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 /* BPF syscall commands, see bpf(2) man-page for details. */
93 BPF_PROG_GET_FD_BY_ID,
95 BPF_OBJ_GET_INFO_BY_FD,
97 BPF_RAW_TRACEPOINT_OPEN,
107 BPF_MAP_TYPE_PROG_ARRAY,
108 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
109 BPF_MAP_TYPE_PERCPU_HASH,
110 BPF_MAP_TYPE_PERCPU_ARRAY,
111 BPF_MAP_TYPE_STACK_TRACE,
112 BPF_MAP_TYPE_CGROUP_ARRAY,
113 BPF_MAP_TYPE_LRU_HASH,
114 BPF_MAP_TYPE_LRU_PERCPU_HASH,
115 BPF_MAP_TYPE_LPM_TRIE,
116 BPF_MAP_TYPE_ARRAY_OF_MAPS,
117 BPF_MAP_TYPE_HASH_OF_MAPS,
119 BPF_MAP_TYPE_SOCKMAP,
122 BPF_MAP_TYPE_SOCKHASH,
126 BPF_PROG_TYPE_UNSPEC,
127 BPF_PROG_TYPE_SOCKET_FILTER,
128 BPF_PROG_TYPE_KPROBE,
129 BPF_PROG_TYPE_SCHED_CLS,
130 BPF_PROG_TYPE_SCHED_ACT,
131 BPF_PROG_TYPE_TRACEPOINT,
133 BPF_PROG_TYPE_PERF_EVENT,
134 BPF_PROG_TYPE_CGROUP_SKB,
135 BPF_PROG_TYPE_CGROUP_SOCK,
136 BPF_PROG_TYPE_LWT_IN,
137 BPF_PROG_TYPE_LWT_OUT,
138 BPF_PROG_TYPE_LWT_XMIT,
139 BPF_PROG_TYPE_SOCK_OPS,
140 BPF_PROG_TYPE_SK_SKB,
141 BPF_PROG_TYPE_CGROUP_DEVICE,
142 BPF_PROG_TYPE_SK_MSG,
143 BPF_PROG_TYPE_RAW_TRACEPOINT,
144 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
145 BPF_PROG_TYPE_LWT_SEG6LOCAL,
148 enum bpf_attach_type {
149 BPF_CGROUP_INET_INGRESS,
150 BPF_CGROUP_INET_EGRESS,
151 BPF_CGROUP_INET_SOCK_CREATE,
153 BPF_SK_SKB_STREAM_PARSER,
154 BPF_SK_SKB_STREAM_VERDICT,
157 BPF_CGROUP_INET4_BIND,
158 BPF_CGROUP_INET6_BIND,
159 BPF_CGROUP_INET4_CONNECT,
160 BPF_CGROUP_INET6_CONNECT,
161 BPF_CGROUP_INET4_POST_BIND,
162 BPF_CGROUP_INET6_POST_BIND,
163 BPF_CGROUP_UDP4_SENDMSG,
164 BPF_CGROUP_UDP6_SENDMSG,
165 __MAX_BPF_ATTACH_TYPE
168 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
170 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
172 * NONE(default): No further bpf programs allowed in the subtree.
174 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
175 * the program in this cgroup yields to sub-cgroup program.
177 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
178 * that cgroup program gets run in addition to the program in this cgroup.
180 * Only one program is allowed to be attached to a cgroup with
181 * NONE or BPF_F_ALLOW_OVERRIDE flag.
182 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
183 * release old program and attach the new one. Attach flags has to match.
185 * Multiple programs are allowed to be attached to a cgroup with
186 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
187 * (those that were attached first, run first)
188 * The programs of sub-cgroup are executed first, then programs of
189 * this cgroup and then programs of parent cgroup.
190 * When children program makes decision (like picking TCP CA or sock bind)
191 * parent program has a chance to override it.
193 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
194 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
196 * cgrp1 (MULTI progs A, B) ->
197 * cgrp2 (OVERRIDE prog C) ->
198 * cgrp3 (MULTI prog D) ->
199 * cgrp4 (OVERRIDE prog E) ->
200 * cgrp5 (NONE prog F)
201 * the event in cgrp5 triggers execution of F,D,A,B in that order.
202 * if prog F is detached, the execution is E,D,A,B
203 * if prog F and D are detached, the execution is E,A,B
204 * if prog F, E and D are detached, the execution is C,A,B
206 * All eligible programs are executed regardless of return code from
209 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
210 #define BPF_F_ALLOW_MULTI (1U << 1)
212 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
213 * verifier will perform strict alignment checking as if the kernel
214 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
215 * and NET_IP_ALIGN defined to 2.
217 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
219 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
220 #define BPF_PSEUDO_MAP_FD 1
222 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
223 * offset to another bpf function
225 #define BPF_PSEUDO_CALL 1
227 /* flags for BPF_MAP_UPDATE_ELEM command */
228 #define BPF_ANY 0 /* create new element or update existing */
229 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
230 #define BPF_EXIST 2 /* update existing element */
232 /* flags for BPF_MAP_CREATE command */
233 #define BPF_F_NO_PREALLOC (1U << 0)
234 /* Instead of having one common LRU list in the
235 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
236 * which can scale and perform better.
237 * Note, the LRU nodes (including free nodes) cannot be moved
238 * across different LRU lists.
240 #define BPF_F_NO_COMMON_LRU (1U << 1)
241 /* Specify numa node during map creation */
242 #define BPF_F_NUMA_NODE (1U << 2)
244 /* flags for BPF_PROG_QUERY */
245 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
247 #define BPF_OBJ_NAME_LEN 16U
249 /* Flags for accessing BPF object */
250 #define BPF_F_RDONLY (1U << 3)
251 #define BPF_F_WRONLY (1U << 4)
253 /* Flag for stack_map, store build_id+offset instead of pointer */
254 #define BPF_F_STACK_BUILD_ID (1U << 5)
256 enum bpf_stack_build_id_status {
257 /* user space need an empty entry to identify end of a trace */
258 BPF_STACK_BUILD_ID_EMPTY = 0,
259 /* with valid build_id and offset */
260 BPF_STACK_BUILD_ID_VALID = 1,
261 /* couldn't get build_id, fallback to ip */
262 BPF_STACK_BUILD_ID_IP = 2,
265 #define BPF_BUILD_ID_SIZE 20
266 struct bpf_stack_build_id {
268 unsigned char build_id[BPF_BUILD_ID_SIZE];
276 struct { /* anonymous struct used by BPF_MAP_CREATE command */
277 __u32 map_type; /* one of enum bpf_map_type */
278 __u32 key_size; /* size of key in bytes */
279 __u32 value_size; /* size of value in bytes */
280 __u32 max_entries; /* max number of entries in a map */
281 __u32 map_flags; /* BPF_MAP_CREATE related
282 * flags defined above.
284 __u32 inner_map_fd; /* fd pointing to the inner map */
285 __u32 numa_node; /* numa node (effective only if
286 * BPF_F_NUMA_NODE is set).
288 char map_name[BPF_OBJ_NAME_LEN];
289 __u32 map_ifindex; /* ifindex of netdev to create on */
290 __u32 btf_fd; /* fd pointing to a BTF type data */
291 __u32 btf_key_type_id; /* BTF type_id of the key */
292 __u32 btf_value_type_id; /* BTF type_id of the value */
295 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
300 __aligned_u64 next_key;
305 struct { /* anonymous struct used by BPF_PROG_LOAD command */
306 __u32 prog_type; /* one of enum bpf_prog_type */
309 __aligned_u64 license;
310 __u32 log_level; /* verbosity level of verifier */
311 __u32 log_size; /* size of user buffer */
312 __aligned_u64 log_buf; /* user supplied buffer */
313 __u32 kern_version; /* checked when prog_type=kprobe */
315 char prog_name[BPF_OBJ_NAME_LEN];
316 __u32 prog_ifindex; /* ifindex of netdev to prep for */
317 /* For some prog types expected attach type must be known at
318 * load time to verify attach type specific parts of prog
319 * (context accesses, allowed helpers, etc).
321 __u32 expected_attach_type;
324 struct { /* anonymous struct used by BPF_OBJ_* commands */
325 __aligned_u64 pathname;
330 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
331 __u32 target_fd; /* container object to attach to */
332 __u32 attach_bpf_fd; /* eBPF program to attach */
337 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
342 __aligned_u64 data_in;
343 __aligned_u64 data_out;
348 struct { /* anonymous struct used by BPF_*_GET_*_ID */
359 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
365 struct { /* anonymous struct used by BPF_PROG_QUERY command */
366 __u32 target_fd; /* container object to query */
370 __aligned_u64 prog_ids;
379 struct { /* anonymous struct for BPF_BTF_LOAD */
381 __aligned_u64 btf_log_buf;
388 __u32 pid; /* input: pid */
389 __u32 fd; /* input: fd */
390 __u32 flags; /* input: flags */
391 __u32 buf_len; /* input/output: buf len */
392 __aligned_u64 buf; /* input/output:
393 * tp_name for tracepoint
395 * filename for uprobe
397 __u32 prog_id; /* output: prod_id */
398 __u32 fd_type; /* output: BPF_FD_TYPE_* */
399 __u64 probe_offset; /* output: probe_offset */
400 __u64 probe_addr; /* output: probe_addr */
402 } __attribute__((aligned(8)));
404 /* The description below is an attempt at providing documentation to eBPF
405 * developers about the multiple available eBPF helper functions. It can be
406 * parsed and used to produce a manual page. The workflow is the following,
407 * and requires the rst2man utility:
409 * $ ./scripts/bpf_helpers_doc.py \
410 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
411 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
412 * $ man /tmp/bpf-helpers.7
414 * Note that in order to produce this external documentation, some RST
415 * formatting is used in the descriptions to get "bold" and "italics" in
416 * manual pages. Also note that the few trailing white spaces are
417 * intentional, removing them would break paragraphs for rst2man.
419 * Start of BPF helper function descriptions:
421 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
423 * Perform a lookup in *map* for an entry associated to *key*.
425 * Map value associated to *key*, or **NULL** if no entry was
428 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
430 * Add or update the value of the entry associated to *key* in
431 * *map* with *value*. *flags* is one of:
434 * The entry for *key* must not exist in the map.
436 * The entry for *key* must already exist in the map.
438 * No condition on the existence of the entry for *key*.
440 * Flag value **BPF_NOEXIST** cannot be used for maps of types
441 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
442 * elements always exist), the helper would return an error.
444 * 0 on success, or a negative error in case of failure.
446 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
448 * Delete entry with *key* from *map*.
450 * 0 on success, or a negative error in case of failure.
452 * int bpf_probe_read(void *dst, u32 size, const void *src)
454 * For tracing programs, safely attempt to read *size* bytes from
455 * address *src* and store the data in *dst*.
457 * 0 on success, or a negative error in case of failure.
459 * u64 bpf_ktime_get_ns(void)
461 * Return the time elapsed since system boot, in nanoseconds.
465 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
467 * This helper is a "printk()-like" facility for debugging. It
468 * prints a message defined by format *fmt* (of size *fmt_size*)
469 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
470 * available. It can take up to three additional **u64**
471 * arguments (as an eBPF helpers, the total number of arguments is
474 * Each time the helper is called, it appends a line to the trace.
475 * The format of the trace is customizable, and the exact output
476 * one will get depends on the options set in
477 * *\/sys/kernel/debug/tracing/trace_options* (see also the
478 * *README* file under the same directory). However, it usually
479 * defaults to something like:
483 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
487 * * ``telnet`` is the name of the current task.
488 * * ``470`` is the PID of the current task.
489 * * ``001`` is the CPU number on which the task is
491 * * In ``.N..``, each character refers to a set of
492 * options (whether irqs are enabled, scheduling
493 * options, whether hard/softirqs are running, level of
494 * preempt_disabled respectively). **N** means that
495 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
497 * * ``419421.045894`` is a timestamp.
498 * * ``0x00000001`` is a fake value used by BPF for the
499 * instruction pointer register.
500 * * ``<formatted msg>`` is the message formatted with
503 * The conversion specifiers supported by *fmt* are similar, but
504 * more limited than for printk(). They are **%d**, **%i**,
505 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
506 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
507 * of field, padding with zeroes, etc.) is available, and the
508 * helper will return **-EINVAL** (but print nothing) if it
509 * encounters an unknown specifier.
511 * Also, note that **bpf_trace_printk**\ () is slow, and should
512 * only be used for debugging purposes. For this reason, a notice
513 * bloc (spanning several lines) is printed to kernel logs and
514 * states that the helper should not be used "for production use"
515 * the first time this helper is used (or more precisely, when
516 * **trace_printk**\ () buffers are allocated). For passing values
517 * to user space, perf events should be preferred.
519 * The number of bytes written to the buffer, or a negative error
520 * in case of failure.
522 * u32 bpf_get_prandom_u32(void)
524 * Get a pseudo-random number.
526 * From a security point of view, this helper uses its own
527 * pseudo-random internal state, and cannot be used to infer the
528 * seed of other random functions in the kernel. However, it is
529 * essential to note that the generator used by the helper is not
530 * cryptographically secure.
532 * A random 32-bit unsigned value.
534 * u32 bpf_get_smp_processor_id(void)
536 * Get the SMP (symmetric multiprocessing) processor id. Note that
537 * all programs run with preemption disabled, which means that the
538 * SMP processor id is stable during all the execution of the
541 * The SMP id of the processor running the program.
543 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
545 * Store *len* bytes from address *from* into the packet
546 * associated to *skb*, at *offset*. *flags* are a combination of
547 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
548 * checksum for the packet after storing the bytes) and
549 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
550 * **->swhash** and *skb*\ **->l4hash** to 0).
552 * A call to this helper is susceptible to change the underlaying
553 * packet buffer. Therefore, at load time, all checks on pointers
554 * previously done by the verifier are invalidated and must be
555 * performed again, if the helper is used in combination with
556 * direct packet access.
558 * 0 on success, or a negative error in case of failure.
560 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
562 * Recompute the layer 3 (e.g. IP) checksum for the packet
563 * associated to *skb*. Computation is incremental, so the helper
564 * must know the former value of the header field that was
565 * modified (*from*), the new value of this field (*to*), and the
566 * number of bytes (2 or 4) for this field, stored in *size*.
567 * Alternatively, it is possible to store the difference between
568 * the previous and the new values of the header field in *to*, by
569 * setting *from* and *size* to 0. For both methods, *offset*
570 * indicates the location of the IP checksum within the packet.
572 * This helper works in combination with **bpf_csum_diff**\ (),
573 * which does not update the checksum in-place, but offers more
574 * flexibility and can handle sizes larger than 2 or 4 for the
575 * checksum to update.
577 * A call to this helper is susceptible to change the underlaying
578 * packet buffer. Therefore, at load time, all checks on pointers
579 * previously done by the verifier are invalidated and must be
580 * performed again, if the helper is used in combination with
581 * direct packet access.
583 * 0 on success, or a negative error in case of failure.
585 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
587 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
588 * packet associated to *skb*. Computation is incremental, so the
589 * helper must know the former value of the header field that was
590 * modified (*from*), the new value of this field (*to*), and the
591 * number of bytes (2 or 4) for this field, stored on the lowest
592 * four bits of *flags*. Alternatively, it is possible to store
593 * the difference between the previous and the new values of the
594 * header field in *to*, by setting *from* and the four lowest
595 * bits of *flags* to 0. For both methods, *offset* indicates the
596 * location of the IP checksum within the packet. In addition to
597 * the size of the field, *flags* can be added (bitwise OR) actual
598 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
599 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
600 * for updates resulting in a null checksum the value is set to
601 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
602 * the checksum is to be computed against a pseudo-header.
604 * This helper works in combination with **bpf_csum_diff**\ (),
605 * which does not update the checksum in-place, but offers more
606 * flexibility and can handle sizes larger than 2 or 4 for the
607 * checksum to update.
609 * A call to this helper is susceptible to change the underlaying
610 * packet buffer. Therefore, at load time, all checks on pointers
611 * previously done by the verifier are invalidated and must be
612 * performed again, if the helper is used in combination with
613 * direct packet access.
615 * 0 on success, or a negative error in case of failure.
617 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
619 * This special helper is used to trigger a "tail call", or in
620 * other words, to jump into another eBPF program. The same stack
621 * frame is used (but values on stack and in registers for the
622 * caller are not accessible to the callee). This mechanism allows
623 * for program chaining, either for raising the maximum number of
624 * available eBPF instructions, or to execute given programs in
625 * conditional blocks. For security reasons, there is an upper
626 * limit to the number of successive tail calls that can be
629 * Upon call of this helper, the program attempts to jump into a
630 * program referenced at index *index* in *prog_array_map*, a
631 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
632 * *ctx*, a pointer to the context.
634 * If the call succeeds, the kernel immediately runs the first
635 * instruction of the new program. This is not a function call,
636 * and it never returns to the previous program. If the call
637 * fails, then the helper has no effect, and the caller continues
638 * to run its subsequent instructions. A call can fail if the
639 * destination program for the jump does not exist (i.e. *index*
640 * is superior to the number of entries in *prog_array_map*), or
641 * if the maximum number of tail calls has been reached for this
642 * chain of programs. This limit is defined in the kernel by the
643 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
644 * which is currently set to 32.
646 * 0 on success, or a negative error in case of failure.
648 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
650 * Clone and redirect the packet associated to *skb* to another
651 * net device of index *ifindex*. Both ingress and egress
652 * interfaces can be used for redirection. The **BPF_F_INGRESS**
653 * value in *flags* is used to make the distinction (ingress path
654 * is selected if the flag is present, egress path otherwise).
655 * This is the only flag supported for now.
657 * In comparison with **bpf_redirect**\ () helper,
658 * **bpf_clone_redirect**\ () has the associated cost of
659 * duplicating the packet buffer, but this can be executed out of
660 * the eBPF program. Conversely, **bpf_redirect**\ () is more
661 * efficient, but it is handled through an action code where the
662 * redirection happens only after the eBPF program has returned.
664 * A call to this helper is susceptible to change the underlaying
665 * packet buffer. Therefore, at load time, all checks on pointers
666 * previously done by the verifier are invalidated and must be
667 * performed again, if the helper is used in combination with
668 * direct packet access.
670 * 0 on success, or a negative error in case of failure.
672 * u64 bpf_get_current_pid_tgid(void)
674 * A 64-bit integer containing the current tgid and pid, and
676 * *current_task*\ **->tgid << 32 \|**
677 * *current_task*\ **->pid**.
679 * u64 bpf_get_current_uid_gid(void)
681 * A 64-bit integer containing the current GID and UID, and
682 * created as such: *current_gid* **<< 32 \|** *current_uid*.
684 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
686 * Copy the **comm** attribute of the current task into *buf* of
687 * *size_of_buf*. The **comm** attribute contains the name of
688 * the executable (excluding the path) for the current task. The
689 * *size_of_buf* must be strictly positive. On success, the
690 * helper makes sure that the *buf* is NUL-terminated. On failure,
691 * it is filled with zeroes.
693 * 0 on success, or a negative error in case of failure.
695 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
697 * Retrieve the classid for the current task, i.e. for the net_cls
698 * cgroup to which *skb* belongs.
700 * This helper can be used on TC egress path, but not on ingress.
702 * The net_cls cgroup provides an interface to tag network packets
703 * based on a user-provided identifier for all traffic coming from
704 * the tasks belonging to the related cgroup. See also the related
705 * kernel documentation, available from the Linux sources in file
706 * *Documentation/cgroup-v1/net_cls.txt*.
708 * The Linux kernel has two versions for cgroups: there are
709 * cgroups v1 and cgroups v2. Both are available to users, who can
710 * use a mixture of them, but note that the net_cls cgroup is for
711 * cgroup v1 only. This makes it incompatible with BPF programs
712 * run on cgroups, which is a cgroup-v2-only feature (a socket can
713 * only hold data for one version of cgroups at a time).
715 * This helper is only available is the kernel was compiled with
716 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
717 * "**y**" or to "**m**".
719 * The classid, or 0 for the default unconfigured classid.
721 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
723 * Push a *vlan_tci* (VLAN tag control information) of protocol
724 * *vlan_proto* to the packet associated to *skb*, then update
725 * the checksum. Note that if *vlan_proto* is different from
726 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
727 * be **ETH_P_8021Q**.
729 * A call to this helper is susceptible to change the underlaying
730 * packet buffer. Therefore, at load time, all checks on pointers
731 * previously done by the verifier are invalidated and must be
732 * performed again, if the helper is used in combination with
733 * direct packet access.
735 * 0 on success, or a negative error in case of failure.
737 * int bpf_skb_vlan_pop(struct sk_buff *skb)
739 * Pop a VLAN header from the packet associated to *skb*.
741 * A call to this helper is susceptible to change the underlaying
742 * packet buffer. Therefore, at load time, all checks on pointers
743 * previously done by the verifier are invalidated and must be
744 * performed again, if the helper is used in combination with
745 * direct packet access.
747 * 0 on success, or a negative error in case of failure.
749 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
751 * Get tunnel metadata. This helper takes a pointer *key* to an
752 * empty **struct bpf_tunnel_key** of **size**, that will be
753 * filled with tunnel metadata for the packet associated to *skb*.
754 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
755 * indicates that the tunnel is based on IPv6 protocol instead of
758 * The **struct bpf_tunnel_key** is an object that generalizes the
759 * principal parameters used by various tunneling protocols into a
760 * single struct. This way, it can be used to easily make a
761 * decision based on the contents of the encapsulation header,
762 * "summarized" in this struct. In particular, it holds the IP
763 * address of the remote end (IPv4 or IPv6, depending on the case)
764 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
765 * this struct exposes the *key*\ **->tunnel_id**, which is
766 * generally mapped to a VNI (Virtual Network Identifier), making
767 * it programmable together with the **bpf_skb_set_tunnel_key**\
770 * Let's imagine that the following code is part of a program
771 * attached to the TC ingress interface, on one end of a GRE
772 * tunnel, and is supposed to filter out all messages coming from
773 * remote ends with IPv4 address other than 10.0.0.1:
778 * struct bpf_tunnel_key key = {};
780 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
782 * return TC_ACT_SHOT; // drop packet
784 * if (key.remote_ipv4 != 0x0a000001)
785 * return TC_ACT_SHOT; // drop packet
787 * return TC_ACT_OK; // accept packet
789 * This interface can also be used with all encapsulation devices
790 * that can operate in "collect metadata" mode: instead of having
791 * one network device per specific configuration, the "collect
792 * metadata" mode only requires a single device where the
793 * configuration can be extracted from this helper.
795 * This can be used together with various tunnels such as VXLan,
796 * Geneve, GRE or IP in IP (IPIP).
798 * 0 on success, or a negative error in case of failure.
800 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
802 * Populate tunnel metadata for packet associated to *skb.* The
803 * tunnel metadata is set to the contents of *key*, of *size*. The
804 * *flags* can be set to a combination of the following values:
806 * **BPF_F_TUNINFO_IPV6**
807 * Indicate that the tunnel is based on IPv6 protocol
809 * **BPF_F_ZERO_CSUM_TX**
810 * For IPv4 packets, add a flag to tunnel metadata
811 * indicating that checksum computation should be skipped
812 * and checksum set to zeroes.
813 * **BPF_F_DONT_FRAGMENT**
814 * Add a flag to tunnel metadata indicating that the
815 * packet should not be fragmented.
816 * **BPF_F_SEQ_NUMBER**
817 * Add a flag to tunnel metadata indicating that a
818 * sequence number should be added to tunnel header before
819 * sending the packet. This flag was added for GRE
820 * encapsulation, but might be used with other protocols
821 * as well in the future.
823 * Here is a typical usage on the transmit path:
827 * struct bpf_tunnel_key key;
829 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
830 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
832 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
833 * helper for additional information.
835 * 0 on success, or a negative error in case of failure.
837 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
839 * Read the value of a perf event counter. This helper relies on a
840 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
841 * the perf event counter is selected when *map* is updated with
842 * perf event file descriptors. The *map* is an array whose size
843 * is the number of available CPUs, and each cell contains a value
844 * relative to one CPU. The value to retrieve is indicated by
845 * *flags*, that contains the index of the CPU to look up, masked
846 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
847 * **BPF_F_CURRENT_CPU** to indicate that the value for the
848 * current CPU should be retrieved.
850 * Note that before Linux 4.13, only hardware perf event can be
853 * Also, be aware that the newer helper
854 * **bpf_perf_event_read_value**\ () is recommended over
855 * **bpf_perf_event_read**\ () in general. The latter has some ABI
856 * quirks where error and counter value are used as a return code
857 * (which is wrong to do since ranges may overlap). This issue is
858 * fixed with **bpf_perf_event_read_value**\ (), which at the same
859 * time provides more features over the **bpf_perf_event_read**\
860 * () interface. Please refer to the description of
861 * **bpf_perf_event_read_value**\ () for details.
863 * The value of the perf event counter read from the map, or a
864 * negative error code in case of failure.
866 * int bpf_redirect(u32 ifindex, u64 flags)
868 * Redirect the packet to another net device of index *ifindex*.
869 * This helper is somewhat similar to **bpf_clone_redirect**\
870 * (), except that the packet is not cloned, which provides
871 * increased performance.
873 * Except for XDP, both ingress and egress interfaces can be used
874 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
875 * to make the distinction (ingress path is selected if the flag
876 * is present, egress path otherwise). Currently, XDP only
877 * supports redirection to the egress interface, and accepts no
880 * The same effect can be attained with the more generic
881 * **bpf_redirect_map**\ (), which requires specific maps to be
882 * used but offers better performance.
884 * For XDP, the helper returns **XDP_REDIRECT** on success or
885 * **XDP_ABORTED** on error. For other program types, the values
886 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
889 * u32 bpf_get_route_realm(struct sk_buff *skb)
891 * Retrieve the realm or the route, that is to say the
892 * **tclassid** field of the destination for the *skb*. The
893 * indentifier retrieved is a user-provided tag, similar to the
894 * one used with the net_cls cgroup (see description for
895 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
896 * held by a route (a destination entry), not by a task.
898 * Retrieving this identifier works with the clsact TC egress hook
899 * (see also **tc-bpf(8)**), or alternatively on conventional
900 * classful egress qdiscs, but not on TC ingress path. In case of
901 * clsact TC egress hook, this has the advantage that, internally,
902 * the destination entry has not been dropped yet in the transmit
903 * path. Therefore, the destination entry does not need to be
904 * artificially held via **netif_keep_dst**\ () for a classful
905 * qdisc until the *skb* is freed.
907 * This helper is available only if the kernel was compiled with
908 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
910 * The realm of the route for the packet associated to *skb*, or 0
913 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
915 * Write raw *data* blob into a special BPF perf event held by
916 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
917 * event must have the following attributes: **PERF_SAMPLE_RAW**
918 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
919 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
921 * The *flags* are used to indicate the index in *map* for which
922 * the value must be put, masked with **BPF_F_INDEX_MASK**.
923 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
924 * to indicate that the index of the current CPU core should be
927 * The value to write, of *size*, is passed through eBPF stack and
930 * The context of the program *ctx* needs also be passed to the
933 * On user space, a program willing to read the values needs to
934 * call **perf_event_open**\ () on the perf event (either for
935 * one or for all CPUs) and to store the file descriptor into the
936 * *map*. This must be done before the eBPF program can send data
937 * into it. An example is available in file
938 * *samples/bpf/trace_output_user.c* in the Linux kernel source
939 * tree (the eBPF program counterpart is in
940 * *samples/bpf/trace_output_kern.c*).
942 * **bpf_perf_event_output**\ () achieves better performance
943 * than **bpf_trace_printk**\ () for sharing data with user
944 * space, and is much better suitable for streaming data from eBPF
947 * Note that this helper is not restricted to tracing use cases
948 * and can be used with programs attached to TC or XDP as well,
949 * where it allows for passing data to user space listeners. Data
952 * * Only custom structs,
953 * * Only the packet payload, or
954 * * A combination of both.
956 * 0 on success, or a negative error in case of failure.
958 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
960 * This helper was provided as an easy way to load data from a
961 * packet. It can be used to load *len* bytes from *offset* from
962 * the packet associated to *skb*, into the buffer pointed by
965 * Since Linux 4.7, usage of this helper has mostly been replaced
966 * by "direct packet access", enabling packet data to be
967 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
968 * pointing respectively to the first byte of packet data and to
969 * the byte after the last byte of packet data. However, it
970 * remains useful if one wishes to read large quantities of data
971 * at once from a packet into the eBPF stack.
973 * 0 on success, or a negative error in case of failure.
975 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
977 * Walk a user or a kernel stack and return its id. To achieve
978 * this, the helper needs *ctx*, which is a pointer to the context
979 * on which the tracing program is executed, and a pointer to a
980 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
982 * The last argument, *flags*, holds the number of stack frames to
983 * skip (from 0 to 255), masked with
984 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
985 * a combination of the following flags:
987 * **BPF_F_USER_STACK**
988 * Collect a user space stack instead of a kernel stack.
989 * **BPF_F_FAST_STACK_CMP**
990 * Compare stacks by hash only.
991 * **BPF_F_REUSE_STACKID**
992 * If two different stacks hash into the same *stackid*,
993 * discard the old one.
995 * The stack id retrieved is a 32 bit long integer handle which
996 * can be further combined with other data (including other stack
997 * ids) and used as a key into maps. This can be useful for
998 * generating a variety of graphs (such as flame graphs or off-cpu
1001 * For walking a stack, this helper is an improvement over
1002 * **bpf_probe_read**\ (), which can be used with unrolled loops
1003 * but is not efficient and consumes a lot of eBPF instructions.
1004 * Instead, **bpf_get_stackid**\ () can collect up to
1005 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1006 * this limit can be controlled with the **sysctl** program, and
1007 * that it should be manually increased in order to profile long
1008 * user stacks (such as stacks for Java programs). To do so, use:
1012 * # sysctl kernel.perf_event_max_stack=<new value>
1015 * The positive or null stack id on success, or a negative error
1016 * in case of failure.
1018 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1020 * Compute a checksum difference, from the raw buffer pointed by
1021 * *from*, of length *from_size* (that must be a multiple of 4),
1022 * towards the raw buffer pointed by *to*, of size *to_size*
1023 * (same remark). An optional *seed* can be added to the value
1024 * (this can be cascaded, the seed may come from a previous call
1027 * This is flexible enough to be used in several ways:
1029 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1030 * checksum, it can be used when pushing new data.
1031 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1032 * checksum, it can be used when removing data from a packet.
1033 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1034 * can be used to compute a diff. Note that *from_size* and
1035 * *to_size* do not need to be equal.
1037 * This helper can be used in combination with
1038 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1039 * which one can feed in the difference computed with
1040 * **bpf_csum_diff**\ ().
1042 * The checksum result, or a negative error code in case of
1045 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1047 * Retrieve tunnel options metadata for the packet associated to
1048 * *skb*, and store the raw tunnel option data to the buffer *opt*
1051 * This helper can be used with encapsulation devices that can
1052 * operate in "collect metadata" mode (please refer to the related
1053 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1054 * more details). A particular example where this can be used is
1055 * in combination with the Geneve encapsulation protocol, where it
1056 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1057 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1058 * the eBPF program. This allows for full customization of these
1061 * The size of the option data retrieved.
1063 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1065 * Set tunnel options metadata for the packet associated to *skb*
1066 * to the option data contained in the raw buffer *opt* of *size*.
1068 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1069 * helper for additional information.
1071 * 0 on success, or a negative error in case of failure.
1073 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1075 * Change the protocol of the *skb* to *proto*. Currently
1076 * supported are transition from IPv4 to IPv6, and from IPv6 to
1077 * IPv4. The helper takes care of the groundwork for the
1078 * transition, including resizing the socket buffer. The eBPF
1079 * program is expected to fill the new headers, if any, via
1080 * **skb_store_bytes**\ () and to recompute the checksums with
1081 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1082 * (). The main case for this helper is to perform NAT64
1083 * operations out of an eBPF program.
1085 * Internally, the GSO type is marked as dodgy so that headers are
1086 * checked and segments are recalculated by the GSO/GRO engine.
1087 * The size for GSO target is adapted as well.
1089 * All values for *flags* are reserved for future usage, and must
1092 * A call to this helper is susceptible to change the underlaying
1093 * packet buffer. Therefore, at load time, all checks on pointers
1094 * previously done by the verifier are invalidated and must be
1095 * performed again, if the helper is used in combination with
1096 * direct packet access.
1098 * 0 on success, or a negative error in case of failure.
1100 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1102 * Change the packet type for the packet associated to *skb*. This
1103 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1104 * the eBPF program does not have a write access to *skb*\
1105 * **->pkt_type** beside this helper. Using a helper here allows
1106 * for graceful handling of errors.
1108 * The major use case is to change incoming *skb*s to
1109 * **PACKET_HOST** in a programmatic way instead of having to
1110 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1113 * Note that *type* only allows certain values. At this time, they
1118 * **PACKET_BROADCAST**
1119 * Send packet to all.
1120 * **PACKET_MULTICAST**
1121 * Send packet to group.
1122 * **PACKET_OTHERHOST**
1123 * Send packet to someone else.
1125 * 0 on success, or a negative error in case of failure.
1127 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1129 * Check whether *skb* is a descendant of the cgroup2 held by
1130 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1132 * The return value depends on the result of the test, and can be:
1134 * * 0, if the *skb* failed the cgroup2 descendant test.
1135 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1136 * * A negative error code, if an error occurred.
1138 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1140 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1141 * not set, in particular if the hash was cleared due to mangling,
1142 * recompute this hash. Later accesses to the hash can be done
1143 * directly with *skb*\ **->hash**.
1145 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1146 * prototype with **bpf_skb_change_proto**\ (), or calling
1147 * **bpf_skb_store_bytes**\ () with the
1148 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1149 * the hash and to trigger a new computation for the next call to
1150 * **bpf_get_hash_recalc**\ ().
1154 * u64 bpf_get_current_task(void)
1156 * A pointer to the current task struct.
1158 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1160 * Attempt in a safe way to write *len* bytes from the buffer
1161 * *src* to *dst* in memory. It only works for threads that are in
1162 * user context, and *dst* must be a valid user space address.
1164 * This helper should not be used to implement any kind of
1165 * security mechanism because of TOC-TOU attacks, but rather to
1166 * debug, divert, and manipulate execution of semi-cooperative
1169 * Keep in mind that this feature is meant for experiments, and it
1170 * has a risk of crashing the system and running programs.
1171 * Therefore, when an eBPF program using this helper is attached,
1172 * a warning including PID and process name is printed to kernel
1175 * 0 on success, or a negative error in case of failure.
1177 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1179 * Check whether the probe is being run is the context of a given
1180 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1181 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1183 * The return value depends on the result of the test, and can be:
1185 * * 0, if the *skb* task belongs to the cgroup2.
1186 * * 1, if the *skb* task does not belong to the cgroup2.
1187 * * A negative error code, if an error occurred.
1189 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1191 * Resize (trim or grow) the packet associated to *skb* to the
1192 * new *len*. The *flags* are reserved for future usage, and must
1195 * The basic idea is that the helper performs the needed work to
1196 * change the size of the packet, then the eBPF program rewrites
1197 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1198 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1199 * and others. This helper is a slow path utility intended for
1200 * replies with control messages. And because it is targeted for
1201 * slow path, the helper itself can afford to be slow: it
1202 * implicitly linearizes, unclones and drops offloads from the
1205 * A call to this helper is susceptible to change the underlaying
1206 * packet buffer. Therefore, at load time, all checks on pointers
1207 * previously done by the verifier are invalidated and must be
1208 * performed again, if the helper is used in combination with
1209 * direct packet access.
1211 * 0 on success, or a negative error in case of failure.
1213 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1215 * Pull in non-linear data in case the *skb* is non-linear and not
1216 * all of *len* are part of the linear section. Make *len* bytes
1217 * from *skb* readable and writable. If a zero value is passed for
1218 * *len*, then the whole length of the *skb* is pulled.
1220 * This helper is only needed for reading and writing with direct
1223 * For direct packet access, testing that offsets to access
1224 * are within packet boundaries (test on *skb*\ **->data_end**) is
1225 * susceptible to fail if offsets are invalid, or if the requested
1226 * data is in non-linear parts of the *skb*. On failure the
1227 * program can just bail out, or in the case of a non-linear
1228 * buffer, use a helper to make the data available. The
1229 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1230 * the data. Another one consists in using **bpf_skb_pull_data**
1231 * to pull in once the non-linear parts, then retesting and
1232 * eventually access the data.
1234 * At the same time, this also makes sure the *skb* is uncloned,
1235 * which is a necessary condition for direct write. As this needs
1236 * to be an invariant for the write part only, the verifier
1237 * detects writes and adds a prologue that is calling
1238 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1239 * the very beginning in case it is indeed cloned.
1241 * A call to this helper is susceptible to change the underlaying
1242 * packet buffer. Therefore, at load time, all checks on pointers
1243 * previously done by the verifier are invalidated and must be
1244 * performed again, if the helper is used in combination with
1245 * direct packet access.
1247 * 0 on success, or a negative error in case of failure.
1249 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1251 * Add the checksum *csum* into *skb*\ **->csum** in case the
1252 * driver has supplied a checksum for the entire packet into that
1253 * field. Return an error otherwise. This helper is intended to be
1254 * used in combination with **bpf_csum_diff**\ (), in particular
1255 * when the checksum needs to be updated after data has been
1256 * written into the packet through direct packet access.
1258 * The checksum on success, or a negative error code in case of
1261 * void bpf_set_hash_invalid(struct sk_buff *skb)
1263 * Invalidate the current *skb*\ **->hash**. It can be used after
1264 * mangling on headers through direct packet access, in order to
1265 * indicate that the hash is outdated and to trigger a
1266 * recalculation the next time the kernel tries to access this
1267 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1269 * int bpf_get_numa_node_id(void)
1271 * Return the id of the current NUMA node. The primary use case
1272 * for this helper is the selection of sockets for the local NUMA
1273 * node, when the program is attached to sockets using the
1274 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1275 * but the helper is also available to other eBPF program types,
1276 * similarly to **bpf_get_smp_processor_id**\ ().
1278 * The id of current NUMA node.
1280 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1282 * Grows headroom of packet associated to *skb* and adjusts the
1283 * offset of the MAC header accordingly, adding *len* bytes of
1284 * space. It automatically extends and reallocates memory as
1287 * This helper can be used on a layer 3 *skb* to push a MAC header
1288 * for redirection into a layer 2 device.
1290 * All values for *flags* are reserved for future usage, and must
1293 * A call to this helper is susceptible to change the underlaying
1294 * packet buffer. Therefore, at load time, all checks on pointers
1295 * previously done by the verifier are invalidated and must be
1296 * performed again, if the helper is used in combination with
1297 * direct packet access.
1299 * 0 on success, or a negative error in case of failure.
1301 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1303 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1304 * it is possible to use a negative value for *delta*. This helper
1305 * can be used to prepare the packet for pushing or popping
1308 * A call to this helper is susceptible to change the underlaying
1309 * packet buffer. Therefore, at load time, all checks on pointers
1310 * previously done by the verifier are invalidated and must be
1311 * performed again, if the helper is used in combination with
1312 * direct packet access.
1314 * 0 on success, or a negative error in case of failure.
1316 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1318 * Copy a NUL terminated string from an unsafe address
1319 * *unsafe_ptr* to *dst*. The *size* should include the
1320 * terminating NUL byte. In case the string length is smaller than
1321 * *size*, the target is not padded with further NUL bytes. If the
1322 * string length is larger than *size*, just *size*-1 bytes are
1323 * copied and the last byte is set to NUL.
1325 * On success, the length of the copied string is returned. This
1326 * makes this helper useful in tracing programs for reading
1327 * strings, and more importantly to get its length at runtime. See
1328 * the following snippet:
1332 * SEC("kprobe/sys_open")
1333 * void bpf_sys_open(struct pt_regs *ctx)
1335 * char buf[PATHLEN]; // PATHLEN is defined to 256
1336 * int res = bpf_probe_read_str(buf, sizeof(buf),
1339 * // Consume buf, for example push it to
1340 * // userspace via bpf_perf_event_output(); we
1341 * // can use res (the string length) as event
1342 * // size, after checking its boundaries.
1345 * In comparison, using **bpf_probe_read()** helper here instead
1346 * to read the string would require to estimate the length at
1347 * compile time, and would often result in copying more memory
1350 * Another useful use case is when parsing individual process
1351 * arguments or individual environment variables navigating
1352 * *current*\ **->mm->arg_start** and *current*\
1353 * **->mm->env_start**: using this helper and the return value,
1354 * one can quickly iterate at the right offset of the memory area.
1356 * On success, the strictly positive length of the string,
1357 * including the trailing NUL character. On error, a negative
1360 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1362 * If the **struct sk_buff** pointed by *skb* has a known socket,
1363 * retrieve the cookie (generated by the kernel) of this socket.
1364 * If no cookie has been set yet, generate a new cookie. Once
1365 * generated, the socket cookie remains stable for the life of the
1366 * socket. This helper can be useful for monitoring per socket
1367 * networking traffic statistics as it provides a unique socket
1368 * identifier per namespace.
1370 * A 8-byte long non-decreasing number on success, or 0 if the
1371 * socket field is missing inside *skb*.
1373 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1375 * The owner UID of the socket associated to *skb*. If the socket
1376 * is **NULL**, or if it is not a full socket (i.e. if it is a
1377 * time-wait or a request socket instead), **overflowuid** value
1378 * is returned (note that **overflowuid** might also be the actual
1379 * UID value for the socket).
1381 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1383 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1388 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1390 * Emulate a call to **setsockopt()** on the socket associated to
1391 * *bpf_socket*, which must be a full socket. The *level* at
1392 * which the option resides and the name *optname* of the option
1393 * must be specified, see **setsockopt(2)** for more information.
1394 * The option value of length *optlen* is pointed by *optval*.
1396 * This helper actually implements a subset of **setsockopt()**.
1397 * It supports the following *level*\ s:
1399 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1400 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1401 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1402 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1403 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1404 * **TCP_BPF_SNDCWND_CLAMP**.
1405 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1406 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1408 * 0 on success, or a negative error in case of failure.
1410 * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags)
1412 * Grow or shrink the room for data in the packet associated to
1413 * *skb* by *len_diff*, and according to the selected *mode*.
1415 * There is a single supported mode at this time:
1417 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1418 * (room space is added or removed below the layer 3 header).
1420 * All values for *flags* are reserved for future usage, and must
1423 * A call to this helper is susceptible to change the underlaying
1424 * packet buffer. Therefore, at load time, all checks on pointers
1425 * previously done by the verifier are invalidated and must be
1426 * performed again, if the helper is used in combination with
1427 * direct packet access.
1429 * 0 on success, or a negative error in case of failure.
1431 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1433 * Redirect the packet to the endpoint referenced by *map* at
1434 * index *key*. Depending on its type, this *map* can contain
1435 * references to net devices (for forwarding packets through other
1436 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1437 * but this is only implemented for native XDP (with driver
1438 * support) as of this writing).
1440 * All values for *flags* are reserved for future usage, and must
1443 * When used to redirect packets to net devices, this helper
1444 * provides a high performance increase over **bpf_redirect**\ ().
1445 * This is due to various implementation details of the underlying
1446 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1447 * () tries to send packet as a "bulk" to the device.
1449 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1451 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1453 * Redirect the packet to the socket referenced by *map* (of type
1454 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1455 * egress interfaces can be used for redirection. The
1456 * **BPF_F_INGRESS** value in *flags* is used to make the
1457 * distinction (ingress path is selected if the flag is present,
1458 * egress path otherwise). This is the only flag supported for now.
1460 * **SK_PASS** on success, or **SK_DROP** on error.
1462 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1464 * Add an entry to, or update a *map* referencing sockets. The
1465 * *skops* is used as a new value for the entry associated to
1466 * *key*. *flags* is one of:
1469 * The entry for *key* must not exist in the map.
1471 * The entry for *key* must already exist in the map.
1473 * No condition on the existence of the entry for *key*.
1475 * If the *map* has eBPF programs (parser and verdict), those will
1476 * be inherited by the socket being added. If the socket is
1477 * already attached to eBPF programs, this results in an error.
1479 * 0 on success, or a negative error in case of failure.
1481 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1483 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1484 * *delta* (which can be positive or negative). Note that this
1485 * operation modifies the address stored in *xdp_md*\ **->data**,
1486 * so the latter must be loaded only after the helper has been
1489 * The use of *xdp_md*\ **->data_meta** is optional and programs
1490 * are not required to use it. The rationale is that when the
1491 * packet is processed with XDP (e.g. as DoS filter), it is
1492 * possible to push further meta data along with it before passing
1493 * to the stack, and to give the guarantee that an ingress eBPF
1494 * program attached as a TC classifier on the same device can pick
1495 * this up for further post-processing. Since TC works with socket
1496 * buffers, it remains possible to set from XDP the **mark** or
1497 * **priority** pointers, or other pointers for the socket buffer.
1498 * Having this scratch space generic and programmable allows for
1499 * more flexibility as the user is free to store whatever meta
1502 * A call to this helper is susceptible to change the underlaying
1503 * packet buffer. Therefore, at load time, all checks on pointers
1504 * previously done by the verifier are invalidated and must be
1505 * performed again, if the helper is used in combination with
1506 * direct packet access.
1508 * 0 on success, or a negative error in case of failure.
1510 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1512 * Read the value of a perf event counter, and store it into *buf*
1513 * of size *buf_size*. This helper relies on a *map* of type
1514 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1515 * counter is selected when *map* is updated with perf event file
1516 * descriptors. The *map* is an array whose size is the number of
1517 * available CPUs, and each cell contains a value relative to one
1518 * CPU. The value to retrieve is indicated by *flags*, that
1519 * contains the index of the CPU to look up, masked with
1520 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1521 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1522 * current CPU should be retrieved.
1524 * This helper behaves in a way close to
1525 * **bpf_perf_event_read**\ () helper, save that instead of
1526 * just returning the value observed, it fills the *buf*
1527 * structure. This allows for additional data to be retrieved: in
1528 * particular, the enabled and running times (in *buf*\
1529 * **->enabled** and *buf*\ **->running**, respectively) are
1530 * copied. In general, **bpf_perf_event_read_value**\ () is
1531 * recommended over **bpf_perf_event_read**\ (), which has some
1532 * ABI issues and provides fewer functionalities.
1534 * These values are interesting, because hardware PMU (Performance
1535 * Monitoring Unit) counters are limited resources. When there are
1536 * more PMU based perf events opened than available counters,
1537 * kernel will multiplex these events so each event gets certain
1538 * percentage (but not all) of the PMU time. In case that
1539 * multiplexing happens, the number of samples or counter value
1540 * will not reflect the case compared to when no multiplexing
1541 * occurs. This makes comparison between different runs difficult.
1542 * Typically, the counter value should be normalized before
1543 * comparing to other experiments. The usual normalization is done
1548 * normalized_counter = counter * t_enabled / t_running
1550 * Where t_enabled is the time enabled for event and t_running is
1551 * the time running for event since last normalization. The
1552 * enabled and running times are accumulated since the perf event
1553 * open. To achieve scaling factor between two invocations of an
1554 * eBPF program, users can can use CPU id as the key (which is
1555 * typical for perf array usage model) to remember the previous
1556 * value and do the calculation inside the eBPF program.
1558 * 0 on success, or a negative error in case of failure.
1560 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1562 * For en eBPF program attached to a perf event, retrieve the
1563 * value of the event counter associated to *ctx* and store it in
1564 * the structure pointed by *buf* and of size *buf_size*. Enabled
1565 * and running times are also stored in the structure (see
1566 * description of helper **bpf_perf_event_read_value**\ () for
1569 * 0 on success, or a negative error in case of failure.
1571 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1573 * Emulate a call to **getsockopt()** on the socket associated to
1574 * *bpf_socket*, which must be a full socket. The *level* at
1575 * which the option resides and the name *optname* of the option
1576 * must be specified, see **getsockopt(2)** for more information.
1577 * The retrieved value is stored in the structure pointed by
1578 * *opval* and of length *optlen*.
1580 * This helper actually implements a subset of **getsockopt()**.
1581 * It supports the following *level*\ s:
1583 * * **IPPROTO_TCP**, which supports *optname*
1584 * **TCP_CONGESTION**.
1585 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1586 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1588 * 0 on success, or a negative error in case of failure.
1590 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1592 * Used for error injection, this helper uses kprobes to override
1593 * the return value of the probed function, and to set it to *rc*.
1594 * The first argument is the context *regs* on which the kprobe
1597 * This helper works by setting setting the PC (program counter)
1598 * to an override function which is run in place of the original
1599 * probed function. This means the probed function is not run at
1600 * all. The replacement function just returns with the required
1603 * This helper has security implications, and thus is subject to
1604 * restrictions. It is only available if the kernel was compiled
1605 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1606 * option, and in this case it only works on functions tagged with
1607 * **ALLOW_ERROR_INJECTION** in the kernel code.
1609 * Also, the helper is only available for the architectures having
1610 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1611 * x86 architecture is the only one to support this feature.
1615 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1617 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1618 * for the full TCP socket associated to *bpf_sock_ops* to
1621 * The primary use of this field is to determine if there should
1622 * be calls to eBPF programs of type
1623 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1624 * code. A program of the same type can change its value, per
1625 * connection and as necessary, when the connection is
1626 * established. This field is directly accessible for reading, but
1627 * this helper must be used for updates in order to return an
1628 * error if an eBPF program tries to set a callback that is not
1629 * supported in the current kernel.
1631 * The supported callback values that *argval* can combine are:
1633 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1634 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1635 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1637 * Here are some examples of where one could call such eBPF
1641 * * When a packet is retransmitted.
1642 * * When the connection terminates.
1643 * * When a packet is sent.
1644 * * When a packet is received.
1646 * Code **-EINVAL** if the socket is not a full TCP socket;
1647 * otherwise, a positive number containing the bits that could not
1648 * be set is returned (which comes down to 0 if all bits were set
1651 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1653 * This helper is used in programs implementing policies at the
1654 * socket level. If the message *msg* is allowed to pass (i.e. if
1655 * the verdict eBPF program returns **SK_PASS**), redirect it to
1656 * the socket referenced by *map* (of type
1657 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1658 * egress interfaces can be used for redirection. The
1659 * **BPF_F_INGRESS** value in *flags* is used to make the
1660 * distinction (ingress path is selected if the flag is present,
1661 * egress path otherwise). This is the only flag supported for now.
1663 * **SK_PASS** on success, or **SK_DROP** on error.
1665 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1667 * For socket policies, apply the verdict of the eBPF program to
1668 * the next *bytes* (number of bytes) of message *msg*.
1670 * For example, this helper can be used in the following cases:
1672 * * A single **sendmsg**\ () or **sendfile**\ () system call
1673 * contains multiple logical messages that the eBPF program is
1674 * supposed to read and for which it should apply a verdict.
1675 * * An eBPF program only cares to read the first *bytes* of a
1676 * *msg*. If the message has a large payload, then setting up
1677 * and calling the eBPF program repeatedly for all bytes, even
1678 * though the verdict is already known, would create unnecessary
1681 * When called from within an eBPF program, the helper sets a
1682 * counter internal to the BPF infrastructure, that is used to
1683 * apply the last verdict to the next *bytes*. If *bytes* is
1684 * smaller than the current data being processed from a
1685 * **sendmsg**\ () or **sendfile**\ () system call, the first
1686 * *bytes* will be sent and the eBPF program will be re-run with
1687 * the pointer for start of data pointing to byte number *bytes*
1688 * **+ 1**. If *bytes* is larger than the current data being
1689 * processed, then the eBPF verdict will be applied to multiple
1690 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1693 * Note that if a socket closes with the internal counter holding
1694 * a non-zero value, this is not a problem because data is not
1695 * being buffered for *bytes* and is sent as it is received.
1699 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1701 * For socket policies, prevent the execution of the verdict eBPF
1702 * program for message *msg* until *bytes* (byte number) have been
1705 * This can be used when one needs a specific number of bytes
1706 * before a verdict can be assigned, even if the data spans
1707 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1708 * case would be a user calling **sendmsg**\ () repeatedly with
1709 * 1-byte long message segments. Obviously, this is bad for
1710 * performance, but it is still valid. If the eBPF program needs
1711 * *bytes* bytes to validate a header, this helper can be used to
1712 * prevent the eBPF program to be called again until *bytes* have
1717 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1719 * For socket policies, pull in non-linear data from user space
1720 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1721 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1724 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1725 * *msg* it can only parse data that the (**data**, **data_end**)
1726 * pointers have already consumed. For **sendmsg**\ () hooks this
1727 * is likely the first scatterlist element. But for calls relying
1728 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1729 * be the range (**0**, **0**) because the data is shared with
1730 * user space and by default the objective is to avoid allowing
1731 * user space to modify data while (or after) eBPF verdict is
1732 * being decided. This helper can be used to pull in data and to
1733 * set the start and end pointer to given values. Data will be
1734 * copied if necessary (i.e. if data was not linear and if start
1735 * and end pointers do not point to the same chunk).
1737 * A call to this helper is susceptible to change the underlaying
1738 * packet buffer. Therefore, at load time, all checks on pointers
1739 * previously done by the verifier are invalidated and must be
1740 * performed again, if the helper is used in combination with
1741 * direct packet access.
1743 * All values for *flags* are reserved for future usage, and must
1746 * 0 on success, or a negative error in case of failure.
1748 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1750 * Bind the socket associated to *ctx* to the address pointed by
1751 * *addr*, of length *addr_len*. This allows for making outgoing
1752 * connection from the desired IP address, which can be useful for
1753 * example when all processes inside a cgroup should use one
1754 * single IP address on a host that has multiple IP configured.
1756 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1757 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1758 * **AF_INET6**). Looking for a free port to bind to can be
1759 * expensive, therefore binding to port is not permitted by the
1760 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1761 * must be set to zero.
1763 * 0 on success, or a negative error in case of failure.
1765 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1767 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1768 * only possible to shrink the packet as of this writing,
1769 * therefore *delta* must be a negative integer.
1771 * A call to this helper is susceptible to change the underlaying
1772 * packet buffer. Therefore, at load time, all checks on pointers
1773 * previously done by the verifier are invalidated and must be
1774 * performed again, if the helper is used in combination with
1775 * direct packet access.
1777 * 0 on success, or a negative error in case of failure.
1779 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1781 * Retrieve the XFRM state (IP transform framework, see also
1782 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1784 * The retrieved value is stored in the **struct bpf_xfrm_state**
1785 * pointed by *xfrm_state* and of length *size*.
1787 * All values for *flags* are reserved for future usage, and must
1790 * This helper is available only if the kernel was compiled with
1791 * **CONFIG_XFRM** configuration option.
1793 * 0 on success, or a negative error in case of failure.
1795 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1797 * Return a user or a kernel stack in bpf program provided buffer.
1798 * To achieve this, the helper needs *ctx*, which is a pointer
1799 * to the context on which the tracing program is executed.
1800 * To store the stacktrace, the bpf program provides *buf* with
1801 * a nonnegative *size*.
1803 * The last argument, *flags*, holds the number of stack frames to
1804 * skip (from 0 to 255), masked with
1805 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1806 * the following flags:
1808 * **BPF_F_USER_STACK**
1809 * Collect a user space stack instead of a kernel stack.
1810 * **BPF_F_USER_BUILD_ID**
1811 * Collect buildid+offset instead of ips for user stack,
1812 * only valid if **BPF_F_USER_STACK** is also specified.
1814 * **bpf_get_stack**\ () can collect up to
1815 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1816 * to sufficient large buffer size. Note that
1817 * this limit can be controlled with the **sysctl** program, and
1818 * that it should be manually increased in order to profile long
1819 * user stacks (such as stacks for Java programs). To do so, use:
1823 * # sysctl kernel.perf_event_max_stack=<new value>
1826 * a non-negative value equal to or less than size on success, or
1827 * a negative error in case of failure.
1829 * int skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1831 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1832 * it provides an easy way to load *len* bytes from *offset*
1833 * from the packet associated to *skb*, into the buffer pointed
1834 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1835 * a fifth argument *start_header* exists in order to select a
1836 * base offset to start from. *start_header* can be one of:
1838 * **BPF_HDR_START_MAC**
1839 * Base offset to load data from is *skb*'s mac header.
1840 * **BPF_HDR_START_NET**
1841 * Base offset to load data from is *skb*'s network header.
1843 * In general, "direct packet access" is the preferred method to
1844 * access packet data, however, this helper is in particular useful
1845 * in socket filters where *skb*\ **->data** does not always point
1846 * to the start of the mac header and where "direct packet access"
1850 * 0 on success, or a negative error in case of failure.
1852 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1854 * Do FIB lookup in kernel tables using parameters in *params*.
1855 * If lookup is successful and result shows packet is to be
1856 * forwarded, the neighbor tables are searched for the nexthop.
1857 * If successful (ie., FIB lookup shows forwarding and nexthop
1858 * is resolved), the nexthop address is returned in ipv4_dst,
1859 * ipv6_dst or mpls_out based on family, smac is set to mac
1860 * address of egress device, dmac is set to nexthop mac address,
1861 * rt_metric is set to metric from route.
1863 * *plen* argument is the size of the passed in struct.
1864 * *flags* argument can be one or more BPF_FIB_LOOKUP_ flags:
1866 * **BPF_FIB_LOOKUP_DIRECT** means do a direct table lookup vs
1867 * full lookup using FIB rules
1868 * **BPF_FIB_LOOKUP_OUTPUT** means do lookup from an egress
1869 * perspective (default is ingress)
1871 * *ctx* is either **struct xdp_md** for XDP programs or
1872 * **struct sk_buff** tc cls_act programs.
1875 * Egress device index on success, 0 if packet needs to continue
1876 * up the stack for further processing or a negative error in case
1879 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1881 * Add an entry to, or update a sockhash *map* referencing sockets.
1882 * The *skops* is used as a new value for the entry associated to
1883 * *key*. *flags* is one of:
1886 * The entry for *key* must not exist in the map.
1888 * The entry for *key* must already exist in the map.
1890 * No condition on the existence of the entry for *key*.
1892 * If the *map* has eBPF programs (parser and verdict), those will
1893 * be inherited by the socket being added. If the socket is
1894 * already attached to eBPF programs, this results in an error.
1896 * 0 on success, or a negative error in case of failure.
1898 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1900 * This helper is used in programs implementing policies at the
1901 * socket level. If the message *msg* is allowed to pass (i.e. if
1902 * the verdict eBPF program returns **SK_PASS**), redirect it to
1903 * the socket referenced by *map* (of type
1904 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1905 * egress interfaces can be used for redirection. The
1906 * **BPF_F_INGRESS** value in *flags* is used to make the
1907 * distinction (ingress path is selected if the flag is present,
1908 * egress path otherwise). This is the only flag supported for now.
1910 * **SK_PASS** on success, or **SK_DROP** on error.
1912 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1914 * This helper is used in programs implementing policies at the
1915 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1916 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1917 * to the socket referenced by *map* (of type
1918 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1919 * egress interfaces can be used for redirection. The
1920 * **BPF_F_INGRESS** value in *flags* is used to make the
1921 * distinction (ingress path is selected if the flag is present,
1922 * egress otherwise). This is the only flag supported for now.
1924 * **SK_PASS** on success, or **SK_DROP** on error.
1926 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
1928 * Encapsulate the packet associated to *skb* within a Layer 3
1929 * protocol header. This header is provided in the buffer at
1930 * address *hdr*, with *len* its size in bytes. *type* indicates
1931 * the protocol of the header and can be one of:
1933 * **BPF_LWT_ENCAP_SEG6**
1934 * IPv6 encapsulation with Segment Routing Header
1935 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
1936 * the IPv6 header is computed by the kernel.
1937 * **BPF_LWT_ENCAP_SEG6_INLINE**
1938 * Only works if *skb* contains an IPv6 packet. Insert a
1939 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
1942 * A call to this helper is susceptible to change the underlaying
1943 * packet buffer. Therefore, at load time, all checks on pointers
1944 * previously done by the verifier are invalidated and must be
1945 * performed again, if the helper is used in combination with
1946 * direct packet access.
1948 * 0 on success, or a negative error in case of failure.
1950 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
1952 * Store *len* bytes from address *from* into the packet
1953 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
1954 * inside the outermost IPv6 Segment Routing Header can be
1955 * modified through this helper.
1957 * A call to this helper is susceptible to change the underlaying
1958 * packet buffer. Therefore, at load time, all checks on pointers
1959 * previously done by the verifier are invalidated and must be
1960 * performed again, if the helper is used in combination with
1961 * direct packet access.
1963 * 0 on success, or a negative error in case of failure.
1965 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
1967 * Adjust the size allocated to TLVs in the outermost IPv6
1968 * Segment Routing Header contained in the packet associated to
1969 * *skb*, at position *offset* by *delta* bytes. Only offsets
1970 * after the segments are accepted. *delta* can be as well
1971 * positive (growing) as negative (shrinking).
1973 * A call to this helper is susceptible to change the underlaying
1974 * packet buffer. Therefore, at load time, all checks on pointers
1975 * previously done by the verifier are invalidated and must be
1976 * performed again, if the helper is used in combination with
1977 * direct packet access.
1979 * 0 on success, or a negative error in case of failure.
1981 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
1983 * Apply an IPv6 Segment Routing action of type *action* to the
1984 * packet associated to *skb*. Each action takes a parameter
1985 * contained at address *param*, and of length *param_len* bytes.
1986 * *action* can be one of:
1988 * **SEG6_LOCAL_ACTION_END_X**
1989 * End.X action: Endpoint with Layer-3 cross-connect.
1990 * Type of *param*: **struct in6_addr**.
1991 * **SEG6_LOCAL_ACTION_END_T**
1992 * End.T action: Endpoint with specific IPv6 table lookup.
1993 * Type of *param*: **int**.
1994 * **SEG6_LOCAL_ACTION_END_B6**
1995 * End.B6 action: Endpoint bound to an SRv6 policy.
1996 * Type of param: **struct ipv6_sr_hdr**.
1997 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
1998 * End.B6.Encap action: Endpoint bound to an SRv6
1999 * encapsulation policy.
2000 * Type of param: **struct ipv6_sr_hdr**.
2002 * A call to this helper is susceptible to change the underlaying
2003 * packet buffer. Therefore, at load time, all checks on pointers
2004 * previously done by the verifier are invalidated and must be
2005 * performed again, if the helper is used in combination with
2006 * direct packet access.
2008 * 0 on success, or a negative error in case of failure.
2010 #define __BPF_FUNC_MAPPER(FN) \
2012 FN(map_lookup_elem), \
2013 FN(map_update_elem), \
2014 FN(map_delete_elem), \
2018 FN(get_prandom_u32), \
2019 FN(get_smp_processor_id), \
2020 FN(skb_store_bytes), \
2021 FN(l3_csum_replace), \
2022 FN(l4_csum_replace), \
2024 FN(clone_redirect), \
2025 FN(get_current_pid_tgid), \
2026 FN(get_current_uid_gid), \
2027 FN(get_current_comm), \
2028 FN(get_cgroup_classid), \
2029 FN(skb_vlan_push), \
2031 FN(skb_get_tunnel_key), \
2032 FN(skb_set_tunnel_key), \
2033 FN(perf_event_read), \
2035 FN(get_route_realm), \
2036 FN(perf_event_output), \
2037 FN(skb_load_bytes), \
2040 FN(skb_get_tunnel_opt), \
2041 FN(skb_set_tunnel_opt), \
2042 FN(skb_change_proto), \
2043 FN(skb_change_type), \
2044 FN(skb_under_cgroup), \
2045 FN(get_hash_recalc), \
2046 FN(get_current_task), \
2047 FN(probe_write_user), \
2048 FN(current_task_under_cgroup), \
2049 FN(skb_change_tail), \
2050 FN(skb_pull_data), \
2052 FN(set_hash_invalid), \
2053 FN(get_numa_node_id), \
2054 FN(skb_change_head), \
2055 FN(xdp_adjust_head), \
2056 FN(probe_read_str), \
2057 FN(get_socket_cookie), \
2058 FN(get_socket_uid), \
2061 FN(skb_adjust_room), \
2063 FN(sk_redirect_map), \
2064 FN(sock_map_update), \
2065 FN(xdp_adjust_meta), \
2066 FN(perf_event_read_value), \
2067 FN(perf_prog_read_value), \
2069 FN(override_return), \
2070 FN(sock_ops_cb_flags_set), \
2071 FN(msg_redirect_map), \
2072 FN(msg_apply_bytes), \
2073 FN(msg_cork_bytes), \
2074 FN(msg_pull_data), \
2076 FN(xdp_adjust_tail), \
2077 FN(skb_get_xfrm_state), \
2079 FN(skb_load_bytes_relative), \
2081 FN(sock_hash_update), \
2082 FN(msg_redirect_hash), \
2083 FN(sk_redirect_hash), \
2084 FN(lwt_push_encap), \
2085 FN(lwt_seg6_store_bytes), \
2086 FN(lwt_seg6_adjust_srh), \
2087 FN(lwt_seg6_action),
2089 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2090 * function eBPF program intends to call
2092 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2094 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2097 #undef __BPF_ENUM_FN
2099 /* All flags used by eBPF helper functions, placed here. */
2101 /* BPF_FUNC_skb_store_bytes flags. */
2102 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2103 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2105 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2106 * First 4 bits are for passing the header field size.
2108 #define BPF_F_HDR_FIELD_MASK 0xfULL
2110 /* BPF_FUNC_l4_csum_replace flags. */
2111 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2112 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2113 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2115 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2116 #define BPF_F_INGRESS (1ULL << 0)
2118 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2119 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2121 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2122 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2123 #define BPF_F_USER_STACK (1ULL << 8)
2124 /* flags used by BPF_FUNC_get_stackid only. */
2125 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2126 #define BPF_F_REUSE_STACKID (1ULL << 10)
2127 /* flags used by BPF_FUNC_get_stack only. */
2128 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2130 /* BPF_FUNC_skb_set_tunnel_key flags. */
2131 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2132 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2133 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2135 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2136 * BPF_FUNC_perf_event_read_value flags.
2138 #define BPF_F_INDEX_MASK 0xffffffffULL
2139 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2140 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2141 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2143 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2144 enum bpf_adj_room_mode {
2148 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2149 enum bpf_hdr_start_off {
2154 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2155 enum bpf_lwt_encap_mode {
2157 BPF_LWT_ENCAP_SEG6_INLINE
2160 /* user accessible mirror of in-kernel sk_buff.
2161 * new fields can only be added to the end of this structure
2167 __u32 queue_mapping;
2173 __u32 ingress_ifindex;
2183 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2185 __u32 remote_ip4; /* Stored in network byte order */
2186 __u32 local_ip4; /* Stored in network byte order */
2187 __u32 remote_ip6[4]; /* Stored in network byte order */
2188 __u32 local_ip6[4]; /* Stored in network byte order */
2189 __u32 remote_port; /* Stored in network byte order */
2190 __u32 local_port; /* stored in host byte order */
2196 struct bpf_tunnel_key {
2200 __u32 remote_ipv6[4];
2208 /* user accessible mirror of in-kernel xfrm_state.
2209 * new fields can only be added to the end of this structure
2211 struct bpf_xfrm_state {
2213 __u32 spi; /* Stored in network byte order */
2216 __u32 remote_ipv4; /* Stored in network byte order */
2217 __u32 remote_ipv6[4]; /* Stored in network byte order */
2221 /* Generic BPF return codes which all BPF program types may support.
2222 * The values are binary compatible with their TC_ACT_* counter-part to
2223 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2226 * XDP is handled seprately, see XDP_*.
2234 /* >127 are reserved for prog type specific return codes */
2244 __u32 src_ip4; /* Allows 1,2,4-byte read.
2245 * Stored in network byte order.
2247 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2248 * Stored in network byte order.
2250 __u32 src_port; /* Allows 4-byte read.
2251 * Stored in host byte order
2255 #define XDP_PACKET_HEADROOM 256
2257 /* User return codes for XDP prog type.
2258 * A valid XDP program must return one of these defined values. All other
2259 * return codes are reserved for future use. Unknown return codes will
2260 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2270 /* user accessible metadata for XDP packet hook
2271 * new fields must be added to the end of this structure
2277 /* Below access go through struct xdp_rxq_info */
2278 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2279 __u32 rx_queue_index; /* rxq->queue_index */
2287 /* user accessible metadata for SK_MSG packet hook, new fields must
2288 * be added to the end of this structure
2295 __u32 remote_ip4; /* Stored in network byte order */
2296 __u32 local_ip4; /* Stored in network byte order */
2297 __u32 remote_ip6[4]; /* Stored in network byte order */
2298 __u32 local_ip6[4]; /* Stored in network byte order */
2299 __u32 remote_port; /* Stored in network byte order */
2300 __u32 local_port; /* stored in host byte order */
2303 #define BPF_TAG_SIZE 8
2305 struct bpf_prog_info {
2308 __u8 tag[BPF_TAG_SIZE];
2309 __u32 jited_prog_len;
2310 __u32 xlated_prog_len;
2311 __aligned_u64 jited_prog_insns;
2312 __aligned_u64 xlated_prog_insns;
2313 __u64 load_time; /* ns since boottime */
2314 __u32 created_by_uid;
2316 __aligned_u64 map_ids;
2317 char name[BPF_OBJ_NAME_LEN];
2319 __u32 gpl_compatible:1;
2322 __u32 nr_jited_ksyms;
2323 __u32 nr_jited_func_lens;
2324 __aligned_u64 jited_ksyms;
2325 __aligned_u64 jited_func_lens;
2326 } __attribute__((aligned(8)));
2328 struct bpf_map_info {
2335 char name[BPF_OBJ_NAME_LEN];
2340 __u32 btf_key_type_id;
2341 __u32 btf_value_type_id;
2342 } __attribute__((aligned(8)));
2344 struct bpf_btf_info {
2348 } __attribute__((aligned(8)));
2350 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2351 * by user and intended to be used by socket (e.g. to bind to, depends on
2352 * attach attach type).
2354 struct bpf_sock_addr {
2355 __u32 user_family; /* Allows 4-byte read, but no write. */
2356 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2357 * Stored in network byte order.
2359 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2360 * Stored in network byte order.
2362 __u32 user_port; /* Allows 4-byte read and write.
2363 * Stored in network byte order
2365 __u32 family; /* Allows 4-byte read, but no write */
2366 __u32 type; /* Allows 4-byte read, but no write */
2367 __u32 protocol; /* Allows 4-byte read, but no write */
2368 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2369 * Stored in network byte order.
2371 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2372 * Stored in network byte order.
2376 /* User bpf_sock_ops struct to access socket values and specify request ops
2377 * and their replies.
2378 * Some of this fields are in network (bigendian) byte order and may need
2379 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2380 * New fields can only be added at the end of this structure
2382 struct bpf_sock_ops {
2385 __u32 args[4]; /* Optionally passed to bpf program */
2386 __u32 reply; /* Returned by bpf program */
2387 __u32 replylong[4]; /* Optionally returned by bpf prog */
2390 __u32 remote_ip4; /* Stored in network byte order */
2391 __u32 local_ip4; /* Stored in network byte order */
2392 __u32 remote_ip6[4]; /* Stored in network byte order */
2393 __u32 local_ip6[4]; /* Stored in network byte order */
2394 __u32 remote_port; /* Stored in network byte order */
2395 __u32 local_port; /* stored in host byte order */
2396 __u32 is_fullsock; /* Some TCP fields are only valid if
2397 * there is a full socket. If not, the
2398 * fields read as zero.
2401 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2402 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2411 __u32 rate_delivered;
2412 __u32 rate_interval_us;
2415 __u32 total_retrans;
2419 __u32 data_segs_out;
2423 __u64 bytes_received;
2427 /* Definitions for bpf_sock_ops_cb_flags */
2428 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2429 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2430 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2431 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2432 * supported cb flags
2435 /* List of known BPF sock_ops operators.
2436 * New entries can only be added at the end
2440 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2441 * -1 if default value should be used
2443 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2444 * window (in packets) or -1 if default
2445 * value should be used
2447 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2448 * active connection is initialized
2450 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2451 * active connection is
2454 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2455 * passive connection is
2458 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2461 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2462 * based on the path and may be
2463 * dependent on the congestion control
2464 * algorithm. In general it indicates
2465 * a congestion threshold. RTTs above
2466 * this indicate congestion
2468 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2469 * Arg1: value of icsk_retransmits
2470 * Arg2: value of icsk_rto
2471 * Arg3: whether RTO has expired
2473 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2474 * Arg1: sequence number of 1st byte
2476 * Arg3: return value of
2477 * tcp_transmit_skb (0 => success)
2479 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2485 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2486 * changes between the TCP and BPF versions. Ideally this should never happen.
2487 * If it does, we need to add code to convert them before calling
2488 * the BPF sock_ops function.
2491 BPF_TCP_ESTABLISHED = 1,
2501 BPF_TCP_CLOSING, /* Now a valid state */
2502 BPF_TCP_NEW_SYN_RECV,
2504 BPF_TCP_MAX_STATES /* Leave at the end! */
2507 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
2508 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
2510 struct bpf_perf_event_value {
2516 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2517 #define BPF_DEVCG_ACC_READ (1ULL << 1)
2518 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2520 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2521 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2523 struct bpf_cgroup_dev_ctx {
2524 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2530 struct bpf_raw_tracepoint_args {
2534 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
2535 * OUTPUT: Do lookup from egress perspective; default is ingress
2537 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
2538 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
2540 struct bpf_fib_lookup {
2542 __u8 family; /* network family, AF_INET, AF_INET6, AF_MPLS */
2544 /* set if lookup is to consider L4 data - e.g., FIB rules */
2549 /* total length of packet from network header - used for MTU check */
2551 __u32 ifindex; /* L3 device index for lookup */
2554 /* inputs to lookup */
2555 __u8 tos; /* AF_INET */
2556 __be32 flowlabel; /* AF_INET6 */
2558 /* output: metric of fib result */
2565 __u32 ipv6_src[4]; /* in6_addr; network order */
2568 /* input to bpf_fib_lookup, *dst is destination address.
2569 * output: bpf_fib_lookup sets to gateway address
2572 /* return for MPLS lookups */
2573 __be32 mpls_out[4]; /* support up to 4 labels */
2575 __u32 ipv6_dst[4]; /* in6_addr; network order */
2579 __be16 h_vlan_proto;
2581 __u8 smac[6]; /* ETH_ALEN */
2582 __u8 dmac[6]; /* ETH_ALEN */
2585 enum bpf_task_fd_type {
2586 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
2587 BPF_FD_TYPE_TRACEPOINT, /* tp name */
2588 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
2589 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
2590 BPF_FD_TYPE_UPROBE, /* filename + offset */
2591 BPF_FD_TYPE_URETPROBE, /* filename + offset */
2594 #endif /* _UAPI__LINUX_BPF_H__ */