 BPF_TCP_NEW_SYN_RECV,
2861 BPF_TCP_MAX_STATES /* Leave at the end! */
2864 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
2865 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
2867 struct bpf_perf_event_value {
2873 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2874 #define BPF_DEVCG_ACC_READ (1ULL << 1)
2875 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2877 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2878 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2880 struct bpf_cgroup_dev_ctx {
2881 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2887 struct bpf_raw_tracepoint_args {
2891 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
2892 * OUTPUT: Do lookup from egress perspective; default is ingress
2894 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
2895 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
2898 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
2899 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
2900 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
2901 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
2902 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
2903 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
2904 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
2905 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
2906 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
2909 struct bpf_fib_lookup {
2910 /* input: network family for lookup (AF_INET, AF_INET6)
2911 * output: network family of egress nexthop
2915 /* set if lookup is to consider L4 data - e.g., FIB rules */
2920 /* total length of packet from network header - used for MTU check */
2923 /* input: L3 device index for lookup
2924 * output: device index from FIB lookup
2929 /* inputs to lookup */
2930 __u8 tos; /* AF_INET */
2931 __be32 flowinfo; /* AF_INET6, flow_label + priority */
2933 /* output: metric of fib result (IPv4/IPv6 only) */
2939 __u32 ipv6_src[4]; /* in6_addr; network order */
2942 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2943 * network header. output: bpf_fib_lookup sets to gateway address
2944 * if FIB lookup returns gateway route
2948 __u32 ipv6_dst[4]; /* in6_addr; network order */
2952 __be16 h_vlan_proto;
2954 __u8 smac[6]; /* ETH_ALEN */
2955 __u8 dmac[6]; /* ETH_ALEN */
2958 enum bpf_task_fd_type {
2959 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
2960 BPF_FD_TYPE_TRACEPOINT, /* tp name */
2961 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
2962 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
2963 BPF_FD_TYPE_UPROBE, /* filename + offset */
2964 BPF_FD_TYPE_URETPROBE, /* filename + offset */
2967 struct bpf_flow_keys {
2970 __u16 addr_proto; /* ETH_P_* of valid addrs */
2984 __u32 ipv6_src[4]; /* in6_addr; network order */
2985 __u32 ipv6_dst[4]; /* in6_addr; network order */
2990 struct bpf_func_info {
2995 #endif /* _UAPI__LINUX_BPF_H__ */