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,
146 BPF_PROG_TYPE_LIRC_MODE2,
149 enum bpf_attach_type {
150 BPF_CGROUP_INET_INGRESS,
151 BPF_CGROUP_INET_EGRESS,
152 BPF_CGROUP_INET_SOCK_CREATE,
154 BPF_SK_SKB_STREAM_PARSER,
155 BPF_SK_SKB_STREAM_VERDICT,
158 BPF_CGROUP_INET4_BIND,
159 BPF_CGROUP_INET6_BIND,
160 BPF_CGROUP_INET4_CONNECT,
161 BPF_CGROUP_INET6_CONNECT,
162 BPF_CGROUP_INET4_POST_BIND,
163 BPF_CGROUP_INET6_POST_BIND,
164 BPF_CGROUP_UDP4_SENDMSG,
165 BPF_CGROUP_UDP6_SENDMSG,
167 __MAX_BPF_ATTACH_TYPE
170 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
172 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
174 * NONE(default): No further bpf programs allowed in the subtree.
176 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
177 * the program in this cgroup yields to sub-cgroup program.
179 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
180 * that cgroup program gets run in addition to the program in this cgroup.
182 * Only one program is allowed to be attached to a cgroup with
183 * NONE or BPF_F_ALLOW_OVERRIDE flag.
184 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
185 * release old program and attach the new one. Attach flags has to match.
187 * Multiple programs are allowed to be attached to a cgroup with
188 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
189 * (those that were attached first, run first)
190 * The programs of sub-cgroup are executed first, then programs of
191 * this cgroup and then programs of parent cgroup.
192 * When children program makes decision (like picking TCP CA or sock bind)
193 * parent program has a chance to override it.
195 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
196 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
198 * cgrp1 (MULTI progs A, B) ->
199 * cgrp2 (OVERRIDE prog C) ->
200 * cgrp3 (MULTI prog D) ->
201 * cgrp4 (OVERRIDE prog E) ->
202 * cgrp5 (NONE prog F)
203 * the event in cgrp5 triggers execution of F,D,A,B in that order.
204 * if prog F is detached, the execution is E,D,A,B
205 * if prog F and D are detached, the execution is E,A,B
206 * if prog F, E and D are detached, the execution is C,A,B
208 * All eligible programs are executed regardless of return code from
211 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
212 #define BPF_F_ALLOW_MULTI (1U << 1)
214 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
215 * verifier will perform strict alignment checking as if the kernel
216 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
217 * and NET_IP_ALIGN defined to 2.
219 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
221 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
222 #define BPF_PSEUDO_MAP_FD 1
224 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
225 * offset to another bpf function
227 #define BPF_PSEUDO_CALL 1
229 /* flags for BPF_MAP_UPDATE_ELEM command */
230 #define BPF_ANY 0 /* create new element or update existing */
231 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
232 #define BPF_EXIST 2 /* update existing element */
234 /* flags for BPF_MAP_CREATE command */
235 #define BPF_F_NO_PREALLOC (1U << 0)
236 /* Instead of having one common LRU list in the
237 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
238 * which can scale and perform better.
239 * Note, the LRU nodes (including free nodes) cannot be moved
240 * across different LRU lists.
242 #define BPF_F_NO_COMMON_LRU (1U << 1)
243 /* Specify numa node during map creation */
244 #define BPF_F_NUMA_NODE (1U << 2)
246 /* flags for BPF_PROG_QUERY */
247 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
249 #define BPF_OBJ_NAME_LEN 16U
251 /* Flags for accessing BPF object */
252 #define BPF_F_RDONLY (1U << 3)
253 #define BPF_F_WRONLY (1U << 4)
255 /* Flag for stack_map, store build_id+offset instead of pointer */
256 #define BPF_F_STACK_BUILD_ID (1U << 5)
258 enum bpf_stack_build_id_status {
259 /* user space need an empty entry to identify end of a trace */
260 BPF_STACK_BUILD_ID_EMPTY = 0,
261 /* with valid build_id and offset */
262 BPF_STACK_BUILD_ID_VALID = 1,
263 /* couldn't get build_id, fallback to ip */
264 BPF_STACK_BUILD_ID_IP = 2,
267 #define BPF_BUILD_ID_SIZE 20
268 struct bpf_stack_build_id {
270 unsigned char build_id[BPF_BUILD_ID_SIZE];
278 struct { /* anonymous struct used by BPF_MAP_CREATE command */
279 __u32 map_type; /* one of enum bpf_map_type */
280 __u32 key_size; /* size of key in bytes */
281 __u32 value_size; /* size of value in bytes */
282 __u32 max_entries; /* max number of entries in a map */
283 __u32 map_flags; /* BPF_MAP_CREATE related
284 * flags defined above.
286 __u32 inner_map_fd; /* fd pointing to the inner map */
287 __u32 numa_node; /* numa node (effective only if
288 * BPF_F_NUMA_NODE is set).
290 char map_name[BPF_OBJ_NAME_LEN];
291 __u32 map_ifindex; /* ifindex of netdev to create on */
292 __u32 btf_fd; /* fd pointing to a BTF type data */
293 __u32 btf_key_type_id; /* BTF type_id of the key */
294 __u32 btf_value_type_id; /* BTF type_id of the value */
297 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
302 __aligned_u64 next_key;
307 struct { /* anonymous struct used by BPF_PROG_LOAD command */
308 __u32 prog_type; /* one of enum bpf_prog_type */
311 __aligned_u64 license;
312 __u32 log_level; /* verbosity level of verifier */
313 __u32 log_size; /* size of user buffer */
314 __aligned_u64 log_buf; /* user supplied buffer */
315 __u32 kern_version; /* checked when prog_type=kprobe */
317 char prog_name[BPF_OBJ_NAME_LEN];
318 __u32 prog_ifindex; /* ifindex of netdev to prep for */
319 /* For some prog types expected attach type must be known at
320 * load time to verify attach type specific parts of prog
321 * (context accesses, allowed helpers, etc).
323 __u32 expected_attach_type;
326 struct { /* anonymous struct used by BPF_OBJ_* commands */
327 __aligned_u64 pathname;
332 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
333 __u32 target_fd; /* container object to attach to */
334 __u32 attach_bpf_fd; /* eBPF program to attach */
339 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
344 __aligned_u64 data_in;
345 __aligned_u64 data_out;
350 struct { /* anonymous struct used by BPF_*_GET_*_ID */
361 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
367 struct { /* anonymous struct used by BPF_PROG_QUERY command */
368 __u32 target_fd; /* container object to query */
372 __aligned_u64 prog_ids;
381 struct { /* anonymous struct for BPF_BTF_LOAD */
383 __aligned_u64 btf_log_buf;
390 __u32 pid; /* input: pid */
391 __u32 fd; /* input: fd */
392 __u32 flags; /* input: flags */
393 __u32 buf_len; /* input/output: buf len */
394 __aligned_u64 buf; /* input/output:
395 * tp_name for tracepoint
397 * filename for uprobe
399 __u32 prog_id; /* output: prod_id */
400 __u32 fd_type; /* output: BPF_FD_TYPE_* */
401 __u64 probe_offset; /* output: probe_offset */
402 __u64 probe_addr; /* output: probe_addr */
404 } __attribute__((aligned(8)));
406 /* The description below is an attempt at providing documentation to eBPF
407 * developers about the multiple available eBPF helper functions. It can be
408 * parsed and used to produce a manual page. The workflow is the following,
409 * and requires the rst2man utility:
411 * $ ./scripts/bpf_helpers_doc.py \
412 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
413 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
414 * $ man /tmp/bpf-helpers.7
416 * Note that in order to produce this external documentation, some RST
417 * formatting is used in the descriptions to get "bold" and "italics" in
418 * manual pages. Also note that the few trailing white spaces are
419 * intentional, removing them would break paragraphs for rst2man.
421 * Start of BPF helper function descriptions:
423 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
425 * Perform a lookup in *map* for an entry associated to *key*.
427 * Map value associated to *key*, or **NULL** if no entry was
430 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
432 * Add or update the value of the entry associated to *key* in
433 * *map* with *value*. *flags* is one of:
436 * The entry for *key* must not exist in the map.
438 * The entry for *key* must already exist in the map.
440 * No condition on the existence of the entry for *key*.
442 * Flag value **BPF_NOEXIST** cannot be used for maps of types
443 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
444 * elements always exist), the helper would return an error.
446 * 0 on success, or a negative error in case of failure.
448 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
450 * Delete entry with *key* from *map*.
452 * 0 on success, or a negative error in case of failure.
454 * int bpf_probe_read(void *dst, u32 size, const void *src)
456 * For tracing programs, safely attempt to read *size* bytes from
457 * address *src* and store the data in *dst*.
459 * 0 on success, or a negative error in case of failure.
461 * u64 bpf_ktime_get_ns(void)
463 * Return the time elapsed since system boot, in nanoseconds.
467 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
469 * This helper is a "printk()-like" facility for debugging. It
470 * prints a message defined by format *fmt* (of size *fmt_size*)
471 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
472 * available. It can take up to three additional **u64**
473 * arguments (as an eBPF helpers, the total number of arguments is
476 * Each time the helper is called, it appends a line to the trace.
477 * The format of the trace is customizable, and the exact output
478 * one will get depends on the options set in
479 * *\/sys/kernel/debug/tracing/trace_options* (see also the
480 * *README* file under the same directory). However, it usually
481 * defaults to something like:
485 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
489 * * ``telnet`` is the name of the current task.
490 * * ``470`` is the PID of the current task.
491 * * ``001`` is the CPU number on which the task is
493 * * In ``.N..``, each character refers to a set of
494 * options (whether irqs are enabled, scheduling
495 * options, whether hard/softirqs are running, level of
496 * preempt_disabled respectively). **N** means that
497 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
499 * * ``419421.045894`` is a timestamp.
500 * * ``0x00000001`` is a fake value used by BPF for the
501 * instruction pointer register.
502 * * ``<formatted msg>`` is the message formatted with
505 * The conversion specifiers supported by *fmt* are similar, but
506 * more limited than for printk(). They are **%d**, **%i**,
507 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
508 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
509 * of field, padding with zeroes, etc.) is available, and the
510 * helper will return **-EINVAL** (but print nothing) if it
511 * encounters an unknown specifier.
513 * Also, note that **bpf_trace_printk**\ () is slow, and should
514 * only be used for debugging purposes. For this reason, a notice
515 * bloc (spanning several lines) is printed to kernel logs and
516 * states that the helper should not be used "for production use"
517 * the first time this helper is used (or more precisely, when
518 * **trace_printk**\ () buffers are allocated). For passing values
519 * to user space, perf events should be preferred.
521 * The number of bytes written to the buffer, or a negative error
522 * in case of failure.
524 * u32 bpf_get_prandom_u32(void)
526 * Get a pseudo-random number.
528 * From a security point of view, this helper uses its own
529 * pseudo-random internal state, and cannot be used to infer the
530 * seed of other random functions in the kernel. However, it is
531 * essential to note that the generator used by the helper is not
532 * cryptographically secure.
534 * A random 32-bit unsigned value.
536 * u32 bpf_get_smp_processor_id(void)
538 * Get the SMP (symmetric multiprocessing) processor id. Note that
539 * all programs run with preemption disabled, which means that the
540 * SMP processor id is stable during all the execution of the
543 * The SMP id of the processor running the program.
545 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
547 * Store *len* bytes from address *from* into the packet
548 * associated to *skb*, at *offset*. *flags* are a combination of
549 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
550 * checksum for the packet after storing the bytes) and
551 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
552 * **->swhash** and *skb*\ **->l4hash** to 0).
554 * A call to this helper is susceptible to change the underlaying
555 * packet buffer. Therefore, at load time, all checks on pointers
556 * previously done by the verifier are invalidated and must be
557 * performed again, if the helper is used in combination with
558 * direct packet access.
560 * 0 on success, or a negative error in case of failure.
562 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
564 * Recompute the layer 3 (e.g. IP) checksum for the packet
565 * associated to *skb*. Computation is incremental, so the helper
566 * must know the former value of the header field that was
567 * modified (*from*), the new value of this field (*to*), and the
568 * number of bytes (2 or 4) for this field, stored in *size*.
569 * Alternatively, it is possible to store the difference between
570 * the previous and the new values of the header field in *to*, by
571 * setting *from* and *size* to 0. For both methods, *offset*
572 * indicates the location of the IP checksum within the packet.
574 * This helper works in combination with **bpf_csum_diff**\ (),
575 * which does not update the checksum in-place, but offers more
576 * flexibility and can handle sizes larger than 2 or 4 for the
577 * checksum to update.
579 * A call to this helper is susceptible to change the underlaying
580 * packet buffer. Therefore, at load time, all checks on pointers
581 * previously done by the verifier are invalidated and must be
582 * performed again, if the helper is used in combination with
583 * direct packet access.
585 * 0 on success, or a negative error in case of failure.
587 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
589 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
590 * packet associated to *skb*. Computation is incremental, so the
591 * helper must know the former value of the header field that was
592 * modified (*from*), the new value of this field (*to*), and the
593 * number of bytes (2 or 4) for this field, stored on the lowest
594 * four bits of *flags*. Alternatively, it is possible to store
595 * the difference between the previous and the new values of the
596 * header field in *to*, by setting *from* and the four lowest
597 * bits of *flags* to 0. For both methods, *offset* indicates the
598 * location of the IP checksum within the packet. In addition to
599 * the size of the field, *flags* can be added (bitwise OR) actual
600 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
601 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
602 * for updates resulting in a null checksum the value is set to
603 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
604 * the checksum is to be computed against a pseudo-header.
606 * This helper works in combination with **bpf_csum_diff**\ (),
607 * which does not update the checksum in-place, but offers more
608 * flexibility and can handle sizes larger than 2 or 4 for the
609 * checksum to update.
611 * A call to this helper is susceptible to change the underlaying
612 * packet buffer. Therefore, at load time, all checks on pointers
613 * previously done by the verifier are invalidated and must be
614 * performed again, if the helper is used in combination with
615 * direct packet access.
617 * 0 on success, or a negative error in case of failure.
619 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
621 * This special helper is used to trigger a "tail call", or in
622 * other words, to jump into another eBPF program. The same stack
623 * frame is used (but values on stack and in registers for the
624 * caller are not accessible to the callee). This mechanism allows
625 * for program chaining, either for raising the maximum number of
626 * available eBPF instructions, or to execute given programs in
627 * conditional blocks. For security reasons, there is an upper
628 * limit to the number of successive tail calls that can be
631 * Upon call of this helper, the program attempts to jump into a
632 * program referenced at index *index* in *prog_array_map*, a
633 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
634 * *ctx*, a pointer to the context.
636 * If the call succeeds, the kernel immediately runs the first
637 * instruction of the new program. This is not a function call,
638 * and it never returns to the previous program. If the call
639 * fails, then the helper has no effect, and the caller continues
640 * to run its subsequent instructions. A call can fail if the
641 * destination program for the jump does not exist (i.e. *index*
642 * is superior to the number of entries in *prog_array_map*), or
643 * if the maximum number of tail calls has been reached for this
644 * chain of programs. This limit is defined in the kernel by the
645 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
646 * which is currently set to 32.
648 * 0 on success, or a negative error in case of failure.
650 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
652 * Clone and redirect the packet associated to *skb* to another
653 * net device of index *ifindex*. Both ingress and egress
654 * interfaces can be used for redirection. The **BPF_F_INGRESS**
655 * value in *flags* is used to make the distinction (ingress path
656 * is selected if the flag is present, egress path otherwise).
657 * This is the only flag supported for now.
659 * In comparison with **bpf_redirect**\ () helper,
660 * **bpf_clone_redirect**\ () has the associated cost of
661 * duplicating the packet buffer, but this can be executed out of
662 * the eBPF program. Conversely, **bpf_redirect**\ () is more
663 * efficient, but it is handled through an action code where the
664 * redirection happens only after the eBPF program has returned.
666 * A call to this helper is susceptible to change the underlaying
667 * packet buffer. Therefore, at load time, all checks on pointers
668 * previously done by the verifier are invalidated and must be
669 * performed again, if the helper is used in combination with
670 * direct packet access.
672 * 0 on success, or a negative error in case of failure.
674 * u64 bpf_get_current_pid_tgid(void)
676 * A 64-bit integer containing the current tgid and pid, and
678 * *current_task*\ **->tgid << 32 \|**
679 * *current_task*\ **->pid**.
681 * u64 bpf_get_current_uid_gid(void)
683 * A 64-bit integer containing the current GID and UID, and
684 * created as such: *current_gid* **<< 32 \|** *current_uid*.
686 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
688 * Copy the **comm** attribute of the current task into *buf* of
689 * *size_of_buf*. The **comm** attribute contains the name of
690 * the executable (excluding the path) for the current task. The
691 * *size_of_buf* must be strictly positive. On success, the
692 * helper makes sure that the *buf* is NUL-terminated. On failure,
693 * it is filled with zeroes.
695 * 0 on success, or a negative error in case of failure.
697 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
699 * Retrieve the classid for the current task, i.e. for the net_cls
700 * cgroup to which *skb* belongs.
702 * This helper can be used on TC egress path, but not on ingress.
704 * The net_cls cgroup provides an interface to tag network packets
705 * based on a user-provided identifier for all traffic coming from
706 * the tasks belonging to the related cgroup. See also the related
707 * kernel documentation, available from the Linux sources in file
708 * *Documentation/cgroup-v1/net_cls.txt*.
710 * The Linux kernel has two versions for cgroups: there are
711 * cgroups v1 and cgroups v2. Both are available to users, who can
712 * use a mixture of them, but note that the net_cls cgroup is for
713 * cgroup v1 only. This makes it incompatible with BPF programs
714 * run on cgroups, which is a cgroup-v2-only feature (a socket can
715 * only hold data for one version of cgroups at a time).
717 * This helper is only available is the kernel was compiled with
718 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
719 * "**y**" or to "**m**".
721 * The classid, or 0 for the default unconfigured classid.
723 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
725 * Push a *vlan_tci* (VLAN tag control information) of protocol
726 * *vlan_proto* to the packet associated to *skb*, then update
727 * the checksum. Note that if *vlan_proto* is different from
728 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
729 * be **ETH_P_8021Q**.
731 * A call to this helper is susceptible to change the underlaying
732 * packet buffer. Therefore, at load time, all checks on pointers
733 * previously done by the verifier are invalidated and must be
734 * performed again, if the helper is used in combination with
735 * direct packet access.
737 * 0 on success, or a negative error in case of failure.
739 * int bpf_skb_vlan_pop(struct sk_buff *skb)
741 * Pop a VLAN header from the packet associated to *skb*.
743 * A call to this helper is susceptible to change the underlaying
744 * packet buffer. Therefore, at load time, all checks on pointers
745 * previously done by the verifier are invalidated and must be
746 * performed again, if the helper is used in combination with
747 * direct packet access.
749 * 0 on success, or a negative error in case of failure.
751 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
753 * Get tunnel metadata. This helper takes a pointer *key* to an
754 * empty **struct bpf_tunnel_key** of **size**, that will be
755 * filled with tunnel metadata for the packet associated to *skb*.
756 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
757 * indicates that the tunnel is based on IPv6 protocol instead of
760 * The **struct bpf_tunnel_key** is an object that generalizes the
761 * principal parameters used by various tunneling protocols into a
762 * single struct. This way, it can be used to easily make a
763 * decision based on the contents of the encapsulation header,
764 * "summarized" in this struct. In particular, it holds the IP
765 * address of the remote end (IPv4 or IPv6, depending on the case)
766 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
767 * this struct exposes the *key*\ **->tunnel_id**, which is
768 * generally mapped to a VNI (Virtual Network Identifier), making
769 * it programmable together with the **bpf_skb_set_tunnel_key**\
772 * Let's imagine that the following code is part of a program
773 * attached to the TC ingress interface, on one end of a GRE
774 * tunnel, and is supposed to filter out all messages coming from
775 * remote ends with IPv4 address other than 10.0.0.1:
780 * struct bpf_tunnel_key key = {};
782 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
784 * return TC_ACT_SHOT; // drop packet
786 * if (key.remote_ipv4 != 0x0a000001)
787 * return TC_ACT_SHOT; // drop packet
789 * return TC_ACT_OK; // accept packet
791 * This interface can also be used with all encapsulation devices
792 * that can operate in "collect metadata" mode: instead of having
793 * one network device per specific configuration, the "collect
794 * metadata" mode only requires a single device where the
795 * configuration can be extracted from this helper.
797 * This can be used together with various tunnels such as VXLan,
798 * Geneve, GRE or IP in IP (IPIP).
800 * 0 on success, or a negative error in case of failure.
802 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
804 * Populate tunnel metadata for packet associated to *skb.* The
805 * tunnel metadata is set to the contents of *key*, of *size*. The
806 * *flags* can be set to a combination of the following values:
808 * **BPF_F_TUNINFO_IPV6**
809 * Indicate that the tunnel is based on IPv6 protocol
811 * **BPF_F_ZERO_CSUM_TX**
812 * For IPv4 packets, add a flag to tunnel metadata
813 * indicating that checksum computation should be skipped
814 * and checksum set to zeroes.
815 * **BPF_F_DONT_FRAGMENT**
816 * Add a flag to tunnel metadata indicating that the
817 * packet should not be fragmented.
818 * **BPF_F_SEQ_NUMBER**
819 * Add a flag to tunnel metadata indicating that a
820 * sequence number should be added to tunnel header before
821 * sending the packet. This flag was added for GRE
822 * encapsulation, but might be used with other protocols
823 * as well in the future.
825 * Here is a typical usage on the transmit path:
829 * struct bpf_tunnel_key key;
831 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
832 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
834 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
835 * helper for additional information.
837 * 0 on success, or a negative error in case of failure.
839 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
841 * Read the value of a perf event counter. This helper relies on a
842 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
843 * the perf event counter is selected when *map* is updated with
844 * perf event file descriptors. The *map* is an array whose size
845 * is the number of available CPUs, and each cell contains a value
846 * relative to one CPU. The value to retrieve is indicated by
847 * *flags*, that contains the index of the CPU to look up, masked
848 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
849 * **BPF_F_CURRENT_CPU** to indicate that the value for the
850 * current CPU should be retrieved.
852 * Note that before Linux 4.13, only hardware perf event can be
855 * Also, be aware that the newer helper
856 * **bpf_perf_event_read_value**\ () is recommended over
857 * **bpf_perf_event_read**\ () in general. The latter has some ABI
858 * quirks where error and counter value are used as a return code
859 * (which is wrong to do since ranges may overlap). This issue is
860 * fixed with **bpf_perf_event_read_value**\ (), which at the same
861 * time provides more features over the **bpf_perf_event_read**\
862 * () interface. Please refer to the description of
863 * **bpf_perf_event_read_value**\ () for details.
865 * The value of the perf event counter read from the map, or a
866 * negative error code in case of failure.
868 * int bpf_redirect(u32 ifindex, u64 flags)
870 * Redirect the packet to another net device of index *ifindex*.
871 * This helper is somewhat similar to **bpf_clone_redirect**\
872 * (), except that the packet is not cloned, which provides
873 * increased performance.
875 * Except for XDP, both ingress and egress interfaces can be used
876 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
877 * to make the distinction (ingress path is selected if the flag
878 * is present, egress path otherwise). Currently, XDP only
879 * supports redirection to the egress interface, and accepts no
882 * The same effect can be attained with the more generic
883 * **bpf_redirect_map**\ (), which requires specific maps to be
884 * used but offers better performance.
886 * For XDP, the helper returns **XDP_REDIRECT** on success or
887 * **XDP_ABORTED** on error. For other program types, the values
888 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
891 * u32 bpf_get_route_realm(struct sk_buff *skb)
893 * Retrieve the realm or the route, that is to say the
894 * **tclassid** field of the destination for the *skb*. The
895 * indentifier retrieved is a user-provided tag, similar to the
896 * one used with the net_cls cgroup (see description for
897 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
898 * held by a route (a destination entry), not by a task.
900 * Retrieving this identifier works with the clsact TC egress hook
901 * (see also **tc-bpf(8)**), or alternatively on conventional
902 * classful egress qdiscs, but not on TC ingress path. In case of
903 * clsact TC egress hook, this has the advantage that, internally,
904 * the destination entry has not been dropped yet in the transmit
905 * path. Therefore, the destination entry does not need to be
906 * artificially held via **netif_keep_dst**\ () for a classful
907 * qdisc until the *skb* is freed.
909 * This helper is available only if the kernel was compiled with
910 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
912 * The realm of the route for the packet associated to *skb*, or 0
915 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
917 * Write raw *data* blob into a special BPF perf event held by
918 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
919 * event must have the following attributes: **PERF_SAMPLE_RAW**
920 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
921 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
923 * The *flags* are used to indicate the index in *map* for which
924 * the value must be put, masked with **BPF_F_INDEX_MASK**.
925 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
926 * to indicate that the index of the current CPU core should be
929 * The value to write, of *size*, is passed through eBPF stack and
932 * The context of the program *ctx* needs also be passed to the
935 * On user space, a program willing to read the values needs to
936 * call **perf_event_open**\ () on the perf event (either for
937 * one or for all CPUs) and to store the file descriptor into the
938 * *map*. This must be done before the eBPF program can send data
939 * into it. An example is available in file
940 * *samples/bpf/trace_output_user.c* in the Linux kernel source
941 * tree (the eBPF program counterpart is in
942 * *samples/bpf/trace_output_kern.c*).
944 * **bpf_perf_event_output**\ () achieves better performance
945 * than **bpf_trace_printk**\ () for sharing data with user
946 * space, and is much better suitable for streaming data from eBPF
949 * Note that this helper is not restricted to tracing use cases
950 * and can be used with programs attached to TC or XDP as well,
951 * where it allows for passing data to user space listeners. Data
954 * * Only custom structs,
955 * * Only the packet payload, or
956 * * A combination of both.
958 * 0 on success, or a negative error in case of failure.
960 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
962 * This helper was provided as an easy way to load data from a
963 * packet. It can be used to load *len* bytes from *offset* from
964 * the packet associated to *skb*, into the buffer pointed by
967 * Since Linux 4.7, usage of this helper has mostly been replaced
968 * by "direct packet access", enabling packet data to be
969 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
970 * pointing respectively to the first byte of packet data and to
971 * the byte after the last byte of packet data. However, it
972 * remains useful if one wishes to read large quantities of data
973 * at once from a packet into the eBPF stack.
975 * 0 on success, or a negative error in case of failure.
977 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
979 * Walk a user or a kernel stack and return its id. To achieve
980 * this, the helper needs *ctx*, which is a pointer to the context
981 * on which the tracing program is executed, and a pointer to a
982 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
984 * The last argument, *flags*, holds the number of stack frames to
985 * skip (from 0 to 255), masked with
986 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
987 * a combination of the following flags:
989 * **BPF_F_USER_STACK**
990 * Collect a user space stack instead of a kernel stack.
991 * **BPF_F_FAST_STACK_CMP**
992 * Compare stacks by hash only.
993 * **BPF_F_REUSE_STACKID**
994 * If two different stacks hash into the same *stackid*,
995 * discard the old one.
997 * The stack id retrieved is a 32 bit long integer handle which
998 * can be further combined with other data (including other stack
999 * ids) and used as a key into maps. This can be useful for
1000 * generating a variety of graphs (such as flame graphs or off-cpu
1003 * For walking a stack, this helper is an improvement over
1004 * **bpf_probe_read**\ (), which can be used with unrolled loops
1005 * but is not efficient and consumes a lot of eBPF instructions.
1006 * Instead, **bpf_get_stackid**\ () can collect up to
1007 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1008 * this limit can be controlled with the **sysctl** program, and
1009 * that it should be manually increased in order to profile long
1010 * user stacks (such as stacks for Java programs). To do so, use:
1014 * # sysctl kernel.perf_event_max_stack=<new value>
1016 * The positive or null stack id on success, or a negative error
1017 * in case of failure.
1019 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1021 * Compute a checksum difference, from the raw buffer pointed by
1022 * *from*, of length *from_size* (that must be a multiple of 4),
1023 * towards the raw buffer pointed by *to*, of size *to_size*
1024 * (same remark). An optional *seed* can be added to the value
1025 * (this can be cascaded, the seed may come from a previous call
1028 * This is flexible enough to be used in several ways:
1030 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1031 * checksum, it can be used when pushing new data.
1032 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1033 * checksum, it can be used when removing data from a packet.
1034 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1035 * can be used to compute a diff. Note that *from_size* and
1036 * *to_size* do not need to be equal.
1038 * This helper can be used in combination with
1039 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1040 * which one can feed in the difference computed with
1041 * **bpf_csum_diff**\ ().
1043 * The checksum result, or a negative error code in case of
1046 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1048 * Retrieve tunnel options metadata for the packet associated to
1049 * *skb*, and store the raw tunnel option data to the buffer *opt*
1052 * This helper can be used with encapsulation devices that can
1053 * operate in "collect metadata" mode (please refer to the related
1054 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1055 * more details). A particular example where this can be used is
1056 * in combination with the Geneve encapsulation protocol, where it
1057 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1058 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1059 * the eBPF program. This allows for full customization of these
1062 * The size of the option data retrieved.
1064 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1066 * Set tunnel options metadata for the packet associated to *skb*
1067 * to the option data contained in the raw buffer *opt* of *size*.
1069 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1070 * helper for additional information.
1072 * 0 on success, or a negative error in case of failure.
1074 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1076 * Change the protocol of the *skb* to *proto*. Currently
1077 * supported are transition from IPv4 to IPv6, and from IPv6 to
1078 * IPv4. The helper takes care of the groundwork for the
1079 * transition, including resizing the socket buffer. The eBPF
1080 * program is expected to fill the new headers, if any, via
1081 * **skb_store_bytes**\ () and to recompute the checksums with
1082 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1083 * (). The main case for this helper is to perform NAT64
1084 * operations out of an eBPF program.
1086 * Internally, the GSO type is marked as dodgy so that headers are
1087 * checked and segments are recalculated by the GSO/GRO engine.
1088 * The size for GSO target is adapted as well.
1090 * All values for *flags* are reserved for future usage, and must
1093 * A call to this helper is susceptible to change the underlaying
1094 * packet buffer. Therefore, at load time, all checks on pointers
1095 * previously done by the verifier are invalidated and must be
1096 * performed again, if the helper is used in combination with
1097 * direct packet access.
1099 * 0 on success, or a negative error in case of failure.
1101 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1103 * Change the packet type for the packet associated to *skb*. This
1104 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1105 * the eBPF program does not have a write access to *skb*\
1106 * **->pkt_type** beside this helper. Using a helper here allows
1107 * for graceful handling of errors.
1109 * The major use case is to change incoming *skb*s to
1110 * **PACKET_HOST** in a programmatic way instead of having to
1111 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1114 * Note that *type* only allows certain values. At this time, they
1119 * **PACKET_BROADCAST**
1120 * Send packet to all.
1121 * **PACKET_MULTICAST**
1122 * Send packet to group.
1123 * **PACKET_OTHERHOST**
1124 * Send packet to someone else.
1126 * 0 on success, or a negative error in case of failure.
1128 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1130 * Check whether *skb* is a descendant of the cgroup2 held by
1131 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1133 * The return value depends on the result of the test, and can be:
1135 * * 0, if the *skb* failed the cgroup2 descendant test.
1136 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1137 * * A negative error code, if an error occurred.
1139 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1141 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1142 * not set, in particular if the hash was cleared due to mangling,
1143 * recompute this hash. Later accesses to the hash can be done
1144 * directly with *skb*\ **->hash**.
1146 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1147 * prototype with **bpf_skb_change_proto**\ (), or calling
1148 * **bpf_skb_store_bytes**\ () with the
1149 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1150 * the hash and to trigger a new computation for the next call to
1151 * **bpf_get_hash_recalc**\ ().
1155 * u64 bpf_get_current_task(void)
1157 * A pointer to the current task struct.
1159 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1161 * Attempt in a safe way to write *len* bytes from the buffer
1162 * *src* to *dst* in memory. It only works for threads that are in
1163 * user context, and *dst* must be a valid user space address.
1165 * This helper should not be used to implement any kind of
1166 * security mechanism because of TOC-TOU attacks, but rather to
1167 * debug, divert, and manipulate execution of semi-cooperative
1170 * Keep in mind that this feature is meant for experiments, and it
1171 * has a risk of crashing the system and running programs.
1172 * Therefore, when an eBPF program using this helper is attached,
1173 * a warning including PID and process name is printed to kernel
1176 * 0 on success, or a negative error in case of failure.
1178 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1180 * Check whether the probe is being run is the context of a given
1181 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1182 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1184 * The return value depends on the result of the test, and can be:
1186 * * 0, if the *skb* task belongs to the cgroup2.
1187 * * 1, if the *skb* task does not belong to the cgroup2.
1188 * * A negative error code, if an error occurred.
1190 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1192 * Resize (trim or grow) the packet associated to *skb* to the
1193 * new *len*. The *flags* are reserved for future usage, and must
1196 * The basic idea is that the helper performs the needed work to
1197 * change the size of the packet, then the eBPF program rewrites
1198 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1199 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1200 * and others. This helper is a slow path utility intended for
1201 * replies with control messages. And because it is targeted for
1202 * slow path, the helper itself can afford to be slow: it
1203 * implicitly linearizes, unclones and drops offloads from the
1206 * A call to this helper is susceptible to change the underlaying
1207 * packet buffer. Therefore, at load time, all checks on pointers
1208 * previously done by the verifier are invalidated and must be
1209 * performed again, if the helper is used in combination with
1210 * direct packet access.
1212 * 0 on success, or a negative error in case of failure.
1214 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1216 * Pull in non-linear data in case the *skb* is non-linear and not
1217 * all of *len* are part of the linear section. Make *len* bytes
1218 * from *skb* readable and writable. If a zero value is passed for
1219 * *len*, then the whole length of the *skb* is pulled.
1221 * This helper is only needed for reading and writing with direct
1224 * For direct packet access, testing that offsets to access
1225 * are within packet boundaries (test on *skb*\ **->data_end**) is
1226 * susceptible to fail if offsets are invalid, or if the requested
1227 * data is in non-linear parts of the *skb*. On failure the
1228 * program can just bail out, or in the case of a non-linear
1229 * buffer, use a helper to make the data available. The
1230 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1231 * the data. Another one consists in using **bpf_skb_pull_data**
1232 * to pull in once the non-linear parts, then retesting and
1233 * eventually access the data.
1235 * At the same time, this also makes sure the *skb* is uncloned,
1236 * which is a necessary condition for direct write. As this needs
1237 * to be an invariant for the write part only, the verifier
1238 * detects writes and adds a prologue that is calling
1239 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1240 * the very beginning in case it is indeed cloned.
1242 * A call to this helper is susceptible to change the underlaying
1243 * packet buffer. Therefore, at load time, all checks on pointers
1244 * previously done by the verifier are invalidated and must be
1245 * performed again, if the helper is used in combination with
1246 * direct packet access.
1248 * 0 on success, or a negative error in case of failure.
1250 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1252 * Add the checksum *csum* into *skb*\ **->csum** in case the
1253 * driver has supplied a checksum for the entire packet into that
1254 * field. Return an error otherwise. This helper is intended to be
1255 * used in combination with **bpf_csum_diff**\ (), in particular
1256 * when the checksum needs to be updated after data has been
1257 * written into the packet through direct packet access.
1259 * The checksum on success, or a negative error code in case of
1262 * void bpf_set_hash_invalid(struct sk_buff *skb)
1264 * Invalidate the current *skb*\ **->hash**. It can be used after
1265 * mangling on headers through direct packet access, in order to
1266 * indicate that the hash is outdated and to trigger a
1267 * recalculation the next time the kernel tries to access this
1268 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1270 * int bpf_get_numa_node_id(void)
1272 * Return the id of the current NUMA node. The primary use case
1273 * for this helper is the selection of sockets for the local NUMA
1274 * node, when the program is attached to sockets using the
1275 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1276 * but the helper is also available to other eBPF program types,
1277 * similarly to **bpf_get_smp_processor_id**\ ().
1279 * The id of current NUMA node.
1281 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1283 * Grows headroom of packet associated to *skb* and adjusts the
1284 * offset of the MAC header accordingly, adding *len* bytes of
1285 * space. It automatically extends and reallocates memory as
1288 * This helper can be used on a layer 3 *skb* to push a MAC header
1289 * for redirection into a layer 2 device.
1291 * All values for *flags* are reserved for future usage, and must
1294 * A call to this helper is susceptible to change the underlaying
1295 * packet buffer. Therefore, at load time, all checks on pointers
1296 * previously done by the verifier are invalidated and must be
1297 * performed again, if the helper is used in combination with
1298 * direct packet access.
1300 * 0 on success, or a negative error in case of failure.
1302 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1304 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1305 * it is possible to use a negative value for *delta*. This helper
1306 * can be used to prepare the packet for pushing or popping
1309 * A call to this helper is susceptible to change the underlaying
1310 * packet buffer. Therefore, at load time, all checks on pointers
1311 * previously done by the verifier are invalidated and must be
1312 * performed again, if the helper is used in combination with
1313 * direct packet access.
1315 * 0 on success, or a negative error in case of failure.
1317 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1319 * Copy a NUL terminated string from an unsafe address
1320 * *unsafe_ptr* to *dst*. The *size* should include the
1321 * terminating NUL byte. In case the string length is smaller than
1322 * *size*, the target is not padded with further NUL bytes. If the
1323 * string length is larger than *size*, just *size*-1 bytes are
1324 * copied and the last byte is set to NUL.
1326 * On success, the length of the copied string is returned. This
1327 * makes this helper useful in tracing programs for reading
1328 * strings, and more importantly to get its length at runtime. See
1329 * the following snippet:
1333 * SEC("kprobe/sys_open")
1334 * void bpf_sys_open(struct pt_regs *ctx)
1336 * char buf[PATHLEN]; // PATHLEN is defined to 256
1337 * int res = bpf_probe_read_str(buf, sizeof(buf),
1340 * // Consume buf, for example push it to
1341 * // userspace via bpf_perf_event_output(); we
1342 * // can use res (the string length) as event
1343 * // size, after checking its boundaries.
1346 * In comparison, using **bpf_probe_read()** helper here instead
1347 * to read the string would require to estimate the length at
1348 * compile time, and would often result in copying more memory
1351 * Another useful use case is when parsing individual process
1352 * arguments or individual environment variables navigating
1353 * *current*\ **->mm->arg_start** and *current*\
1354 * **->mm->env_start**: using this helper and the return value,
1355 * one can quickly iterate at the right offset of the memory area.
1357 * On success, the strictly positive length of the string,
1358 * including the trailing NUL character. On error, a negative
1361 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1363 * If the **struct sk_buff** pointed by *skb* has a known socket,
1364 * retrieve the cookie (generated by the kernel) of this socket.
1365 * If no cookie has been set yet, generate a new cookie. Once
1366 * generated, the socket cookie remains stable for the life of the
1367 * socket. This helper can be useful for monitoring per socket
1368 * networking traffic statistics as it provides a unique socket
1369 * identifier per namespace.
1371 * A 8-byte long non-decreasing number on success, or 0 if the
1372 * socket field is missing inside *skb*.
1374 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1376 * The owner UID of the socket associated to *skb*. If the socket
1377 * is **NULL**, or if it is not a full socket (i.e. if it is a
1378 * time-wait or a request socket instead), **overflowuid** value
1379 * is returned (note that **overflowuid** might also be the actual
1380 * UID value for the socket).
1382 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1384 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1389 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1391 * Emulate a call to **setsockopt()** on the socket associated to
1392 * *bpf_socket*, which must be a full socket. The *level* at
1393 * which the option resides and the name *optname* of the option
1394 * must be specified, see **setsockopt(2)** for more information.
1395 * The option value of length *optlen* is pointed by *optval*.
1397 * This helper actually implements a subset of **setsockopt()**.
1398 * It supports the following *level*\ s:
1400 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1401 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1402 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1403 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1404 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1405 * **TCP_BPF_SNDCWND_CLAMP**.
1406 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1407 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1409 * 0 on success, or a negative error in case of failure.
1411 * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags)
1413 * Grow or shrink the room for data in the packet associated to
1414 * *skb* by *len_diff*, and according to the selected *mode*.
1416 * There is a single supported mode at this time:
1418 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1419 * (room space is added or removed below the layer 3 header).
1421 * All values for *flags* are reserved for future usage, and must
1424 * A call to this helper is susceptible to change the underlaying
1425 * packet buffer. Therefore, at load time, all checks on pointers
1426 * previously done by the verifier are invalidated and must be
1427 * performed again, if the helper is used in combination with
1428 * direct packet access.
1430 * 0 on success, or a negative error in case of failure.
1432 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1434 * Redirect the packet to the endpoint referenced by *map* at
1435 * index *key*. Depending on its type, this *map* can contain
1436 * references to net devices (for forwarding packets through other
1437 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1438 * but this is only implemented for native XDP (with driver
1439 * support) as of this writing).
1441 * All values for *flags* are reserved for future usage, and must
1444 * When used to redirect packets to net devices, this helper
1445 * provides a high performance increase over **bpf_redirect**\ ().
1446 * This is due to various implementation details of the underlying
1447 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1448 * () tries to send packet as a "bulk" to the device.
1450 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1452 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1454 * Redirect the packet to the socket referenced by *map* (of type
1455 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1456 * egress interfaces can be used for redirection. The
1457 * **BPF_F_INGRESS** value in *flags* is used to make the
1458 * distinction (ingress path is selected if the flag is present,
1459 * egress path otherwise). This is the only flag supported for now.
1461 * **SK_PASS** on success, or **SK_DROP** on error.
1463 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1465 * Add an entry to, or update a *map* referencing sockets. The
1466 * *skops* is used as a new value for the entry associated to
1467 * *key*. *flags* is one of:
1470 * The entry for *key* must not exist in the map.
1472 * The entry for *key* must already exist in the map.
1474 * No condition on the existence of the entry for *key*.
1476 * If the *map* has eBPF programs (parser and verdict), those will
1477 * be inherited by the socket being added. If the socket is
1478 * already attached to eBPF programs, this results in an error.
1480 * 0 on success, or a negative error in case of failure.
1482 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1484 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1485 * *delta* (which can be positive or negative). Note that this
1486 * operation modifies the address stored in *xdp_md*\ **->data**,
1487 * so the latter must be loaded only after the helper has been
1490 * The use of *xdp_md*\ **->data_meta** is optional and programs
1491 * are not required to use it. The rationale is that when the
1492 * packet is processed with XDP (e.g. as DoS filter), it is
1493 * possible to push further meta data along with it before passing
1494 * to the stack, and to give the guarantee that an ingress eBPF
1495 * program attached as a TC classifier on the same device can pick
1496 * this up for further post-processing. Since TC works with socket
1497 * buffers, it remains possible to set from XDP the **mark** or
1498 * **priority** pointers, or other pointers for the socket buffer.
1499 * Having this scratch space generic and programmable allows for
1500 * more flexibility as the user is free to store whatever meta
1503 * A call to this helper is susceptible to change the underlaying
1504 * packet buffer. Therefore, at load time, all checks on pointers
1505 * previously done by the verifier are invalidated and must be
1506 * performed again, if the helper is used in combination with
1507 * direct packet access.
1509 * 0 on success, or a negative error in case of failure.
1511 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1513 * Read the value of a perf event counter, and store it into *buf*
1514 * of size *buf_size*. This helper relies on a *map* of type
1515 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1516 * counter is selected when *map* is updated with perf event file
1517 * descriptors. The *map* is an array whose size is the number of
1518 * available CPUs, and each cell contains a value relative to one
1519 * CPU. The value to retrieve is indicated by *flags*, that
1520 * contains the index of the CPU to look up, masked with
1521 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1522 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1523 * current CPU should be retrieved.
1525 * This helper behaves in a way close to
1526 * **bpf_perf_event_read**\ () helper, save that instead of
1527 * just returning the value observed, it fills the *buf*
1528 * structure. This allows for additional data to be retrieved: in
1529 * particular, the enabled and running times (in *buf*\
1530 * **->enabled** and *buf*\ **->running**, respectively) are
1531 * copied. In general, **bpf_perf_event_read_value**\ () is
1532 * recommended over **bpf_perf_event_read**\ (), which has some
1533 * ABI issues and provides fewer functionalities.
1535 * These values are interesting, because hardware PMU (Performance
1536 * Monitoring Unit) counters are limited resources. When there are
1537 * more PMU based perf events opened than available counters,
1538 * kernel will multiplex these events so each event gets certain
1539 * percentage (but not all) of the PMU time. In case that
1540 * multiplexing happens, the number of samples or counter value
1541 * will not reflect the case compared to when no multiplexing
1542 * occurs. This makes comparison between different runs difficult.
1543 * Typically, the counter value should be normalized before
1544 * comparing to other experiments. The usual normalization is done
1549 * normalized_counter = counter * t_enabled / t_running
1551 * Where t_enabled is the time enabled for event and t_running is
1552 * the time running for event since last normalization. The
1553 * enabled and running times are accumulated since the perf event
1554 * open. To achieve scaling factor between two invocations of an
1555 * eBPF program, users can can use CPU id as the key (which is
1556 * typical for perf array usage model) to remember the previous
1557 * value and do the calculation inside the eBPF program.
1559 * 0 on success, or a negative error in case of failure.
1561 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1563 * For en eBPF program attached to a perf event, retrieve the
1564 * value of the event counter associated to *ctx* and store it in
1565 * the structure pointed by *buf* and of size *buf_size*. Enabled
1566 * and running times are also stored in the structure (see
1567 * description of helper **bpf_perf_event_read_value**\ () for
1570 * 0 on success, or a negative error in case of failure.
1572 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1574 * Emulate a call to **getsockopt()** on the socket associated to
1575 * *bpf_socket*, which must be a full socket. The *level* at
1576 * which the option resides and the name *optname* of the option
1577 * must be specified, see **getsockopt(2)** for more information.
1578 * The retrieved value is stored in the structure pointed by
1579 * *opval* and of length *optlen*.
1581 * This helper actually implements a subset of **getsockopt()**.
1582 * It supports the following *level*\ s:
1584 * * **IPPROTO_TCP**, which supports *optname*
1585 * **TCP_CONGESTION**.
1586 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1587 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1589 * 0 on success, or a negative error in case of failure.
1591 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1593 * Used for error injection, this helper uses kprobes to override
1594 * the return value of the probed function, and to set it to *rc*.
1595 * The first argument is the context *regs* on which the kprobe
1598 * This helper works by setting setting the PC (program counter)
1599 * to an override function which is run in place of the original
1600 * probed function. This means the probed function is not run at
1601 * all. The replacement function just returns with the required
1604 * This helper has security implications, and thus is subject to
1605 * restrictions. It is only available if the kernel was compiled
1606 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1607 * option, and in this case it only works on functions tagged with
1608 * **ALLOW_ERROR_INJECTION** in the kernel code.
1610 * Also, the helper is only available for the architectures having
1611 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1612 * x86 architecture is the only one to support this feature.
1616 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1618 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1619 * for the full TCP socket associated to *bpf_sock_ops* to
1622 * The primary use of this field is to determine if there should
1623 * be calls to eBPF programs of type
1624 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1625 * code. A program of the same type can change its value, per
1626 * connection and as necessary, when the connection is
1627 * established. This field is directly accessible for reading, but
1628 * this helper must be used for updates in order to return an
1629 * error if an eBPF program tries to set a callback that is not
1630 * supported in the current kernel.
1632 * The supported callback values that *argval* can combine are:
1634 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1635 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1636 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1638 * Here are some examples of where one could call such eBPF
1642 * * When a packet is retransmitted.
1643 * * When the connection terminates.
1644 * * When a packet is sent.
1645 * * When a packet is received.
1647 * Code **-EINVAL** if the socket is not a full TCP socket;
1648 * otherwise, a positive number containing the bits that could not
1649 * be set is returned (which comes down to 0 if all bits were set
1652 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1654 * This helper is used in programs implementing policies at the
1655 * socket level. If the message *msg* is allowed to pass (i.e. if
1656 * the verdict eBPF program returns **SK_PASS**), redirect it to
1657 * the socket referenced by *map* (of type
1658 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1659 * egress interfaces can be used for redirection. The
1660 * **BPF_F_INGRESS** value in *flags* is used to make the
1661 * distinction (ingress path is selected if the flag is present,
1662 * egress path otherwise). This is the only flag supported for now.
1664 * **SK_PASS** on success, or **SK_DROP** on error.
1666 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1668 * For socket policies, apply the verdict of the eBPF program to
1669 * the next *bytes* (number of bytes) of message *msg*.
1671 * For example, this helper can be used in the following cases:
1673 * * A single **sendmsg**\ () or **sendfile**\ () system call
1674 * contains multiple logical messages that the eBPF program is
1675 * supposed to read and for which it should apply a verdict.
1676 * * An eBPF program only cares to read the first *bytes* of a
1677 * *msg*. If the message has a large payload, then setting up
1678 * and calling the eBPF program repeatedly for all bytes, even
1679 * though the verdict is already known, would create unnecessary
1682 * When called from within an eBPF program, the helper sets a
1683 * counter internal to the BPF infrastructure, that is used to
1684 * apply the last verdict to the next *bytes*. If *bytes* is
1685 * smaller than the current data being processed from a
1686 * **sendmsg**\ () or **sendfile**\ () system call, the first
1687 * *bytes* will be sent and the eBPF program will be re-run with
1688 * the pointer for start of data pointing to byte number *bytes*
1689 * **+ 1**. If *bytes* is larger than the current data being
1690 * processed, then the eBPF verdict will be applied to multiple
1691 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1694 * Note that if a socket closes with the internal counter holding
1695 * a non-zero value, this is not a problem because data is not
1696 * being buffered for *bytes* and is sent as it is received.
1700 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1702 * For socket policies, prevent the execution of the verdict eBPF
1703 * program for message *msg* until *bytes* (byte number) have been
1706 * This can be used when one needs a specific number of bytes
1707 * before a verdict can be assigned, even if the data spans
1708 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1709 * case would be a user calling **sendmsg**\ () repeatedly with
1710 * 1-byte long message segments. Obviously, this is bad for
1711 * performance, but it is still valid. If the eBPF program needs
1712 * *bytes* bytes to validate a header, this helper can be used to
1713 * prevent the eBPF program to be called again until *bytes* have
1718 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1720 * For socket policies, pull in non-linear data from user space
1721 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1722 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1725 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1726 * *msg* it can only parse data that the (**data**, **data_end**)
1727 * pointers have already consumed. For **sendmsg**\ () hooks this
1728 * is likely the first scatterlist element. But for calls relying
1729 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1730 * be the range (**0**, **0**) because the data is shared with
1731 * user space and by default the objective is to avoid allowing
1732 * user space to modify data while (or after) eBPF verdict is
1733 * being decided. This helper can be used to pull in data and to
1734 * set the start and end pointer to given values. Data will be
1735 * copied if necessary (i.e. if data was not linear and if start
1736 * and end pointers do not point to the same chunk).
1738 * A call to this helper is susceptible to change the underlaying
1739 * packet buffer. Therefore, at load time, all checks on pointers
1740 * previously done by the verifier are invalidated and must be
1741 * performed again, if the helper is used in combination with
1742 * direct packet access.
1744 * All values for *flags* are reserved for future usage, and must
1747 * 0 on success, or a negative error in case of failure.
1749 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1751 * Bind the socket associated to *ctx* to the address pointed by
1752 * *addr*, of length *addr_len*. This allows for making outgoing
1753 * connection from the desired IP address, which can be useful for
1754 * example when all processes inside a cgroup should use one
1755 * single IP address on a host that has multiple IP configured.
1757 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1758 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1759 * **AF_INET6**). Looking for a free port to bind to can be
1760 * expensive, therefore binding to port is not permitted by the
1761 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1762 * must be set to zero.
1764 * 0 on success, or a negative error in case of failure.
1766 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1768 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1769 * only possible to shrink the packet as of this writing,
1770 * therefore *delta* must be a negative integer.
1772 * A call to this helper is susceptible to change the underlaying
1773 * packet buffer. Therefore, at load time, all checks on pointers
1774 * previously done by the verifier are invalidated and must be
1775 * performed again, if the helper is used in combination with
1776 * direct packet access.
1778 * 0 on success, or a negative error in case of failure.
1780 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1782 * Retrieve the XFRM state (IP transform framework, see also
1783 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1785 * The retrieved value is stored in the **struct bpf_xfrm_state**
1786 * pointed by *xfrm_state* and of length *size*.
1788 * All values for *flags* are reserved for future usage, and must
1791 * This helper is available only if the kernel was compiled with
1792 * **CONFIG_XFRM** configuration option.
1794 * 0 on success, or a negative error in case of failure.
1796 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1798 * Return a user or a kernel stack in bpf program provided buffer.
1799 * To achieve this, the helper needs *ctx*, which is a pointer
1800 * to the context on which the tracing program is executed.
1801 * To store the stacktrace, the bpf program provides *buf* with
1802 * a nonnegative *size*.
1804 * The last argument, *flags*, holds the number of stack frames to
1805 * skip (from 0 to 255), masked with
1806 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1807 * the following flags:
1809 * **BPF_F_USER_STACK**
1810 * Collect a user space stack instead of a kernel stack.
1811 * **BPF_F_USER_BUILD_ID**
1812 * Collect buildid+offset instead of ips for user stack,
1813 * only valid if **BPF_F_USER_STACK** is also specified.
1815 * **bpf_get_stack**\ () can collect up to
1816 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1817 * to sufficient large buffer size. Note that
1818 * this limit can be controlled with the **sysctl** program, and
1819 * that it should be manually increased in order to profile long
1820 * user stacks (such as stacks for Java programs). To do so, use:
1824 * # sysctl kernel.perf_event_max_stack=<new value>
1826 * A non-negative value equal to or less than *size* on success,
1827 * or 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"
1849 * 0 on success, or a negative error in case of failure.
1851 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1853 * Do FIB lookup in kernel tables using parameters in *params*.
1854 * If lookup is successful and result shows packet is to be
1855 * forwarded, the neighbor tables are searched for the nexthop.
1856 * If successful (ie., FIB lookup shows forwarding and nexthop
1857 * is resolved), the nexthop address is returned in ipv4_dst
1858 * or ipv6_dst based on family, smac is set to mac address of
1859 * egress device, dmac is set to nexthop mac address, rt_metric
1860 * is set to metric from route (IPv4/IPv6 only).
1862 * *plen* argument is the size of the passed in struct.
1863 * *flags* argument can be a combination of one or more of the
1866 * **BPF_FIB_LOOKUP_DIRECT**
1867 * Do a direct table lookup vs full lookup using FIB
1869 * **BPF_FIB_LOOKUP_OUTPUT**
1870 * Perform lookup from an egress perspective (default is
1873 * *ctx* is either **struct xdp_md** for XDP programs or
1874 * **struct sk_buff** tc cls_act programs.
1876 * Egress device index on success, 0 if packet needs to continue
1877 * up the stack for further processing or a negative error in case
1880 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1882 * Add an entry to, or update a sockhash *map* referencing sockets.
1883 * The *skops* is used as a new value for the entry associated to
1884 * *key*. *flags* is one of:
1887 * The entry for *key* must not exist in the map.
1889 * The entry for *key* must already exist in the map.
1891 * No condition on the existence of the entry for *key*.
1893 * If the *map* has eBPF programs (parser and verdict), those will
1894 * be inherited by the socket being added. If the socket is
1895 * already attached to eBPF programs, this results in an error.
1897 * 0 on success, or a negative error in case of failure.
1899 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1901 * This helper is used in programs implementing policies at the
1902 * socket level. If the message *msg* is allowed to pass (i.e. if
1903 * the verdict eBPF program returns **SK_PASS**), redirect it to
1904 * the socket referenced by *map* (of type
1905 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1906 * egress interfaces can be used for redirection. The
1907 * **BPF_F_INGRESS** value in *flags* is used to make the
1908 * distinction (ingress path is selected if the flag is present,
1909 * egress path otherwise). This is the only flag supported for now.
1911 * **SK_PASS** on success, or **SK_DROP** on error.
1913 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1915 * This helper is used in programs implementing policies at the
1916 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1917 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1918 * to the socket referenced by *map* (of type
1919 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1920 * egress interfaces can be used for redirection. The
1921 * **BPF_F_INGRESS** value in *flags* is used to make the
1922 * distinction (ingress path is selected if the flag is present,
1923 * egress otherwise). This is the only flag supported for now.
1925 * **SK_PASS** on success, or **SK_DROP** on error.
1927 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
1929 * Encapsulate the packet associated to *skb* within a Layer 3
1930 * protocol header. This header is provided in the buffer at
1931 * address *hdr*, with *len* its size in bytes. *type* indicates
1932 * the protocol of the header and can be one of:
1934 * **BPF_LWT_ENCAP_SEG6**
1935 * IPv6 encapsulation with Segment Routing Header
1936 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
1937 * the IPv6 header is computed by the kernel.
1938 * **BPF_LWT_ENCAP_SEG6_INLINE**
1939 * Only works if *skb* contains an IPv6 packet. Insert a
1940 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
1943 * A call to this helper is susceptible to change the underlaying
1944 * packet buffer. Therefore, at load time, all checks on pointers
1945 * previously done by the verifier are invalidated and must be
1946 * performed again, if the helper is used in combination with
1947 * direct packet access.
1949 * 0 on success, or a negative error in case of failure.
1951 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
1953 * Store *len* bytes from address *from* into the packet
1954 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
1955 * inside the outermost IPv6 Segment Routing Header can be
1956 * modified through this helper.
1958 * A call to this helper is susceptible to change the underlaying
1959 * packet buffer. Therefore, at load time, all checks on pointers
1960 * previously done by the verifier are invalidated and must be
1961 * performed again, if the helper is used in combination with
1962 * direct packet access.
1964 * 0 on success, or a negative error in case of failure.
1966 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
1968 * Adjust the size allocated to TLVs in the outermost IPv6
1969 * Segment Routing Header contained in the packet associated to
1970 * *skb*, at position *offset* by *delta* bytes. Only offsets
1971 * after the segments are accepted. *delta* can be as well
1972 * positive (growing) as negative (shrinking).
1974 * A call to this helper is susceptible to change the underlaying
1975 * packet buffer. Therefore, at load time, all checks on pointers
1976 * previously done by the verifier are invalidated and must be
1977 * performed again, if the helper is used in combination with
1978 * direct packet access.
1980 * 0 on success, or a negative error in case of failure.
1982 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
1984 * Apply an IPv6 Segment Routing action of type *action* to the
1985 * packet associated to *skb*. Each action takes a parameter
1986 * contained at address *param*, and of length *param_len* bytes.
1987 * *action* can be one of:
1989 * **SEG6_LOCAL_ACTION_END_X**
1990 * End.X action: Endpoint with Layer-3 cross-connect.
1991 * Type of *param*: **struct in6_addr**.
1992 * **SEG6_LOCAL_ACTION_END_T**
1993 * End.T action: Endpoint with specific IPv6 table lookup.
1994 * Type of *param*: **int**.
1995 * **SEG6_LOCAL_ACTION_END_B6**
1996 * End.B6 action: Endpoint bound to an SRv6 policy.
1997 * Type of param: **struct ipv6_sr_hdr**.
1998 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
1999 * End.B6.Encap action: Endpoint bound to an SRv6
2000 * encapsulation policy.
2001 * Type of param: **struct ipv6_sr_hdr**.
2003 * A call to this helper is susceptible to change the underlaying
2004 * packet buffer. Therefore, at load time, all checks on pointers
2005 * previously done by the verifier are invalidated and must be
2006 * performed again, if the helper is used in combination with
2007 * direct packet access.
2009 * 0 on success, or a negative error in case of failure.
2011 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2013 * This helper is used in programs implementing IR decoding, to
2014 * report a successfully decoded key press with *scancode*,
2015 * *toggle* value in the given *protocol*. The scancode will be
2016 * translated to a keycode using the rc keymap, and reported as
2017 * an input key down event. After a period a key up event is
2018 * generated. This period can be extended by calling either
2019 * **bpf_rc_keydown** () again with the same values, or calling
2020 * **bpf_rc_repeat** ().
2022 * Some protocols include a toggle bit, in case the button was
2023 * released and pressed again between consecutive scancodes.
2025 * The *ctx* should point to the lirc sample as passed into
2028 * The *protocol* is the decoded protocol number (see
2029 * **enum rc_proto** for some predefined values).
2031 * This helper is only available is the kernel was compiled with
2032 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2038 * int bpf_rc_repeat(void *ctx)
2040 * This helper is used in programs implementing IR decoding, to
2041 * report a successfully decoded repeat key message. This delays
2042 * the generation of a key up event for previously generated
2045 * Some IR protocols like NEC have a special IR message for
2046 * repeating last button, for when a button is held down.
2048 * The *ctx* should point to the lirc sample as passed into
2051 * This helper is only available is the kernel was compiled with
2052 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2058 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2060 * Return the cgroup v2 id of the socket associated with the *skb*.
2061 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2062 * helper for cgroup v1 by providing a tag resp. identifier that
2063 * can be matched on or used for map lookups e.g. to implement
2064 * policy. The cgroup v2 id of a given path in the hierarchy is
2065 * exposed in user space through the f_handle API in order to get
2066 * to the same 64-bit id.
2068 * This helper can be used on TC egress path, but not on ingress,
2069 * and is available only if the kernel was compiled with the
2070 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2072 * The id is returned or 0 in case the id could not be retrieved.
2074 #define __BPF_FUNC_MAPPER(FN) \
2076 FN(map_lookup_elem), \
2077 FN(map_update_elem), \
2078 FN(map_delete_elem), \
2082 FN(get_prandom_u32), \
2083 FN(get_smp_processor_id), \
2084 FN(skb_store_bytes), \
2085 FN(l3_csum_replace), \
2086 FN(l4_csum_replace), \
2088 FN(clone_redirect), \
2089 FN(get_current_pid_tgid), \
2090 FN(get_current_uid_gid), \
2091 FN(get_current_comm), \
2092 FN(get_cgroup_classid), \
2093 FN(skb_vlan_push), \
2095 FN(skb_get_tunnel_key), \
2096 FN(skb_set_tunnel_key), \
2097 FN(perf_event_read), \
2099 FN(get_route_realm), \
2100 FN(perf_event_output), \
2101 FN(skb_load_bytes), \
2104 FN(skb_get_tunnel_opt), \
2105 FN(skb_set_tunnel_opt), \
2106 FN(skb_change_proto), \
2107 FN(skb_change_type), \
2108 FN(skb_under_cgroup), \
2109 FN(get_hash_recalc), \
2110 FN(get_current_task), \
2111 FN(probe_write_user), \
2112 FN(current_task_under_cgroup), \
2113 FN(skb_change_tail), \
2114 FN(skb_pull_data), \
2116 FN(set_hash_invalid), \
2117 FN(get_numa_node_id), \
2118 FN(skb_change_head), \
2119 FN(xdp_adjust_head), \
2120 FN(probe_read_str), \
2121 FN(get_socket_cookie), \
2122 FN(get_socket_uid), \
2125 FN(skb_adjust_room), \
2127 FN(sk_redirect_map), \
2128 FN(sock_map_update), \
2129 FN(xdp_adjust_meta), \
2130 FN(perf_event_read_value), \
2131 FN(perf_prog_read_value), \
2133 FN(override_return), \
2134 FN(sock_ops_cb_flags_set), \
2135 FN(msg_redirect_map), \
2136 FN(msg_apply_bytes), \
2137 FN(msg_cork_bytes), \
2138 FN(msg_pull_data), \
2140 FN(xdp_adjust_tail), \
2141 FN(skb_get_xfrm_state), \
2143 FN(skb_load_bytes_relative), \
2145 FN(sock_hash_update), \
2146 FN(msg_redirect_hash), \
2147 FN(sk_redirect_hash), \
2148 FN(lwt_push_encap), \
2149 FN(lwt_seg6_store_bytes), \
2150 FN(lwt_seg6_adjust_srh), \
2151 FN(lwt_seg6_action), \
2156 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2157 * function eBPF program intends to call
2159 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2161 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2164 #undef __BPF_ENUM_FN
2166 /* All flags used by eBPF helper functions, placed here. */
2168 /* BPF_FUNC_skb_store_bytes flags. */
2169 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2170 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2172 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2173 * First 4 bits are for passing the header field size.
2175 #define BPF_F_HDR_FIELD_MASK 0xfULL
2177 /* BPF_FUNC_l4_csum_replace flags. */
2178 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2179 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2180 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2182 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2183 #define BPF_F_INGRESS (1ULL << 0)
2185 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2186 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2188 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2189 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2190 #define BPF_F_USER_STACK (1ULL << 8)
2191 /* flags used by BPF_FUNC_get_stackid only. */
2192 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2193 #define BPF_F_REUSE_STACKID (1ULL << 10)
2194 /* flags used by BPF_FUNC_get_stack only. */
2195 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2197 /* BPF_FUNC_skb_set_tunnel_key flags. */
2198 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2199 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2200 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2202 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2203 * BPF_FUNC_perf_event_read_value flags.
2205 #define BPF_F_INDEX_MASK 0xffffffffULL
2206 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2207 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2208 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2210 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2211 enum bpf_adj_room_mode {
2215 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2216 enum bpf_hdr_start_off {
2221 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2222 enum bpf_lwt_encap_mode {
2224 BPF_LWT_ENCAP_SEG6_INLINE
2227 /* user accessible mirror of in-kernel sk_buff.
2228 * new fields can only be added to the end of this structure
2234 __u32 queue_mapping;
2240 __u32 ingress_ifindex;
2250 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2252 __u32 remote_ip4; /* Stored in network byte order */
2253 __u32 local_ip4; /* Stored in network byte order */
2254 __u32 remote_ip6[4]; /* Stored in network byte order */
2255 __u32 local_ip6[4]; /* Stored in network byte order */
2256 __u32 remote_port; /* Stored in network byte order */
2257 __u32 local_port; /* stored in host byte order */
2263 struct bpf_tunnel_key {
2267 __u32 remote_ipv6[4];
2275 /* user accessible mirror of in-kernel xfrm_state.
2276 * new fields can only be added to the end of this structure
2278 struct bpf_xfrm_state {
2280 __u32 spi; /* Stored in network byte order */
2283 __u32 remote_ipv4; /* Stored in network byte order */
2284 __u32 remote_ipv6[4]; /* Stored in network byte order */
2288 /* Generic BPF return codes which all BPF program types may support.
2289 * The values are binary compatible with their TC_ACT_* counter-part to
2290 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2293 * XDP is handled seprately, see XDP_*.
2301 /* >127 are reserved for prog type specific return codes */
2311 __u32 src_ip4; /* Allows 1,2,4-byte read.
2312 * Stored in network byte order.
2314 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2315 * Stored in network byte order.
2317 __u32 src_port; /* Allows 4-byte read.
2318 * Stored in host byte order
2322 #define XDP_PACKET_HEADROOM 256
2324 /* User return codes for XDP prog type.
2325 * A valid XDP program must return one of these defined values. All other
2326 * return codes are reserved for future use. Unknown return codes will
2327 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2337 /* user accessible metadata for XDP packet hook
2338 * new fields must be added to the end of this structure
2344 /* Below access go through struct xdp_rxq_info */
2345 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2346 __u32 rx_queue_index; /* rxq->queue_index */
2354 /* user accessible metadata for SK_MSG packet hook, new fields must
2355 * be added to the end of this structure
2362 __u32 remote_ip4; /* Stored in network byte order */
2363 __u32 local_ip4; /* Stored in network byte order */
2364 __u32 remote_ip6[4]; /* Stored in network byte order */
2365 __u32 local_ip6[4]; /* Stored in network byte order */
2366 __u32 remote_port; /* Stored in network byte order */
2367 __u32 local_port; /* stored in host byte order */
2370 #define BPF_TAG_SIZE 8
2372 struct bpf_prog_info {
2375 __u8 tag[BPF_TAG_SIZE];
2376 __u32 jited_prog_len;
2377 __u32 xlated_prog_len;
2378 __aligned_u64 jited_prog_insns;
2379 __aligned_u64 xlated_prog_insns;
2380 __u64 load_time; /* ns since boottime */
2381 __u32 created_by_uid;
2383 __aligned_u64 map_ids;
2384 char name[BPF_OBJ_NAME_LEN];
2386 __u32 gpl_compatible:1;
2389 __u32 nr_jited_ksyms;
2390 __u32 nr_jited_func_lens;
2391 __aligned_u64 jited_ksyms;
2392 __aligned_u64 jited_func_lens;
2393 } __attribute__((aligned(8)));
2395 struct bpf_map_info {
2402 char name[BPF_OBJ_NAME_LEN];
2407 __u32 btf_key_type_id;
2408 __u32 btf_value_type_id;
2409 } __attribute__((aligned(8)));
2411 struct bpf_btf_info {
2415 } __attribute__((aligned(8)));
2417 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2418 * by user and intended to be used by socket (e.g. to bind to, depends on
2419 * attach attach type).
2421 struct bpf_sock_addr {
2422 __u32 user_family; /* Allows 4-byte read, but no write. */
2423 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2424 * Stored in network byte order.
2426 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2427 * Stored in network byte order.
2429 __u32 user_port; /* Allows 4-byte read and write.
2430 * Stored in network byte order
2432 __u32 family; /* Allows 4-byte read, but no write */
2433 __u32 type; /* Allows 4-byte read, but no write */
2434 __u32 protocol; /* Allows 4-byte read, but no write */
2435 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2436 * Stored in network byte order.
2438 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2439 * Stored in network byte order.
2443 /* User bpf_sock_ops struct to access socket values and specify request ops
2444 * and their replies.
2445 * Some of this fields are in network (bigendian) byte order and may need
2446 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2447 * New fields can only be added at the end of this structure
2449 struct bpf_sock_ops {
2452 __u32 args[4]; /* Optionally passed to bpf program */
2453 __u32 reply; /* Returned by bpf program */
2454 __u32 replylong[4]; /* Optionally returned by bpf prog */
2457 __u32 remote_ip4; /* Stored in network byte order */
2458 __u32 local_ip4; /* Stored in network byte order */
2459 __u32 remote_ip6[4]; /* Stored in network byte order */
2460 __u32 local_ip6[4]; /* Stored in network byte order */
2461 __u32 remote_port; /* Stored in network byte order */
2462 __u32 local_port; /* stored in host byte order */
2463 __u32 is_fullsock; /* Some TCP fields are only valid if
2464 * there is a full socket. If not, the
2465 * fields read as zero.
2468 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2469 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2478 __u32 rate_delivered;
2479 __u32 rate_interval_us;
2482 __u32 total_retrans;
2486 __u32 data_segs_out;
2490 __u64 bytes_received;
2494 /* Definitions for bpf_sock_ops_cb_flags */
2495 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2496 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2497 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2498 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2499 * supported cb flags
2502 /* List of known BPF sock_ops operators.
2503 * New entries can only be added at the end
2507 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2508 * -1 if default value should be used
2510 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2511 * window (in packets) or -1 if default
2512 * value should be used
2514 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2515 * active connection is initialized
2517 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2518 * active connection is
2521 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2522 * passive connection is
2525 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2528 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2529 * based on the path and may be
2530 * dependent on the congestion control
2531 * algorithm. In general it indicates
2532 * a congestion threshold. RTTs above
2533 * this indicate congestion
2535 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2536 * Arg1: value of icsk_retransmits
2537 * Arg2: value of icsk_rto
2538 * Arg3: whether RTO has expired
2540 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2541 * Arg1: sequence number of 1st byte
2543 * Arg3: return value of
2544 * tcp_transmit_skb (0 => success)
2546 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2552 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2553 * changes between the TCP and BPF versions. Ideally this should never happen.
2554 * If it does, we need to add code to convert them before calling
2555 * the BPF sock_ops function.
2558 BPF_TCP_ESTABLISHED = 1,
2568 BPF_TCP_CLOSING, /* Now a valid state */
2569 BPF_TCP_NEW_SYN_RECV,
2571 BPF_TCP_MAX_STATES /* Leave at the end! */
2574 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
2575 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
2577 struct bpf_perf_event_value {
2583 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2584 #define BPF_DEVCG_ACC_READ (1ULL << 1)
2585 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2587 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2588 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2590 struct bpf_cgroup_dev_ctx {
2591 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2597 struct bpf_raw_tracepoint_args {
2601 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
2602 * OUTPUT: Do lookup from egress perspective; default is ingress
2604 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
2605 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
2607 struct bpf_fib_lookup {
2608 /* input: network family for lookup (AF_INET, AF_INET6)
2609 * output: network family of egress nexthop
2613 /* set if lookup is to consider L4 data - e.g., FIB rules */
2618 /* total length of packet from network header - used for MTU check */
2620 __u32 ifindex; /* L3 device index for lookup */
2623 /* inputs to lookup */
2624 __u8 tos; /* AF_INET */
2625 __be32 flowlabel; /* AF_INET6 */
2627 /* output: metric of fib result (IPv4/IPv6 only) */
2633 __u32 ipv6_src[4]; /* in6_addr; network order */
2636 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2637 * network header. output: bpf_fib_lookup sets to gateway address
2638 * if FIB lookup returns gateway route
2642 __u32 ipv6_dst[4]; /* in6_addr; network order */
2646 __be16 h_vlan_proto;
2648 __u8 smac[6]; /* ETH_ALEN */
2649 __u8 dmac[6]; /* ETH_ALEN */
2652 enum bpf_task_fd_type {
2653 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
2654 BPF_FD_TYPE_TRACEPOINT, /* tp name */
2655 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
2656 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
2657 BPF_FD_TYPE_UPROBE, /* filename + offset */
2658 BPF_FD_TYPE_URETPROBE, /* filename + offset */
2661 #endif /* _UAPI__LINUX_BPF_H__ */