1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_ALU64 0x07 /* alu mode in double word width */
20 #define BPF_DW 0x18 /* double word (64-bit) */
21 #define BPF_XADD 0xc0 /* exclusive add */
24 #define BPF_MOV 0xb0 /* mov reg to reg */
25 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
27 /* change endianness of a register */
28 #define BPF_END 0xd0 /* flags for endianness conversion: */
29 #define BPF_TO_LE 0x00 /* convert to little-endian */
30 #define BPF_TO_BE 0x08 /* convert to big-endian */
31 #define BPF_FROM_LE BPF_TO_LE
32 #define BPF_FROM_BE BPF_TO_BE
35 #define BPF_JNE 0x50 /* jump != */
36 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
37 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
38 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
39 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
40 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
41 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
42 #define BPF_CALL 0x80 /* function call */
43 #define BPF_EXIT 0x90 /* function return */
45 /* Register numbers */
61 /* BPF has 10 general purpose 64-bit registers and stack frame. */
62 #define MAX_BPF_REG __MAX_BPF_REG
65 __u8 code; /* opcode */
66 __u8 dst_reg:4; /* dest register */
67 __u8 src_reg:4; /* source register */
68 __s16 off; /* signed offset */
69 __s32 imm; /* signed immediate constant */
72 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
73 struct bpf_lpm_trie_key {
74 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
75 __u8 data[0]; /* Arbitrary size */
78 struct bpf_cgroup_storage_key {
79 __u64 cgroup_inode_id; /* cgroup inode id */
80 __u32 attach_type; /* program attach type */
83 /* BPF syscall commands, see bpf(2) man-page for details. */
98 BPF_PROG_GET_FD_BY_ID,
100 BPF_OBJ_GET_INFO_BY_FD,
102 BPF_RAW_TRACEPOINT_OPEN,
104 BPF_BTF_GET_FD_BY_ID,
106 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
113 BPF_MAP_TYPE_PROG_ARRAY,
114 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
115 BPF_MAP_TYPE_PERCPU_HASH,
116 BPF_MAP_TYPE_PERCPU_ARRAY,
117 BPF_MAP_TYPE_STACK_TRACE,
118 BPF_MAP_TYPE_CGROUP_ARRAY,
119 BPF_MAP_TYPE_LRU_HASH,
120 BPF_MAP_TYPE_LRU_PERCPU_HASH,
121 BPF_MAP_TYPE_LPM_TRIE,
122 BPF_MAP_TYPE_ARRAY_OF_MAPS,
123 BPF_MAP_TYPE_HASH_OF_MAPS,
125 BPF_MAP_TYPE_SOCKMAP,
128 BPF_MAP_TYPE_SOCKHASH,
129 BPF_MAP_TYPE_CGROUP_STORAGE,
130 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
131 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
137 BPF_PROG_TYPE_UNSPEC,
138 BPF_PROG_TYPE_SOCKET_FILTER,
139 BPF_PROG_TYPE_KPROBE,
140 BPF_PROG_TYPE_SCHED_CLS,
141 BPF_PROG_TYPE_SCHED_ACT,
142 BPF_PROG_TYPE_TRACEPOINT,
144 BPF_PROG_TYPE_PERF_EVENT,
145 BPF_PROG_TYPE_CGROUP_SKB,
146 BPF_PROG_TYPE_CGROUP_SOCK,
147 BPF_PROG_TYPE_LWT_IN,
148 BPF_PROG_TYPE_LWT_OUT,
149 BPF_PROG_TYPE_LWT_XMIT,
150 BPF_PROG_TYPE_SOCK_OPS,
151 BPF_PROG_TYPE_SK_SKB,
152 BPF_PROG_TYPE_CGROUP_DEVICE,
153 BPF_PROG_TYPE_SK_MSG,
154 BPF_PROG_TYPE_RAW_TRACEPOINT,
155 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
156 BPF_PROG_TYPE_LWT_SEG6LOCAL,
157 BPF_PROG_TYPE_LIRC_MODE2,
158 BPF_PROG_TYPE_SK_REUSEPORT,
159 BPF_PROG_TYPE_FLOW_DISSECTOR,
162 enum bpf_attach_type {
163 BPF_CGROUP_INET_INGRESS,
164 BPF_CGROUP_INET_EGRESS,
165 BPF_CGROUP_INET_SOCK_CREATE,
167 BPF_SK_SKB_STREAM_PARSER,
168 BPF_SK_SKB_STREAM_VERDICT,
171 BPF_CGROUP_INET4_BIND,
172 BPF_CGROUP_INET6_BIND,
173 BPF_CGROUP_INET4_CONNECT,
174 BPF_CGROUP_INET6_CONNECT,
175 BPF_CGROUP_INET4_POST_BIND,
176 BPF_CGROUP_INET6_POST_BIND,
177 BPF_CGROUP_UDP4_SENDMSG,
178 BPF_CGROUP_UDP6_SENDMSG,
181 __MAX_BPF_ATTACH_TYPE
184 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
186 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
188 * NONE(default): No further bpf programs allowed in the subtree.
190 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
191 * the program in this cgroup yields to sub-cgroup program.
193 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
194 * that cgroup program gets run in addition to the program in this cgroup.
196 * Only one program is allowed to be attached to a cgroup with
197 * NONE or BPF_F_ALLOW_OVERRIDE flag.
198 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
199 * release old program and attach the new one. Attach flags has to match.
201 * Multiple programs are allowed to be attached to a cgroup with
202 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
203 * (those that were attached first, run first)
204 * The programs of sub-cgroup are executed first, then programs of
205 * this cgroup and then programs of parent cgroup.
206 * When children program makes decision (like picking TCP CA or sock bind)
207 * parent program has a chance to override it.
209 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
210 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
212 * cgrp1 (MULTI progs A, B) ->
213 * cgrp2 (OVERRIDE prog C) ->
214 * cgrp3 (MULTI prog D) ->
215 * cgrp4 (OVERRIDE prog E) ->
216 * cgrp5 (NONE prog F)
217 * the event in cgrp5 triggers execution of F,D,A,B in that order.
218 * if prog F is detached, the execution is E,D,A,B
219 * if prog F and D are detached, the execution is E,A,B
220 * if prog F, E and D are detached, the execution is C,A,B
222 * All eligible programs are executed regardless of return code from
225 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
226 #define BPF_F_ALLOW_MULTI (1U << 1)
228 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
229 * verifier will perform strict alignment checking as if the kernel
230 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
231 * and NET_IP_ALIGN defined to 2.
233 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
235 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
236 #define BPF_PSEUDO_MAP_FD 1
238 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
239 * offset to another bpf function
241 #define BPF_PSEUDO_CALL 1
243 /* flags for BPF_MAP_UPDATE_ELEM command */
244 #define BPF_ANY 0 /* create new element or update existing */
245 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
246 #define BPF_EXIST 2 /* update existing element */
248 /* flags for BPF_MAP_CREATE command */
249 #define BPF_F_NO_PREALLOC (1U << 0)
250 /* Instead of having one common LRU list in the
251 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
252 * which can scale and perform better.
253 * Note, the LRU nodes (including free nodes) cannot be moved
254 * across different LRU lists.
256 #define BPF_F_NO_COMMON_LRU (1U << 1)
257 /* Specify numa node during map creation */
258 #define BPF_F_NUMA_NODE (1U << 2)
260 /* flags for BPF_PROG_QUERY */
261 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
263 #define BPF_OBJ_NAME_LEN 16U
265 /* Flags for accessing BPF object */
266 #define BPF_F_RDONLY (1U << 3)
267 #define BPF_F_WRONLY (1U << 4)
269 /* Flag for stack_map, store build_id+offset instead of pointer */
270 #define BPF_F_STACK_BUILD_ID (1U << 5)
272 enum bpf_stack_build_id_status {
273 /* user space need an empty entry to identify end of a trace */
274 BPF_STACK_BUILD_ID_EMPTY = 0,
275 /* with valid build_id and offset */
276 BPF_STACK_BUILD_ID_VALID = 1,
277 /* couldn't get build_id, fallback to ip */
278 BPF_STACK_BUILD_ID_IP = 2,
281 #define BPF_BUILD_ID_SIZE 20
282 struct bpf_stack_build_id {
284 unsigned char build_id[BPF_BUILD_ID_SIZE];
292 struct { /* anonymous struct used by BPF_MAP_CREATE command */
293 __u32 map_type; /* one of enum bpf_map_type */
294 __u32 key_size; /* size of key in bytes */
295 __u32 value_size; /* size of value in bytes */
296 __u32 max_entries; /* max number of entries in a map */
297 __u32 map_flags; /* BPF_MAP_CREATE related
298 * flags defined above.
300 __u32 inner_map_fd; /* fd pointing to the inner map */
301 __u32 numa_node; /* numa node (effective only if
302 * BPF_F_NUMA_NODE is set).
304 char map_name[BPF_OBJ_NAME_LEN];
305 __u32 map_ifindex; /* ifindex of netdev to create on */
306 __u32 btf_fd; /* fd pointing to a BTF type data */
307 __u32 btf_key_type_id; /* BTF type_id of the key */
308 __u32 btf_value_type_id; /* BTF type_id of the value */
311 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
316 __aligned_u64 next_key;
321 struct { /* anonymous struct used by BPF_PROG_LOAD command */
322 __u32 prog_type; /* one of enum bpf_prog_type */
325 __aligned_u64 license;
326 __u32 log_level; /* verbosity level of verifier */
327 __u32 log_size; /* size of user buffer */
328 __aligned_u64 log_buf; /* user supplied buffer */
329 __u32 kern_version; /* checked when prog_type=kprobe */
331 char prog_name[BPF_OBJ_NAME_LEN];
332 __u32 prog_ifindex; /* ifindex of netdev to prep for */
333 /* For some prog types expected attach type must be known at
334 * load time to verify attach type specific parts of prog
335 * (context accesses, allowed helpers, etc).
337 __u32 expected_attach_type;
340 struct { /* anonymous struct used by BPF_OBJ_* commands */
341 __aligned_u64 pathname;
346 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
347 __u32 target_fd; /* container object to attach to */
348 __u32 attach_bpf_fd; /* eBPF program to attach */
353 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
358 __aligned_u64 data_in;
359 __aligned_u64 data_out;
364 struct { /* anonymous struct used by BPF_*_GET_*_ID */
375 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
381 struct { /* anonymous struct used by BPF_PROG_QUERY command */
382 __u32 target_fd; /* container object to query */
386 __aligned_u64 prog_ids;
395 struct { /* anonymous struct for BPF_BTF_LOAD */
397 __aligned_u64 btf_log_buf;
404 __u32 pid; /* input: pid */
405 __u32 fd; /* input: fd */
406 __u32 flags; /* input: flags */
407 __u32 buf_len; /* input/output: buf len */
408 __aligned_u64 buf; /* input/output:
409 * tp_name for tracepoint
411 * filename for uprobe
413 __u32 prog_id; /* output: prod_id */
414 __u32 fd_type; /* output: BPF_FD_TYPE_* */
415 __u64 probe_offset; /* output: probe_offset */
416 __u64 probe_addr; /* output: probe_addr */
418 } __attribute__((aligned(8)));
420 /* The description below is an attempt at providing documentation to eBPF
421 * developers about the multiple available eBPF helper functions. It can be
422 * parsed and used to produce a manual page. The workflow is the following,
423 * and requires the rst2man utility:
425 * $ ./scripts/bpf_helpers_doc.py \
426 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
427 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
428 * $ man /tmp/bpf-helpers.7
430 * Note that in order to produce this external documentation, some RST
431 * formatting is used in the descriptions to get "bold" and "italics" in
432 * manual pages. Also note that the few trailing white spaces are
433 * intentional, removing them would break paragraphs for rst2man.
435 * Start of BPF helper function descriptions:
437 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
439 * Perform a lookup in *map* for an entry associated to *key*.
441 * Map value associated to *key*, or **NULL** if no entry was
444 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
446 * Add or update the value of the entry associated to *key* in
447 * *map* with *value*. *flags* is one of:
450 * The entry for *key* must not exist in the map.
452 * The entry for *key* must already exist in the map.
454 * No condition on the existence of the entry for *key*.
456 * Flag value **BPF_NOEXIST** cannot be used for maps of types
457 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
458 * elements always exist), the helper would return an error.
460 * 0 on success, or a negative error in case of failure.
462 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
464 * Delete entry with *key* from *map*.
466 * 0 on success, or a negative error in case of failure.
468 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
470 * Push an element *value* in *map*. *flags* is one of:
473 * If the queue/stack is full, the oldest element is removed to
474 * make room for this.
476 * 0 on success, or a negative error in case of failure.
478 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
480 * Pop an element from *map*.
482 * 0 on success, or a negative error in case of failure.
484 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
486 * Get an element from *map* without removing it.
488 * 0 on success, or a negative error in case of failure.
490 * int bpf_probe_read(void *dst, u32 size, const void *src)
492 * For tracing programs, safely attempt to read *size* bytes from
493 * address *src* and store the data in *dst*.
495 * 0 on success, or a negative error in case of failure.
497 * u64 bpf_ktime_get_ns(void)
499 * Return the time elapsed since system boot, in nanoseconds.
503 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
505 * This helper is a "printk()-like" facility for debugging. It
506 * prints a message defined by format *fmt* (of size *fmt_size*)
507 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
508 * available. It can take up to three additional **u64**
509 * arguments (as an eBPF helpers, the total number of arguments is
512 * Each time the helper is called, it appends a line to the trace.
513 * The format of the trace is customizable, and the exact output
514 * one will get depends on the options set in
515 * *\/sys/kernel/debug/tracing/trace_options* (see also the
516 * *README* file under the same directory). However, it usually
517 * defaults to something like:
521 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
525 * * ``telnet`` is the name of the current task.
526 * * ``470`` is the PID of the current task.
527 * * ``001`` is the CPU number on which the task is
529 * * In ``.N..``, each character refers to a set of
530 * options (whether irqs are enabled, scheduling
531 * options, whether hard/softirqs are running, level of
532 * preempt_disabled respectively). **N** means that
533 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
535 * * ``419421.045894`` is a timestamp.
536 * * ``0x00000001`` is a fake value used by BPF for the
537 * instruction pointer register.
538 * * ``<formatted msg>`` is the message formatted with
541 * The conversion specifiers supported by *fmt* are similar, but
542 * more limited than for printk(). They are **%d**, **%i**,
543 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
544 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
545 * of field, padding with zeroes, etc.) is available, and the
546 * helper will return **-EINVAL** (but print nothing) if it
547 * encounters an unknown specifier.
549 * Also, note that **bpf_trace_printk**\ () is slow, and should
550 * only be used for debugging purposes. For this reason, a notice
551 * bloc (spanning several lines) is printed to kernel logs and
552 * states that the helper should not be used "for production use"
553 * the first time this helper is used (or more precisely, when
554 * **trace_printk**\ () buffers are allocated). For passing values
555 * to user space, perf events should be preferred.
557 * The number of bytes written to the buffer, or a negative error
558 * in case of failure.
560 * u32 bpf_get_prandom_u32(void)
562 * Get a pseudo-random number.
564 * From a security point of view, this helper uses its own
565 * pseudo-random internal state, and cannot be used to infer the
566 * seed of other random functions in the kernel. However, it is
567 * essential to note that the generator used by the helper is not
568 * cryptographically secure.
570 * A random 32-bit unsigned value.
572 * u32 bpf_get_smp_processor_id(void)
574 * Get the SMP (symmetric multiprocessing) processor id. Note that
575 * all programs run with preemption disabled, which means that the
576 * SMP processor id is stable during all the execution of the
579 * The SMP id of the processor running the program.
581 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
583 * Store *len* bytes from address *from* into the packet
584 * associated to *skb*, at *offset*. *flags* are a combination of
585 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
586 * checksum for the packet after storing the bytes) and
587 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
588 * **->swhash** and *skb*\ **->l4hash** to 0).
590 * A call to this helper is susceptible to change the underlaying
591 * packet buffer. Therefore, at load time, all checks on pointers
592 * previously done by the verifier are invalidated and must be
593 * performed again, if the helper is used in combination with
594 * direct packet access.
596 * 0 on success, or a negative error in case of failure.
598 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
600 * Recompute the layer 3 (e.g. IP) checksum for the packet
601 * associated to *skb*. Computation is incremental, so the helper
602 * must know the former value of the header field that was
603 * modified (*from*), the new value of this field (*to*), and the
604 * number of bytes (2 or 4) for this field, stored in *size*.
605 * Alternatively, it is possible to store the difference between
606 * the previous and the new values of the header field in *to*, by
607 * setting *from* and *size* to 0. For both methods, *offset*
608 * indicates the location of the IP checksum within the packet.
610 * This helper works in combination with **bpf_csum_diff**\ (),
611 * which does not update the checksum in-place, but offers more
612 * flexibility and can handle sizes larger than 2 or 4 for the
613 * checksum to update.
615 * A call to this helper is susceptible to change the underlaying
616 * packet buffer. Therefore, at load time, all checks on pointers
617 * previously done by the verifier are invalidated and must be
618 * performed again, if the helper is used in combination with
619 * direct packet access.
621 * 0 on success, or a negative error in case of failure.
623 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
625 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
626 * packet associated to *skb*. Computation is incremental, so the
627 * helper must know the former value of the header field that was
628 * modified (*from*), the new value of this field (*to*), and the
629 * number of bytes (2 or 4) for this field, stored on the lowest
630 * four bits of *flags*. Alternatively, it is possible to store
631 * the difference between the previous and the new values of the
632 * header field in *to*, by setting *from* and the four lowest
633 * bits of *flags* to 0. For both methods, *offset* indicates the
634 * location of the IP checksum within the packet. In addition to
635 * the size of the field, *flags* can be added (bitwise OR) actual
636 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
637 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
638 * for updates resulting in a null checksum the value is set to
639 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
640 * the checksum is to be computed against a pseudo-header.
642 * This helper works in combination with **bpf_csum_diff**\ (),
643 * which does not update the checksum in-place, but offers more
644 * flexibility and can handle sizes larger than 2 or 4 for the
645 * checksum to update.
647 * A call to this helper is susceptible to change the underlaying
648 * packet buffer. Therefore, at load time, all checks on pointers
649 * previously done by the verifier are invalidated and must be
650 * performed again, if the helper is used in combination with
651 * direct packet access.
653 * 0 on success, or a negative error in case of failure.
655 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
657 * This special helper is used to trigger a "tail call", or in
658 * other words, to jump into another eBPF program. The same stack
659 * frame is used (but values on stack and in registers for the
660 * caller are not accessible to the callee). This mechanism allows
661 * for program chaining, either for raising the maximum number of
662 * available eBPF instructions, or to execute given programs in
663 * conditional blocks. For security reasons, there is an upper
664 * limit to the number of successive tail calls that can be
667 * Upon call of this helper, the program attempts to jump into a
668 * program referenced at index *index* in *prog_array_map*, a
669 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
670 * *ctx*, a pointer to the context.
672 * If the call succeeds, the kernel immediately runs the first
673 * instruction of the new program. This is not a function call,
674 * and it never returns to the previous program. If the call
675 * fails, then the helper has no effect, and the caller continues
676 * to run its subsequent instructions. A call can fail if the
677 * destination program for the jump does not exist (i.e. *index*
678 * is superior to the number of entries in *prog_array_map*), or
679 * if the maximum number of tail calls has been reached for this
680 * chain of programs. This limit is defined in the kernel by the
681 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
682 * which is currently set to 32.
684 * 0 on success, or a negative error in case of failure.
686 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
688 * Clone and redirect the packet associated to *skb* to another
689 * net device of index *ifindex*. Both ingress and egress
690 * interfaces can be used for redirection. The **BPF_F_INGRESS**
691 * value in *flags* is used to make the distinction (ingress path
692 * is selected if the flag is present, egress path otherwise).
693 * This is the only flag supported for now.
695 * In comparison with **bpf_redirect**\ () helper,
696 * **bpf_clone_redirect**\ () has the associated cost of
697 * duplicating the packet buffer, but this can be executed out of
698 * the eBPF program. Conversely, **bpf_redirect**\ () is more
699 * efficient, but it is handled through an action code where the
700 * redirection happens only after the eBPF program has returned.
702 * A call to this helper is susceptible to change the underlaying
703 * packet buffer. Therefore, at load time, all checks on pointers
704 * previously done by the verifier are invalidated and must be
705 * performed again, if the helper is used in combination with
706 * direct packet access.
708 * 0 on success, or a negative error in case of failure.
710 * u64 bpf_get_current_pid_tgid(void)
712 * A 64-bit integer containing the current tgid and pid, and
714 * *current_task*\ **->tgid << 32 \|**
715 * *current_task*\ **->pid**.
717 * u64 bpf_get_current_uid_gid(void)
719 * A 64-bit integer containing the current GID and UID, and
720 * created as such: *current_gid* **<< 32 \|** *current_uid*.
722 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
724 * Copy the **comm** attribute of the current task into *buf* of
725 * *size_of_buf*. The **comm** attribute contains the name of
726 * the executable (excluding the path) for the current task. The
727 * *size_of_buf* must be strictly positive. On success, the
728 * helper makes sure that the *buf* is NUL-terminated. On failure,
729 * it is filled with zeroes.
731 * 0 on success, or a negative error in case of failure.
733 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
735 * Retrieve the classid for the current task, i.e. for the net_cls
736 * cgroup to which *skb* belongs.
738 * This helper can be used on TC egress path, but not on ingress.
740 * The net_cls cgroup provides an interface to tag network packets
741 * based on a user-provided identifier for all traffic coming from
742 * the tasks belonging to the related cgroup. See also the related
743 * kernel documentation, available from the Linux sources in file
744 * *Documentation/cgroup-v1/net_cls.txt*.
746 * The Linux kernel has two versions for cgroups: there are
747 * cgroups v1 and cgroups v2. Both are available to users, who can
748 * use a mixture of them, but note that the net_cls cgroup is for
749 * cgroup v1 only. This makes it incompatible with BPF programs
750 * run on cgroups, which is a cgroup-v2-only feature (a socket can
751 * only hold data for one version of cgroups at a time).
753 * This helper is only available is the kernel was compiled with
754 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
755 * "**y**" or to "**m**".
757 * The classid, or 0 for the default unconfigured classid.
759 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
761 * Push a *vlan_tci* (VLAN tag control information) of protocol
762 * *vlan_proto* to the packet associated to *skb*, then update
763 * the checksum. Note that if *vlan_proto* is different from
764 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
765 * be **ETH_P_8021Q**.
767 * A call to this helper is susceptible to change the underlaying
768 * packet buffer. Therefore, at load time, all checks on pointers
769 * previously done by the verifier are invalidated and must be
770 * performed again, if the helper is used in combination with
771 * direct packet access.
773 * 0 on success, or a negative error in case of failure.
775 * int bpf_skb_vlan_pop(struct sk_buff *skb)
777 * Pop a VLAN header from the packet associated to *skb*.
779 * A call to this helper is susceptible to change the underlaying
780 * packet buffer. Therefore, at load time, all checks on pointers
781 * previously done by the verifier are invalidated and must be
782 * performed again, if the helper is used in combination with
783 * direct packet access.
785 * 0 on success, or a negative error in case of failure.
787 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
789 * Get tunnel metadata. This helper takes a pointer *key* to an
790 * empty **struct bpf_tunnel_key** of **size**, that will be
791 * filled with tunnel metadata for the packet associated to *skb*.
792 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
793 * indicates that the tunnel is based on IPv6 protocol instead of
796 * The **struct bpf_tunnel_key** is an object that generalizes the
797 * principal parameters used by various tunneling protocols into a
798 * single struct. This way, it can be used to easily make a
799 * decision based on the contents of the encapsulation header,
800 * "summarized" in this struct. In particular, it holds the IP
801 * address of the remote end (IPv4 or IPv6, depending on the case)
802 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
803 * this struct exposes the *key*\ **->tunnel_id**, which is
804 * generally mapped to a VNI (Virtual Network Identifier), making
805 * it programmable together with the **bpf_skb_set_tunnel_key**\
808 * Let's imagine that the following code is part of a program
809 * attached to the TC ingress interface, on one end of a GRE
810 * tunnel, and is supposed to filter out all messages coming from
811 * remote ends with IPv4 address other than 10.0.0.1:
816 * struct bpf_tunnel_key key = {};
818 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
820 * return TC_ACT_SHOT; // drop packet
822 * if (key.remote_ipv4 != 0x0a000001)
823 * return TC_ACT_SHOT; // drop packet
825 * return TC_ACT_OK; // accept packet
827 * This interface can also be used with all encapsulation devices
828 * that can operate in "collect metadata" mode: instead of having
829 * one network device per specific configuration, the "collect
830 * metadata" mode only requires a single device where the
831 * configuration can be extracted from this helper.
833 * This can be used together with various tunnels such as VXLan,
834 * Geneve, GRE or IP in IP (IPIP).
836 * 0 on success, or a negative error in case of failure.
838 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
840 * Populate tunnel metadata for packet associated to *skb.* The
841 * tunnel metadata is set to the contents of *key*, of *size*. The
842 * *flags* can be set to a combination of the following values:
844 * **BPF_F_TUNINFO_IPV6**
845 * Indicate that the tunnel is based on IPv6 protocol
847 * **BPF_F_ZERO_CSUM_TX**
848 * For IPv4 packets, add a flag to tunnel metadata
849 * indicating that checksum computation should be skipped
850 * and checksum set to zeroes.
851 * **BPF_F_DONT_FRAGMENT**
852 * Add a flag to tunnel metadata indicating that the
853 * packet should not be fragmented.
854 * **BPF_F_SEQ_NUMBER**
855 * Add a flag to tunnel metadata indicating that a
856 * sequence number should be added to tunnel header before
857 * sending the packet. This flag was added for GRE
858 * encapsulation, but might be used with other protocols
859 * as well in the future.
861 * Here is a typical usage on the transmit path:
865 * struct bpf_tunnel_key key;
867 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
868 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
870 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
871 * helper for additional information.
873 * 0 on success, or a negative error in case of failure.
875 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
877 * Read the value of a perf event counter. This helper relies on a
878 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
879 * the perf event counter is selected when *map* is updated with
880 * perf event file descriptors. The *map* is an array whose size
881 * is the number of available CPUs, and each cell contains a value
882 * relative to one CPU. The value to retrieve is indicated by
883 * *flags*, that contains the index of the CPU to look up, masked
884 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
885 * **BPF_F_CURRENT_CPU** to indicate that the value for the
886 * current CPU should be retrieved.
888 * Note that before Linux 4.13, only hardware perf event can be
891 * Also, be aware that the newer helper
892 * **bpf_perf_event_read_value**\ () is recommended over
893 * **bpf_perf_event_read**\ () in general. The latter has some ABI
894 * quirks where error and counter value are used as a return code
895 * (which is wrong to do since ranges may overlap). This issue is
896 * fixed with **bpf_perf_event_read_value**\ (), which at the same
897 * time provides more features over the **bpf_perf_event_read**\
898 * () interface. Please refer to the description of
899 * **bpf_perf_event_read_value**\ () for details.
901 * The value of the perf event counter read from the map, or a
902 * negative error code in case of failure.
904 * int bpf_redirect(u32 ifindex, u64 flags)
906 * Redirect the packet to another net device of index *ifindex*.
907 * This helper is somewhat similar to **bpf_clone_redirect**\
908 * (), except that the packet is not cloned, which provides
909 * increased performance.
911 * Except for XDP, both ingress and egress interfaces can be used
912 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
913 * to make the distinction (ingress path is selected if the flag
914 * is present, egress path otherwise). Currently, XDP only
915 * supports redirection to the egress interface, and accepts no
918 * The same effect can be attained with the more generic
919 * **bpf_redirect_map**\ (), which requires specific maps to be
920 * used but offers better performance.
922 * For XDP, the helper returns **XDP_REDIRECT** on success or
923 * **XDP_ABORTED** on error. For other program types, the values
924 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
927 * u32 bpf_get_route_realm(struct sk_buff *skb)
929 * Retrieve the realm or the route, that is to say the
930 * **tclassid** field of the destination for the *skb*. The
931 * indentifier retrieved is a user-provided tag, similar to the
932 * one used with the net_cls cgroup (see description for
933 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
934 * held by a route (a destination entry), not by a task.
936 * Retrieving this identifier works with the clsact TC egress hook
937 * (see also **tc-bpf(8)**), or alternatively on conventional
938 * classful egress qdiscs, but not on TC ingress path. In case of
939 * clsact TC egress hook, this has the advantage that, internally,
940 * the destination entry has not been dropped yet in the transmit
941 * path. Therefore, the destination entry does not need to be
942 * artificially held via **netif_keep_dst**\ () for a classful
943 * qdisc until the *skb* is freed.
945 * This helper is available only if the kernel was compiled with
946 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
948 * The realm of the route for the packet associated to *skb*, or 0
951 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
953 * Write raw *data* blob into a special BPF perf event held by
954 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
955 * event must have the following attributes: **PERF_SAMPLE_RAW**
956 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
957 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
959 * The *flags* are used to indicate the index in *map* for which
960 * the value must be put, masked with **BPF_F_INDEX_MASK**.
961 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
962 * to indicate that the index of the current CPU core should be
965 * The value to write, of *size*, is passed through eBPF stack and
968 * The context of the program *ctx* needs also be passed to the
971 * On user space, a program willing to read the values needs to
972 * call **perf_event_open**\ () on the perf event (either for
973 * one or for all CPUs) and to store the file descriptor into the
974 * *map*. This must be done before the eBPF program can send data
975 * into it. An example is available in file
976 * *samples/bpf/trace_output_user.c* in the Linux kernel source
977 * tree (the eBPF program counterpart is in
978 * *samples/bpf/trace_output_kern.c*).
980 * **bpf_perf_event_output**\ () achieves better performance
981 * than **bpf_trace_printk**\ () for sharing data with user
982 * space, and is much better suitable for streaming data from eBPF
985 * Note that this helper is not restricted to tracing use cases
986 * and can be used with programs attached to TC or XDP as well,
987 * where it allows for passing data to user space listeners. Data
990 * * Only custom structs,
991 * * Only the packet payload, or
992 * * A combination of both.
994 * 0 on success, or a negative error in case of failure.
996 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
998 * This helper was provided as an easy way to load data from a
999 * packet. It can be used to load *len* bytes from *offset* from
1000 * the packet associated to *skb*, into the buffer pointed by
1003 * Since Linux 4.7, usage of this helper has mostly been replaced
1004 * by "direct packet access", enabling packet data to be
1005 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1006 * pointing respectively to the first byte of packet data and to
1007 * the byte after the last byte of packet data. However, it
1008 * remains useful if one wishes to read large quantities of data
1009 * at once from a packet into the eBPF stack.
1011 * 0 on success, or a negative error in case of failure.
1013 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1015 * Walk a user or a kernel stack and return its id. To achieve
1016 * this, the helper needs *ctx*, which is a pointer to the context
1017 * on which the tracing program is executed, and a pointer to a
1018 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1020 * The last argument, *flags*, holds the number of stack frames to
1021 * skip (from 0 to 255), masked with
1022 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1023 * a combination of the following flags:
1025 * **BPF_F_USER_STACK**
1026 * Collect a user space stack instead of a kernel stack.
1027 * **BPF_F_FAST_STACK_CMP**
1028 * Compare stacks by hash only.
1029 * **BPF_F_REUSE_STACKID**
1030 * If two different stacks hash into the same *stackid*,
1031 * discard the old one.
1033 * The stack id retrieved is a 32 bit long integer handle which
1034 * can be further combined with other data (including other stack
1035 * ids) and used as a key into maps. This can be useful for
1036 * generating a variety of graphs (such as flame graphs or off-cpu
1039 * For walking a stack, this helper is an improvement over
1040 * **bpf_probe_read**\ (), which can be used with unrolled loops
1041 * but is not efficient and consumes a lot of eBPF instructions.
1042 * Instead, **bpf_get_stackid**\ () can collect up to
1043 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1044 * this limit can be controlled with the **sysctl** program, and
1045 * that it should be manually increased in order to profile long
1046 * user stacks (such as stacks for Java programs). To do so, use:
1050 * # sysctl kernel.perf_event_max_stack=<new value>
1052 * The positive or null stack id on success, or a negative error
1053 * in case of failure.
1055 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1057 * Compute a checksum difference, from the raw buffer pointed by
1058 * *from*, of length *from_size* (that must be a multiple of 4),
1059 * towards the raw buffer pointed by *to*, of size *to_size*
1060 * (same remark). An optional *seed* can be added to the value
1061 * (this can be cascaded, the seed may come from a previous call
1064 * This is flexible enough to be used in several ways:
1066 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1067 * checksum, it can be used when pushing new data.
1068 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1069 * checksum, it can be used when removing data from a packet.
1070 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1071 * can be used to compute a diff. Note that *from_size* and
1072 * *to_size* do not need to be equal.
1074 * This helper can be used in combination with
1075 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1076 * which one can feed in the difference computed with
1077 * **bpf_csum_diff**\ ().
1079 * The checksum result, or a negative error code in case of
1082 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1084 * Retrieve tunnel options metadata for the packet associated to
1085 * *skb*, and store the raw tunnel option data to the buffer *opt*
1088 * This helper can be used with encapsulation devices that can
1089 * operate in "collect metadata" mode (please refer to the related
1090 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1091 * more details). A particular example where this can be used is
1092 * in combination with the Geneve encapsulation protocol, where it
1093 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1094 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1095 * the eBPF program. This allows for full customization of these
1098 * The size of the option data retrieved.
1100 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1102 * Set tunnel options metadata for the packet associated to *skb*
1103 * to the option data contained in the raw buffer *opt* of *size*.
1105 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1106 * helper for additional information.
1108 * 0 on success, or a negative error in case of failure.
1110 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1112 * Change the protocol of the *skb* to *proto*. Currently
1113 * supported are transition from IPv4 to IPv6, and from IPv6 to
1114 * IPv4. The helper takes care of the groundwork for the
1115 * transition, including resizing the socket buffer. The eBPF
1116 * program is expected to fill the new headers, if any, via
1117 * **skb_store_bytes**\ () and to recompute the checksums with
1118 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1119 * (). The main case for this helper is to perform NAT64
1120 * operations out of an eBPF program.
1122 * Internally, the GSO type is marked as dodgy so that headers are
1123 * checked and segments are recalculated by the GSO/GRO engine.
1124 * The size for GSO target is adapted as well.
1126 * All values for *flags* are reserved for future usage, and must
1129 * A call to this helper is susceptible to change the underlaying
1130 * packet buffer. Therefore, at load time, all checks on pointers
1131 * previously done by the verifier are invalidated and must be
1132 * performed again, if the helper is used in combination with
1133 * direct packet access.
1135 * 0 on success, or a negative error in case of failure.
1137 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1139 * Change the packet type for the packet associated to *skb*. This
1140 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1141 * the eBPF program does not have a write access to *skb*\
1142 * **->pkt_type** beside this helper. Using a helper here allows
1143 * for graceful handling of errors.
1145 * The major use case is to change incoming *skb*s to
1146 * **PACKET_HOST** in a programmatic way instead of having to
1147 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1150 * Note that *type* only allows certain values. At this time, they
1155 * **PACKET_BROADCAST**
1156 * Send packet to all.
1157 * **PACKET_MULTICAST**
1158 * Send packet to group.
1159 * **PACKET_OTHERHOST**
1160 * Send packet to someone else.
1162 * 0 on success, or a negative error in case of failure.
1164 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1166 * Check whether *skb* is a descendant of the cgroup2 held by
1167 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1169 * The return value depends on the result of the test, and can be:
1171 * * 0, if the *skb* failed the cgroup2 descendant test.
1172 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1173 * * A negative error code, if an error occurred.
1175 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1177 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1178 * not set, in particular if the hash was cleared due to mangling,
1179 * recompute this hash. Later accesses to the hash can be done
1180 * directly with *skb*\ **->hash**.
1182 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1183 * prototype with **bpf_skb_change_proto**\ (), or calling
1184 * **bpf_skb_store_bytes**\ () with the
1185 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1186 * the hash and to trigger a new computation for the next call to
1187 * **bpf_get_hash_recalc**\ ().
1191 * u64 bpf_get_current_task(void)
1193 * A pointer to the current task struct.
1195 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1197 * Attempt in a safe way to write *len* bytes from the buffer
1198 * *src* to *dst* in memory. It only works for threads that are in
1199 * user context, and *dst* must be a valid user space address.
1201 * This helper should not be used to implement any kind of
1202 * security mechanism because of TOC-TOU attacks, but rather to
1203 * debug, divert, and manipulate execution of semi-cooperative
1206 * Keep in mind that this feature is meant for experiments, and it
1207 * has a risk of crashing the system and running programs.
1208 * Therefore, when an eBPF program using this helper is attached,
1209 * a warning including PID and process name is printed to kernel
1212 * 0 on success, or a negative error in case of failure.
1214 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1216 * Check whether the probe is being run is the context of a given
1217 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1218 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1220 * The return value depends on the result of the test, and can be:
1222 * * 0, if the *skb* task belongs to the cgroup2.
1223 * * 1, if the *skb* task does not belong to the cgroup2.
1224 * * A negative error code, if an error occurred.
1226 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1228 * Resize (trim or grow) the packet associated to *skb* to the
1229 * new *len*. The *flags* are reserved for future usage, and must
1232 * The basic idea is that the helper performs the needed work to
1233 * change the size of the packet, then the eBPF program rewrites
1234 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1235 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1236 * and others. This helper is a slow path utility intended for
1237 * replies with control messages. And because it is targeted for
1238 * slow path, the helper itself can afford to be slow: it
1239 * implicitly linearizes, unclones and drops offloads from the
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 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1252 * Pull in non-linear data in case the *skb* is non-linear and not
1253 * all of *len* are part of the linear section. Make *len* bytes
1254 * from *skb* readable and writable. If a zero value is passed for
1255 * *len*, then the whole length of the *skb* is pulled.
1257 * This helper is only needed for reading and writing with direct
1260 * For direct packet access, testing that offsets to access
1261 * are within packet boundaries (test on *skb*\ **->data_end**) is
1262 * susceptible to fail if offsets are invalid, or if the requested
1263 * data is in non-linear parts of the *skb*. On failure the
1264 * program can just bail out, or in the case of a non-linear
1265 * buffer, use a helper to make the data available. The
1266 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1267 * the data. Another one consists in using **bpf_skb_pull_data**
1268 * to pull in once the non-linear parts, then retesting and
1269 * eventually access the data.
1271 * At the same time, this also makes sure the *skb* is uncloned,
1272 * which is a necessary condition for direct write. As this needs
1273 * to be an invariant for the write part only, the verifier
1274 * detects writes and adds a prologue that is calling
1275 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1276 * the very beginning in case it is indeed cloned.
1278 * A call to this helper is susceptible to change the underlaying
1279 * packet buffer. Therefore, at load time, all checks on pointers
1280 * previously done by the verifier are invalidated and must be
1281 * performed again, if the helper is used in combination with
1282 * direct packet access.
1284 * 0 on success, or a negative error in case of failure.
1286 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1288 * Add the checksum *csum* into *skb*\ **->csum** in case the
1289 * driver has supplied a checksum for the entire packet into that
1290 * field. Return an error otherwise. This helper is intended to be
1291 * used in combination with **bpf_csum_diff**\ (), in particular
1292 * when the checksum needs to be updated after data has been
1293 * written into the packet through direct packet access.
1295 * The checksum on success, or a negative error code in case of
1298 * void bpf_set_hash_invalid(struct sk_buff *skb)
1300 * Invalidate the current *skb*\ **->hash**. It can be used after
1301 * mangling on headers through direct packet access, in order to
1302 * indicate that the hash is outdated and to trigger a
1303 * recalculation the next time the kernel tries to access this
1304 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1306 * int bpf_get_numa_node_id(void)
1308 * Return the id of the current NUMA node. The primary use case
1309 * for this helper is the selection of sockets for the local NUMA
1310 * node, when the program is attached to sockets using the
1311 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1312 * but the helper is also available to other eBPF program types,
1313 * similarly to **bpf_get_smp_processor_id**\ ().
1315 * The id of current NUMA node.
1317 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1319 * Grows headroom of packet associated to *skb* and adjusts the
1320 * offset of the MAC header accordingly, adding *len* bytes of
1321 * space. It automatically extends and reallocates memory as
1324 * This helper can be used on a layer 3 *skb* to push a MAC header
1325 * for redirection into a layer 2 device.
1327 * All values for *flags* are reserved for future usage, and must
1330 * A call to this helper is susceptible to change the underlaying
1331 * packet buffer. Therefore, at load time, all checks on pointers
1332 * previously done by the verifier are invalidated and must be
1333 * performed again, if the helper is used in combination with
1334 * direct packet access.
1336 * 0 on success, or a negative error in case of failure.
1338 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1340 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1341 * it is possible to use a negative value for *delta*. This helper
1342 * can be used to prepare the packet for pushing or popping
1345 * A call to this helper is susceptible to change the underlaying
1346 * packet buffer. Therefore, at load time, all checks on pointers
1347 * previously done by the verifier are invalidated and must be
1348 * performed again, if the helper is used in combination with
1349 * direct packet access.
1351 * 0 on success, or a negative error in case of failure.
1353 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1355 * Copy a NUL terminated string from an unsafe address
1356 * *unsafe_ptr* to *dst*. The *size* should include the
1357 * terminating NUL byte. In case the string length is smaller than
1358 * *size*, the target is not padded with further NUL bytes. If the
1359 * string length is larger than *size*, just *size*-1 bytes are
1360 * copied and the last byte is set to NUL.
1362 * On success, the length of the copied string is returned. This
1363 * makes this helper useful in tracing programs for reading
1364 * strings, and more importantly to get its length at runtime. See
1365 * the following snippet:
1369 * SEC("kprobe/sys_open")
1370 * void bpf_sys_open(struct pt_regs *ctx)
1372 * char buf[PATHLEN]; // PATHLEN is defined to 256
1373 * int res = bpf_probe_read_str(buf, sizeof(buf),
1376 * // Consume buf, for example push it to
1377 * // userspace via bpf_perf_event_output(); we
1378 * // can use res (the string length) as event
1379 * // size, after checking its boundaries.
1382 * In comparison, using **bpf_probe_read()** helper here instead
1383 * to read the string would require to estimate the length at
1384 * compile time, and would often result in copying more memory
1387 * Another useful use case is when parsing individual process
1388 * arguments or individual environment variables navigating
1389 * *current*\ **->mm->arg_start** and *current*\
1390 * **->mm->env_start**: using this helper and the return value,
1391 * one can quickly iterate at the right offset of the memory area.
1393 * On success, the strictly positive length of the string,
1394 * including the trailing NUL character. On error, a negative
1397 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1399 * If the **struct sk_buff** pointed by *skb* has a known socket,
1400 * retrieve the cookie (generated by the kernel) of this socket.
1401 * If no cookie has been set yet, generate a new cookie. Once
1402 * generated, the socket cookie remains stable for the life of the
1403 * socket. This helper can be useful for monitoring per socket
1404 * networking traffic statistics as it provides a unique socket
1405 * identifier per namespace.
1407 * A 8-byte long non-decreasing number on success, or 0 if the
1408 * socket field is missing inside *skb*.
1410 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1412 * Equivalent to bpf_get_socket_cookie() helper that accepts
1413 * *skb*, but gets socket from **struct bpf_sock_addr** contex.
1415 * A 8-byte long non-decreasing number.
1417 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1419 * Equivalent to bpf_get_socket_cookie() helper that accepts
1420 * *skb*, but gets socket from **struct bpf_sock_ops** contex.
1422 * A 8-byte long non-decreasing number.
1424 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1426 * The owner UID of the socket associated to *skb*. If the socket
1427 * is **NULL**, or if it is not a full socket (i.e. if it is a
1428 * time-wait or a request socket instead), **overflowuid** value
1429 * is returned (note that **overflowuid** might also be the actual
1430 * UID value for the socket).
1432 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1434 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1439 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1441 * Emulate a call to **setsockopt()** on the socket associated to
1442 * *bpf_socket*, which must be a full socket. The *level* at
1443 * which the option resides and the name *optname* of the option
1444 * must be specified, see **setsockopt(2)** for more information.
1445 * The option value of length *optlen* is pointed by *optval*.
1447 * This helper actually implements a subset of **setsockopt()**.
1448 * It supports the following *level*\ s:
1450 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1451 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1452 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1453 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1454 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1455 * **TCP_BPF_SNDCWND_CLAMP**.
1456 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1457 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1459 * 0 on success, or a negative error in case of failure.
1461 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1463 * Grow or shrink the room for data in the packet associated to
1464 * *skb* by *len_diff*, and according to the selected *mode*.
1466 * There is a single supported mode at this time:
1468 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1469 * (room space is added or removed below the layer 3 header).
1471 * All values for *flags* are reserved for future usage, and must
1474 * A call to this helper is susceptible to change the underlaying
1475 * packet buffer. Therefore, at load time, all checks on pointers
1476 * previously done by the verifier are invalidated and must be
1477 * performed again, if the helper is used in combination with
1478 * direct packet access.
1480 * 0 on success, or a negative error in case of failure.
1482 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1484 * Redirect the packet to the endpoint referenced by *map* at
1485 * index *key*. Depending on its type, this *map* can contain
1486 * references to net devices (for forwarding packets through other
1487 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1488 * but this is only implemented for native XDP (with driver
1489 * support) as of this writing).
1491 * All values for *flags* are reserved for future usage, and must
1494 * When used to redirect packets to net devices, this helper
1495 * provides a high performance increase over **bpf_redirect**\ ().
1496 * This is due to various implementation details of the underlying
1497 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1498 * () tries to send packet as a "bulk" to the device.
1500 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1502 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1504 * Redirect the packet to the socket referenced by *map* (of type
1505 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1506 * egress interfaces can be used for redirection. The
1507 * **BPF_F_INGRESS** value in *flags* is used to make the
1508 * distinction (ingress path is selected if the flag is present,
1509 * egress path otherwise). This is the only flag supported for now.
1511 * **SK_PASS** on success, or **SK_DROP** on error.
1513 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1515 * Add an entry to, or update a *map* referencing sockets. The
1516 * *skops* is used as a new value for the entry associated to
1517 * *key*. *flags* is one of:
1520 * The entry for *key* must not exist in the map.
1522 * The entry for *key* must already exist in the map.
1524 * No condition on the existence of the entry for *key*.
1526 * If the *map* has eBPF programs (parser and verdict), those will
1527 * be inherited by the socket being added. If the socket is
1528 * already attached to eBPF programs, this results in an error.
1530 * 0 on success, or a negative error in case of failure.
1532 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1534 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1535 * *delta* (which can be positive or negative). Note that this
1536 * operation modifies the address stored in *xdp_md*\ **->data**,
1537 * so the latter must be loaded only after the helper has been
1540 * The use of *xdp_md*\ **->data_meta** is optional and programs
1541 * are not required to use it. The rationale is that when the
1542 * packet is processed with XDP (e.g. as DoS filter), it is
1543 * possible to push further meta data along with it before passing
1544 * to the stack, and to give the guarantee that an ingress eBPF
1545 * program attached as a TC classifier on the same device can pick
1546 * this up for further post-processing. Since TC works with socket
1547 * buffers, it remains possible to set from XDP the **mark** or
1548 * **priority** pointers, or other pointers for the socket buffer.
1549 * Having this scratch space generic and programmable allows for
1550 * more flexibility as the user is free to store whatever meta
1553 * A call to this helper is susceptible to change the underlaying
1554 * packet buffer. Therefore, at load time, all checks on pointers
1555 * previously done by the verifier are invalidated and must be
1556 * performed again, if the helper is used in combination with
1557 * direct packet access.
1559 * 0 on success, or a negative error in case of failure.
1561 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1563 * Read the value of a perf event counter, and store it into *buf*
1564 * of size *buf_size*. This helper relies on a *map* of type
1565 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1566 * counter is selected when *map* is updated with perf event file
1567 * descriptors. The *map* is an array whose size is the number of
1568 * available CPUs, and each cell contains a value relative to one
1569 * CPU. The value to retrieve is indicated by *flags*, that
1570 * contains the index of the CPU to look up, masked with
1571 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1572 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1573 * current CPU should be retrieved.
1575 * This helper behaves in a way close to
1576 * **bpf_perf_event_read**\ () helper, save that instead of
1577 * just returning the value observed, it fills the *buf*
1578 * structure. This allows for additional data to be retrieved: in
1579 * particular, the enabled and running times (in *buf*\
1580 * **->enabled** and *buf*\ **->running**, respectively) are
1581 * copied. In general, **bpf_perf_event_read_value**\ () is
1582 * recommended over **bpf_perf_event_read**\ (), which has some
1583 * ABI issues and provides fewer functionalities.
1585 * These values are interesting, because hardware PMU (Performance
1586 * Monitoring Unit) counters are limited resources. When there are
1587 * more PMU based perf events opened than available counters,
1588 * kernel will multiplex these events so each event gets certain
1589 * percentage (but not all) of the PMU time. In case that
1590 * multiplexing happens, the number of samples or counter value
1591 * will not reflect the case compared to when no multiplexing
1592 * occurs. This makes comparison between different runs difficult.
1593 * Typically, the counter value should be normalized before
1594 * comparing to other experiments. The usual normalization is done
1599 * normalized_counter = counter * t_enabled / t_running
1601 * Where t_enabled is the time enabled for event and t_running is
1602 * the time running for event since last normalization. The
1603 * enabled and running times are accumulated since the perf event
1604 * open. To achieve scaling factor between two invocations of an
1605 * eBPF program, users can can use CPU id as the key (which is
1606 * typical for perf array usage model) to remember the previous
1607 * value and do the calculation inside the eBPF program.
1609 * 0 on success, or a negative error in case of failure.
1611 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1613 * For en eBPF program attached to a perf event, retrieve the
1614 * value of the event counter associated to *ctx* and store it in
1615 * the structure pointed by *buf* and of size *buf_size*. Enabled
1616 * and running times are also stored in the structure (see
1617 * description of helper **bpf_perf_event_read_value**\ () for
1620 * 0 on success, or a negative error in case of failure.
1622 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1624 * Emulate a call to **getsockopt()** on the socket associated to
1625 * *bpf_socket*, which must be a full socket. The *level* at
1626 * which the option resides and the name *optname* of the option
1627 * must be specified, see **getsockopt(2)** for more information.
1628 * The retrieved value is stored in the structure pointed by
1629 * *opval* and of length *optlen*.
1631 * This helper actually implements a subset of **getsockopt()**.
1632 * It supports the following *level*\ s:
1634 * * **IPPROTO_TCP**, which supports *optname*
1635 * **TCP_CONGESTION**.
1636 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1637 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1639 * 0 on success, or a negative error in case of failure.
1641 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1643 * Used for error injection, this helper uses kprobes to override
1644 * the return value of the probed function, and to set it to *rc*.
1645 * The first argument is the context *regs* on which the kprobe
1648 * This helper works by setting setting the PC (program counter)
1649 * to an override function which is run in place of the original
1650 * probed function. This means the probed function is not run at
1651 * all. The replacement function just returns with the required
1654 * This helper has security implications, and thus is subject to
1655 * restrictions. It is only available if the kernel was compiled
1656 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1657 * option, and in this case it only works on functions tagged with
1658 * **ALLOW_ERROR_INJECTION** in the kernel code.
1660 * Also, the helper is only available for the architectures having
1661 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1662 * x86 architecture is the only one to support this feature.
1666 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1668 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1669 * for the full TCP socket associated to *bpf_sock_ops* to
1672 * The primary use of this field is to determine if there should
1673 * be calls to eBPF programs of type
1674 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1675 * code. A program of the same type can change its value, per
1676 * connection and as necessary, when the connection is
1677 * established. This field is directly accessible for reading, but
1678 * this helper must be used for updates in order to return an
1679 * error if an eBPF program tries to set a callback that is not
1680 * supported in the current kernel.
1682 * The supported callback values that *argval* can combine are:
1684 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1685 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1686 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1688 * Here are some examples of where one could call such eBPF
1692 * * When a packet is retransmitted.
1693 * * When the connection terminates.
1694 * * When a packet is sent.
1695 * * When a packet is received.
1697 * Code **-EINVAL** if the socket is not a full TCP socket;
1698 * otherwise, a positive number containing the bits that could not
1699 * be set is returned (which comes down to 0 if all bits were set
1702 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1704 * This helper is used in programs implementing policies at the
1705 * socket level. If the message *msg* is allowed to pass (i.e. if
1706 * the verdict eBPF program returns **SK_PASS**), redirect it to
1707 * the socket referenced by *map* (of type
1708 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1709 * egress interfaces can be used for redirection. The
1710 * **BPF_F_INGRESS** value in *flags* is used to make the
1711 * distinction (ingress path is selected if the flag is present,
1712 * egress path otherwise). This is the only flag supported for now.
1714 * **SK_PASS** on success, or **SK_DROP** on error.
1716 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1718 * For socket policies, apply the verdict of the eBPF program to
1719 * the next *bytes* (number of bytes) of message *msg*.
1721 * For example, this helper can be used in the following cases:
1723 * * A single **sendmsg**\ () or **sendfile**\ () system call
1724 * contains multiple logical messages that the eBPF program is
1725 * supposed to read and for which it should apply a verdict.
1726 * * An eBPF program only cares to read the first *bytes* of a
1727 * *msg*. If the message has a large payload, then setting up
1728 * and calling the eBPF program repeatedly for all bytes, even
1729 * though the verdict is already known, would create unnecessary
1732 * When called from within an eBPF program, the helper sets a
1733 * counter internal to the BPF infrastructure, that is used to
1734 * apply the last verdict to the next *bytes*. If *bytes* is
1735 * smaller than the current data being processed from a
1736 * **sendmsg**\ () or **sendfile**\ () system call, the first
1737 * *bytes* will be sent and the eBPF program will be re-run with
1738 * the pointer for start of data pointing to byte number *bytes*
1739 * **+ 1**. If *bytes* is larger than the current data being
1740 * processed, then the eBPF verdict will be applied to multiple
1741 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1744 * Note that if a socket closes with the internal counter holding
1745 * a non-zero value, this is not a problem because data is not
1746 * being buffered for *bytes* and is sent as it is received.
1750 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1752 * For socket policies, prevent the execution of the verdict eBPF
1753 * program for message *msg* until *bytes* (byte number) have been
1756 * This can be used when one needs a specific number of bytes
1757 * before a verdict can be assigned, even if the data spans
1758 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1759 * case would be a user calling **sendmsg**\ () repeatedly with
1760 * 1-byte long message segments. Obviously, this is bad for
1761 * performance, but it is still valid. If the eBPF program needs
1762 * *bytes* bytes to validate a header, this helper can be used to
1763 * prevent the eBPF program to be called again until *bytes* have
1768 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1770 * For socket policies, pull in non-linear data from user space
1771 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1772 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1775 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1776 * *msg* it can only parse data that the (**data**, **data_end**)
1777 * pointers have already consumed. For **sendmsg**\ () hooks this
1778 * is likely the first scatterlist element. But for calls relying
1779 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1780 * be the range (**0**, **0**) because the data is shared with
1781 * user space and by default the objective is to avoid allowing
1782 * user space to modify data while (or after) eBPF verdict is
1783 * being decided. This helper can be used to pull in data and to
1784 * set the start and end pointer to given values. Data will be
1785 * copied if necessary (i.e. if data was not linear and if start
1786 * and end pointers do not point to the same chunk).
1788 * A call to this helper is susceptible to change the underlaying
1789 * packet buffer. Therefore, at load time, all checks on pointers
1790 * previously done by the verifier are invalidated and must be
1791 * performed again, if the helper is used in combination with
1792 * direct packet access.
1794 * All values for *flags* are reserved for future usage, and must
1797 * 0 on success, or a negative error in case of failure.
1799 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1801 * Bind the socket associated to *ctx* to the address pointed by
1802 * *addr*, of length *addr_len*. This allows for making outgoing
1803 * connection from the desired IP address, which can be useful for
1804 * example when all processes inside a cgroup should use one
1805 * single IP address on a host that has multiple IP configured.
1807 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1808 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1809 * **AF_INET6**). Looking for a free port to bind to can be
1810 * expensive, therefore binding to port is not permitted by the
1811 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1812 * must be set to zero.
1814 * 0 on success, or a negative error in case of failure.
1816 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1818 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1819 * only possible to shrink the packet as of this writing,
1820 * therefore *delta* must be a negative integer.
1822 * A call to this helper is susceptible to change the underlaying
1823 * packet buffer. Therefore, at load time, all checks on pointers
1824 * previously done by the verifier are invalidated and must be
1825 * performed again, if the helper is used in combination with
1826 * direct packet access.
1828 * 0 on success, or a negative error in case of failure.
1830 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1832 * Retrieve the XFRM state (IP transform framework, see also
1833 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1835 * The retrieved value is stored in the **struct bpf_xfrm_state**
1836 * pointed by *xfrm_state* and of length *size*.
1838 * All values for *flags* are reserved for future usage, and must
1841 * This helper is available only if the kernel was compiled with
1842 * **CONFIG_XFRM** configuration option.
1844 * 0 on success, or a negative error in case of failure.
1846 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1848 * Return a user or a kernel stack in bpf program provided buffer.
1849 * To achieve this, the helper needs *ctx*, which is a pointer
1850 * to the context on which the tracing program is executed.
1851 * To store the stacktrace, the bpf program provides *buf* with
1852 * a nonnegative *size*.
1854 * The last argument, *flags*, holds the number of stack frames to
1855 * skip (from 0 to 255), masked with
1856 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1857 * the following flags:
1859 * **BPF_F_USER_STACK**
1860 * Collect a user space stack instead of a kernel stack.
1861 * **BPF_F_USER_BUILD_ID**
1862 * Collect buildid+offset instead of ips for user stack,
1863 * only valid if **BPF_F_USER_STACK** is also specified.
1865 * **bpf_get_stack**\ () can collect up to
1866 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1867 * to sufficient large buffer size. Note that
1868 * this limit can be controlled with the **sysctl** program, and
1869 * that it should be manually increased in order to profile long
1870 * user stacks (such as stacks for Java programs). To do so, use:
1874 * # sysctl kernel.perf_event_max_stack=<new value>
1876 * A non-negative value equal to or less than *size* on success,
1877 * or a negative error in case of failure.
1879 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1881 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1882 * it provides an easy way to load *len* bytes from *offset*
1883 * from the packet associated to *skb*, into the buffer pointed
1884 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1885 * a fifth argument *start_header* exists in order to select a
1886 * base offset to start from. *start_header* can be one of:
1888 * **BPF_HDR_START_MAC**
1889 * Base offset to load data from is *skb*'s mac header.
1890 * **BPF_HDR_START_NET**
1891 * Base offset to load data from is *skb*'s network header.
1893 * In general, "direct packet access" is the preferred method to
1894 * access packet data, however, this helper is in particular useful
1895 * in socket filters where *skb*\ **->data** does not always point
1896 * to the start of the mac header and where "direct packet access"
1899 * 0 on success, or a negative error in case of failure.
1901 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1903 * Do FIB lookup in kernel tables using parameters in *params*.
1904 * If lookup is successful and result shows packet is to be
1905 * forwarded, the neighbor tables are searched for the nexthop.
1906 * If successful (ie., FIB lookup shows forwarding and nexthop
1907 * is resolved), the nexthop address is returned in ipv4_dst
1908 * or ipv6_dst based on family, smac is set to mac address of
1909 * egress device, dmac is set to nexthop mac address, rt_metric
1910 * is set to metric from route (IPv4/IPv6 only), and ifindex
1911 * is set to the device index of the nexthop from the FIB lookup.
1913 * *plen* argument is the size of the passed in struct.
1914 * *flags* argument can be a combination of one or more of the
1917 * **BPF_FIB_LOOKUP_DIRECT**
1918 * Do a direct table lookup vs full lookup using FIB
1920 * **BPF_FIB_LOOKUP_OUTPUT**
1921 * Perform lookup from an egress perspective (default is
1924 * *ctx* is either **struct xdp_md** for XDP programs or
1925 * **struct sk_buff** tc cls_act programs.
1927 * * < 0 if any input argument is invalid
1928 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1929 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1930 * packet is not forwarded or needs assist from full stack
1932 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1934 * Add an entry to, or update a sockhash *map* referencing sockets.
1935 * The *skops* is used as a new value for the entry associated to
1936 * *key*. *flags* is one of:
1939 * The entry for *key* must not exist in the map.
1941 * The entry for *key* must already exist in the map.
1943 * No condition on the existence of the entry for *key*.
1945 * If the *map* has eBPF programs (parser and verdict), those will
1946 * be inherited by the socket being added. If the socket is
1947 * already attached to eBPF programs, this results in an error.
1949 * 0 on success, or a negative error in case of failure.
1951 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1953 * This helper is used in programs implementing policies at the
1954 * socket level. If the message *msg* is allowed to pass (i.e. if
1955 * the verdict eBPF program returns **SK_PASS**), redirect it to
1956 * the socket referenced by *map* (of type
1957 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1958 * egress interfaces can be used for redirection. The
1959 * **BPF_F_INGRESS** value in *flags* is used to make the
1960 * distinction (ingress path is selected if the flag is present,
1961 * egress path otherwise). This is the only flag supported for now.
1963 * **SK_PASS** on success, or **SK_DROP** on error.
1965 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1967 * This helper is used in programs implementing policies at the
1968 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1969 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1970 * to the socket referenced by *map* (of type
1971 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1972 * egress interfaces can be used for redirection. The
1973 * **BPF_F_INGRESS** value in *flags* is used to make the
1974 * distinction (ingress path is selected if the flag is present,
1975 * egress otherwise). This is the only flag supported for now.
1977 * **SK_PASS** on success, or **SK_DROP** on error.
1979 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
1981 * Encapsulate the packet associated to *skb* within a Layer 3
1982 * protocol header. This header is provided in the buffer at
1983 * address *hdr*, with *len* its size in bytes. *type* indicates
1984 * the protocol of the header and can be one of:
1986 * **BPF_LWT_ENCAP_SEG6**
1987 * IPv6 encapsulation with Segment Routing Header
1988 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
1989 * the IPv6 header is computed by the kernel.
1990 * **BPF_LWT_ENCAP_SEG6_INLINE**
1991 * Only works if *skb* contains an IPv6 packet. Insert a
1992 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
1995 * A call to this helper is susceptible to change the underlaying
1996 * packet buffer. Therefore, at load time, all checks on pointers
1997 * previously done by the verifier are invalidated and must be
1998 * performed again, if the helper is used in combination with
1999 * direct packet access.
2001 * 0 on success, or a negative error in case of failure.
2003 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2005 * Store *len* bytes from address *from* into the packet
2006 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2007 * inside the outermost IPv6 Segment Routing Header can be
2008 * modified through this helper.
2010 * A call to this helper is susceptible to change the underlaying
2011 * packet buffer. Therefore, at load time, all checks on pointers
2012 * previously done by the verifier are invalidated and must be
2013 * performed again, if the helper is used in combination with
2014 * direct packet access.
2016 * 0 on success, or a negative error in case of failure.
2018 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2020 * Adjust the size allocated to TLVs in the outermost IPv6
2021 * Segment Routing Header contained in the packet associated to
2022 * *skb*, at position *offset* by *delta* bytes. Only offsets
2023 * after the segments are accepted. *delta* can be as well
2024 * positive (growing) as negative (shrinking).
2026 * A call to this helper is susceptible to change the underlaying
2027 * packet buffer. Therefore, at load time, all checks on pointers
2028 * previously done by the verifier are invalidated and must be
2029 * performed again, if the helper is used in combination with
2030 * direct packet access.
2032 * 0 on success, or a negative error in case of failure.
2034 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2036 * Apply an IPv6 Segment Routing action of type *action* to the
2037 * packet associated to *skb*. Each action takes a parameter
2038 * contained at address *param*, and of length *param_len* bytes.
2039 * *action* can be one of:
2041 * **SEG6_LOCAL_ACTION_END_X**
2042 * End.X action: Endpoint with Layer-3 cross-connect.
2043 * Type of *param*: **struct in6_addr**.
2044 * **SEG6_LOCAL_ACTION_END_T**
2045 * End.T action: Endpoint with specific IPv6 table lookup.
2046 * Type of *param*: **int**.
2047 * **SEG6_LOCAL_ACTION_END_B6**
2048 * End.B6 action: Endpoint bound to an SRv6 policy.
2049 * Type of param: **struct ipv6_sr_hdr**.
2050 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2051 * End.B6.Encap action: Endpoint bound to an SRv6
2052 * encapsulation policy.
2053 * Type of param: **struct ipv6_sr_hdr**.
2055 * A call to this helper is susceptible to change the underlaying
2056 * packet buffer. Therefore, at load time, all checks on pointers
2057 * previously done by the verifier are invalidated and must be
2058 * performed again, if the helper is used in combination with
2059 * direct packet access.
2061 * 0 on success, or a negative error in case of failure.
2063 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2065 * This helper is used in programs implementing IR decoding, to
2066 * report a successfully decoded key press with *scancode*,
2067 * *toggle* value in the given *protocol*. The scancode will be
2068 * translated to a keycode using the rc keymap, and reported as
2069 * an input key down event. After a period a key up event is
2070 * generated. This period can be extended by calling either
2071 * **bpf_rc_keydown** () again with the same values, or calling
2072 * **bpf_rc_repeat** ().
2074 * Some protocols include a toggle bit, in case the button was
2075 * released and pressed again between consecutive scancodes.
2077 * The *ctx* should point to the lirc sample as passed into
2080 * The *protocol* is the decoded protocol number (see
2081 * **enum rc_proto** for some predefined values).
2083 * This helper is only available is the kernel was compiled with
2084 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2089 * int bpf_rc_repeat(void *ctx)
2091 * This helper is used in programs implementing IR decoding, to
2092 * report a successfully decoded repeat key message. This delays
2093 * the generation of a key up event for previously generated
2096 * Some IR protocols like NEC have a special IR message for
2097 * repeating last button, for when a button is held down.
2099 * The *ctx* should point to the lirc sample as passed into
2102 * This helper is only available is the kernel was compiled with
2103 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2108 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2110 * Return the cgroup v2 id of the socket associated with the *skb*.
2111 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2112 * helper for cgroup v1 by providing a tag resp. identifier that
2113 * can be matched on or used for map lookups e.g. to implement
2114 * policy. The cgroup v2 id of a given path in the hierarchy is
2115 * exposed in user space through the f_handle API in order to get
2116 * to the same 64-bit id.
2118 * This helper can be used on TC egress path, but not on ingress,
2119 * and is available only if the kernel was compiled with the
2120 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2122 * The id is returned or 0 in case the id could not be retrieved.
2124 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2126 * Return id of cgroup v2 that is ancestor of cgroup associated
2127 * with the *skb* at the *ancestor_level*. The root cgroup is at
2128 * *ancestor_level* zero and each step down the hierarchy
2129 * increments the level. If *ancestor_level* == level of cgroup
2130 * associated with *skb*, then return value will be same as that
2131 * of **bpf_skb_cgroup_id**\ ().
2133 * The helper is useful to implement policies based on cgroups
2134 * that are upper in hierarchy than immediate cgroup associated
2137 * The format of returned id and helper limitations are same as in
2138 * **bpf_skb_cgroup_id**\ ().
2140 * The id is returned or 0 in case the id could not be retrieved.
2142 * u64 bpf_get_current_cgroup_id(void)
2144 * A 64-bit integer containing the current cgroup id based
2145 * on the cgroup within which the current task is running.
2147 * void* get_local_storage(void *map, u64 flags)
2149 * Get the pointer to the local storage area.
2150 * The type and the size of the local storage is defined
2151 * by the *map* argument.
2152 * The *flags* meaning is specific for each map type,
2153 * and has to be 0 for cgroup local storage.
2155 * Depending on the bpf program type, a local storage area
2156 * can be shared between multiple instances of the bpf program,
2157 * running simultaneously.
2159 * A user should care about the synchronization by himself.
2160 * For example, by using the BPF_STX_XADD instruction to alter
2163 * Pointer to the local storage area.
2165 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2167 * Select a SO_REUSEPORT sk from a BPF_MAP_TYPE_REUSEPORT_ARRAY map
2168 * It checks the selected sk is matching the incoming
2169 * request in the skb.
2171 * 0 on success, or a negative error in case of failure.
2173 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2175 * Look for TCP socket matching *tuple*, optionally in a child
2176 * network namespace *netns*. The return value must be checked,
2177 * and if non-NULL, released via **bpf_sk_release**\ ().
2179 * The *ctx* should point to the context of the program, such as
2180 * the skb or socket (depending on the hook in use). This is used
2181 * to determine the base network namespace for the lookup.
2183 * *tuple_size* must be one of:
2185 * **sizeof**\ (*tuple*\ **->ipv4**)
2186 * Look for an IPv4 socket.
2187 * **sizeof**\ (*tuple*\ **->ipv6**)
2188 * Look for an IPv6 socket.
2190 * If the *netns* is zero, then the socket lookup table in the
2191 * netns associated with the *ctx* will be used. For the TC hooks,
2192 * this in the netns of the device in the skb. For socket hooks,
2193 * this in the netns of the socket. If *netns* is non-zero, then
2194 * it specifies the ID of the netns relative to the netns
2195 * associated with the *ctx*.
2197 * All values for *flags* are reserved for future usage, and must
2200 * This helper is available only if the kernel was compiled with
2201 * **CONFIG_NET** configuration option.
2203 * Pointer to *struct bpf_sock*, or NULL in case of failure.
2205 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2207 * Look for UDP socket matching *tuple*, optionally in a child
2208 * network namespace *netns*. The return value must be checked,
2209 * and if non-NULL, released via **bpf_sk_release**\ ().
2211 * The *ctx* should point to the context of the program, such as
2212 * the skb or socket (depending on the hook in use). This is used
2213 * to determine the base network namespace for the lookup.
2215 * *tuple_size* must be one of:
2217 * **sizeof**\ (*tuple*\ **->ipv4**)
2218 * Look for an IPv4 socket.
2219 * **sizeof**\ (*tuple*\ **->ipv6**)
2220 * Look for an IPv6 socket.
2222 * If the *netns* is zero, then the socket lookup table in the
2223 * netns associated with the *ctx* will be used. For the TC hooks,
2224 * this in the netns of the device in the skb. For socket hooks,
2225 * this in the netns of the socket. If *netns* is non-zero, then
2226 * it specifies the ID of the netns relative to the netns
2227 * associated with the *ctx*.
2229 * All values for *flags* are reserved for future usage, and must
2232 * This helper is available only if the kernel was compiled with
2233 * **CONFIG_NET** configuration option.
2235 * Pointer to *struct bpf_sock*, or NULL in case of failure.
2237 * int bpf_sk_release(struct bpf_sock *sk)
2239 * Release the reference held by *sock*. *sock* must be a non-NULL
2240 * pointer that was returned from bpf_sk_lookup_xxx\ ().
2242 * 0 on success, or a negative error in case of failure.
2244 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2246 * For socket policies, insert *len* bytes into msg at offset
2249 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2250 * *msg* it may want to insert metadata or options into the msg.
2251 * This can later be read and used by any of the lower layer BPF
2254 * This helper may fail if under memory pressure (a malloc
2255 * fails) in these cases BPF programs will get an appropriate
2256 * error and BPF programs will need to handle them.
2259 * 0 on success, or a negative error in case of failure.
2261 #define __BPF_FUNC_MAPPER(FN) \
2263 FN(map_lookup_elem), \
2264 FN(map_update_elem), \
2265 FN(map_delete_elem), \
2269 FN(get_prandom_u32), \
2270 FN(get_smp_processor_id), \
2271 FN(skb_store_bytes), \
2272 FN(l3_csum_replace), \
2273 FN(l4_csum_replace), \
2275 FN(clone_redirect), \
2276 FN(get_current_pid_tgid), \
2277 FN(get_current_uid_gid), \
2278 FN(get_current_comm), \
2279 FN(get_cgroup_classid), \
2280 FN(skb_vlan_push), \
2282 FN(skb_get_tunnel_key), \
2283 FN(skb_set_tunnel_key), \
2284 FN(perf_event_read), \
2286 FN(get_route_realm), \
2287 FN(perf_event_output), \
2288 FN(skb_load_bytes), \
2291 FN(skb_get_tunnel_opt), \
2292 FN(skb_set_tunnel_opt), \
2293 FN(skb_change_proto), \
2294 FN(skb_change_type), \
2295 FN(skb_under_cgroup), \
2296 FN(get_hash_recalc), \
2297 FN(get_current_task), \
2298 FN(probe_write_user), \
2299 FN(current_task_under_cgroup), \
2300 FN(skb_change_tail), \
2301 FN(skb_pull_data), \
2303 FN(set_hash_invalid), \
2304 FN(get_numa_node_id), \
2305 FN(skb_change_head), \
2306 FN(xdp_adjust_head), \
2307 FN(probe_read_str), \
2308 FN(get_socket_cookie), \
2309 FN(get_socket_uid), \
2312 FN(skb_adjust_room), \
2314 FN(sk_redirect_map), \
2315 FN(sock_map_update), \
2316 FN(xdp_adjust_meta), \
2317 FN(perf_event_read_value), \
2318 FN(perf_prog_read_value), \
2320 FN(override_return), \
2321 FN(sock_ops_cb_flags_set), \
2322 FN(msg_redirect_map), \
2323 FN(msg_apply_bytes), \
2324 FN(msg_cork_bytes), \
2325 FN(msg_pull_data), \
2327 FN(xdp_adjust_tail), \
2328 FN(skb_get_xfrm_state), \
2330 FN(skb_load_bytes_relative), \
2332 FN(sock_hash_update), \
2333 FN(msg_redirect_hash), \
2334 FN(sk_redirect_hash), \
2335 FN(lwt_push_encap), \
2336 FN(lwt_seg6_store_bytes), \
2337 FN(lwt_seg6_adjust_srh), \
2338 FN(lwt_seg6_action), \
2341 FN(skb_cgroup_id), \
2342 FN(get_current_cgroup_id), \
2343 FN(get_local_storage), \
2344 FN(sk_select_reuseport), \
2345 FN(skb_ancestor_cgroup_id), \
2346 FN(sk_lookup_tcp), \
2347 FN(sk_lookup_udp), \
2349 FN(map_push_elem), \
2351 FN(map_peek_elem), \
2354 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2355 * function eBPF program intends to call
2357 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2359 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2362 #undef __BPF_ENUM_FN
2364 /* All flags used by eBPF helper functions, placed here. */
2366 /* BPF_FUNC_skb_store_bytes flags. */
2367 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2368 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2370 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2371 * First 4 bits are for passing the header field size.
2373 #define BPF_F_HDR_FIELD_MASK 0xfULL
2375 /* BPF_FUNC_l4_csum_replace flags. */
2376 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2377 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2378 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2380 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2381 #define BPF_F_INGRESS (1ULL << 0)
2383 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2384 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2386 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2387 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2388 #define BPF_F_USER_STACK (1ULL << 8)
2389 /* flags used by BPF_FUNC_get_stackid only. */
2390 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2391 #define BPF_F_REUSE_STACKID (1ULL << 10)
2392 /* flags used by BPF_FUNC_get_stack only. */
2393 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2395 /* BPF_FUNC_skb_set_tunnel_key flags. */
2396 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2397 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2398 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2400 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2401 * BPF_FUNC_perf_event_read_value flags.
2403 #define BPF_F_INDEX_MASK 0xffffffffULL
2404 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2405 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2406 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2408 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2409 enum bpf_adj_room_mode {
2413 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2414 enum bpf_hdr_start_off {
2419 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2420 enum bpf_lwt_encap_mode {
2422 BPF_LWT_ENCAP_SEG6_INLINE
2425 /* user accessible mirror of in-kernel sk_buff.
2426 * new fields can only be added to the end of this structure
2432 __u32 queue_mapping;
2438 __u32 ingress_ifindex;
2448 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2450 __u32 remote_ip4; /* Stored in network byte order */
2451 __u32 local_ip4; /* Stored in network byte order */
2452 __u32 remote_ip6[4]; /* Stored in network byte order */
2453 __u32 local_ip6[4]; /* Stored in network byte order */
2454 __u32 remote_port; /* Stored in network byte order */
2455 __u32 local_port; /* stored in host byte order */
2459 struct bpf_flow_keys *flow_keys;
2462 struct bpf_tunnel_key {
2466 __u32 remote_ipv6[4];
2470 __u16 tunnel_ext; /* Padding, future use. */
2474 /* user accessible mirror of in-kernel xfrm_state.
2475 * new fields can only be added to the end of this structure
2477 struct bpf_xfrm_state {
2479 __u32 spi; /* Stored in network byte order */
2481 __u16 ext; /* Padding, future use. */
2483 __u32 remote_ipv4; /* Stored in network byte order */
2484 __u32 remote_ipv6[4]; /* Stored in network byte order */
2488 /* Generic BPF return codes which all BPF program types may support.
2489 * The values are binary compatible with their TC_ACT_* counter-part to
2490 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2493 * XDP is handled seprately, see XDP_*.
2501 /* >127 are reserved for prog type specific return codes */
2511 __u32 src_ip4; /* Allows 1,2,4-byte read.
2512 * Stored in network byte order.
2514 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2515 * Stored in network byte order.
2517 __u32 src_port; /* Allows 4-byte read.
2518 * Stored in host byte order
2522 struct bpf_sock_tuple {
2539 #define XDP_PACKET_HEADROOM 256
2541 /* User return codes for XDP prog type.
2542 * A valid XDP program must return one of these defined values. All other
2543 * return codes are reserved for future use. Unknown return codes will
2544 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2554 /* user accessible metadata for XDP packet hook
2555 * new fields must be added to the end of this structure
2561 /* Below access go through struct xdp_rxq_info */
2562 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2563 __u32 rx_queue_index; /* rxq->queue_index */
2571 /* user accessible metadata for SK_MSG packet hook, new fields must
2572 * be added to the end of this structure
2579 __u32 remote_ip4; /* Stored in network byte order */
2580 __u32 local_ip4; /* Stored in network byte order */
2581 __u32 remote_ip6[4]; /* Stored in network byte order */
2582 __u32 local_ip6[4]; /* Stored in network byte order */
2583 __u32 remote_port; /* Stored in network byte order */
2584 __u32 local_port; /* stored in host byte order */
2587 struct sk_reuseport_md {
2589 * Start of directly accessible data. It begins from
2590 * the tcp/udp header.
2593 void *data_end; /* End of directly accessible data */
2595 * Total length of packet (starting from the tcp/udp header).
2596 * Note that the directly accessible bytes (data_end - data)
2597 * could be less than this "len". Those bytes could be
2598 * indirectly read by a helper "bpf_skb_load_bytes()".
2602 * Eth protocol in the mac header (network byte order). e.g.
2603 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2606 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2607 __u32 bind_inany; /* Is sock bound to an INANY address? */
2608 __u32 hash; /* A hash of the packet 4 tuples */
2611 #define BPF_TAG_SIZE 8
2613 struct bpf_prog_info {
2616 __u8 tag[BPF_TAG_SIZE];
2617 __u32 jited_prog_len;
2618 __u32 xlated_prog_len;
2619 __aligned_u64 jited_prog_insns;
2620 __aligned_u64 xlated_prog_insns;
2621 __u64 load_time; /* ns since boottime */
2622 __u32 created_by_uid;
2624 __aligned_u64 map_ids;
2625 char name[BPF_OBJ_NAME_LEN];
2627 __u32 gpl_compatible:1;
2630 __u32 nr_jited_ksyms;
2631 __u32 nr_jited_func_lens;
2632 __aligned_u64 jited_ksyms;
2633 __aligned_u64 jited_func_lens;
2634 } __attribute__((aligned(8)));
2636 struct bpf_map_info {
2643 char name[BPF_OBJ_NAME_LEN];
2649 __u32 btf_key_type_id;
2650 __u32 btf_value_type_id;
2651 } __attribute__((aligned(8)));
2653 struct bpf_btf_info {
2657 } __attribute__((aligned(8)));
2659 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2660 * by user and intended to be used by socket (e.g. to bind to, depends on
2661 * attach attach type).
2663 struct bpf_sock_addr {
2664 __u32 user_family; /* Allows 4-byte read, but no write. */
2665 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2666 * Stored in network byte order.
2668 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2669 * Stored in network byte order.
2671 __u32 user_port; /* Allows 4-byte read and write.
2672 * Stored in network byte order
2674 __u32 family; /* Allows 4-byte read, but no write */
2675 __u32 type; /* Allows 4-byte read, but no write */
2676 __u32 protocol; /* Allows 4-byte read, but no write */
2677 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2678 * Stored in network byte order.
2680 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2681 * Stored in network byte order.
2685 /* User bpf_sock_ops struct to access socket values and specify request ops
2686 * and their replies.
2687 * Some of this fields are in network (bigendian) byte order and may need
2688 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2689 * New fields can only be added at the end of this structure
2691 struct bpf_sock_ops {
2694 __u32 args[4]; /* Optionally passed to bpf program */
2695 __u32 reply; /* Returned by bpf program */
2696 __u32 replylong[4]; /* Optionally returned by bpf prog */
2699 __u32 remote_ip4; /* Stored in network byte order */
2700 __u32 local_ip4; /* Stored in network byte order */
2701 __u32 remote_ip6[4]; /* Stored in network byte order */
2702 __u32 local_ip6[4]; /* Stored in network byte order */
2703 __u32 remote_port; /* Stored in network byte order */
2704 __u32 local_port; /* stored in host byte order */
2705 __u32 is_fullsock; /* Some TCP fields are only valid if
2706 * there is a full socket. If not, the
2707 * fields read as zero.
2710 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2711 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2720 __u32 rate_delivered;
2721 __u32 rate_interval_us;
2724 __u32 total_retrans;
2728 __u32 data_segs_out;
2732 __u64 bytes_received;
2736 /* Definitions for bpf_sock_ops_cb_flags */
2737 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2738 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2739 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2740 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2741 * supported cb flags
2744 /* List of known BPF sock_ops operators.
2745 * New entries can only be added at the end
2749 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2750 * -1 if default value should be used
2752 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2753 * window (in packets) or -1 if default
2754 * value should be used
2756 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2757 * active connection is initialized
2759 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2760 * active connection is
2763 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2764 * passive connection is
2767 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2770 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2771 * based on the path and may be
2772 * dependent on the congestion control
2773 * algorithm. In general it indicates
2774 * a congestion threshold. RTTs above
2775 * this indicate congestion
2777 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2778 * Arg1: value of icsk_retransmits
2779 * Arg2: value of icsk_rto
2780 * Arg3: whether RTO has expired
2782 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2783 * Arg1: sequence number of 1st byte
2785 * Arg3: return value of
2786 * tcp_transmit_skb (0 => success)
2788 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2792 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
2793 * socket transition to LISTEN state.
2797 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2798 * changes between the TCP and BPF versions. Ideally this should never happen.
2799 * If it does, we need to add code to convert them before calling
2800 * the BPF sock_ops function.
2803 BPF_TCP_ESTABLISHED = 1,
2813 BPF_TCP_CLOSING, /* Now a valid state */
2814 BPF_TCP_NEW_SYN_RECV,
2816 BPF_TCP_MAX_STATES /* Leave at the end! */
2819 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
2820 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
2822 struct bpf_perf_event_value {
2828 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2829 #define BPF_DEVCG_ACC_READ (1ULL << 1)
2830 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2832 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2833 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2835 struct bpf_cgroup_dev_ctx {
2836 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2842 struct bpf_raw_tracepoint_args {
2846 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
2847 * OUTPUT: Do lookup from egress perspective; default is ingress
2849 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
2850 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
2853 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
2854 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
2855 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
2856 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
2857 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
2858 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
2859 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
2860 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
2861 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
2864 struct bpf_fib_lookup {
2865 /* input: network family for lookup (AF_INET, AF_INET6)
2866 * output: network family of egress nexthop
2870 /* set if lookup is to consider L4 data - e.g., FIB rules */
2875 /* total length of packet from network header - used for MTU check */
2878 /* input: L3 device index for lookup
2879 * output: device index from FIB lookup
2884 /* inputs to lookup */
2885 __u8 tos; /* AF_INET */
2886 __be32 flowinfo; /* AF_INET6, flow_label + priority */
2888 /* output: metric of fib result (IPv4/IPv6 only) */
2894 __u32 ipv6_src[4]; /* in6_addr; network order */
2897 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2898 * network header. output: bpf_fib_lookup sets to gateway address
2899 * if FIB lookup returns gateway route
2903 __u32 ipv6_dst[4]; /* in6_addr; network order */
2907 __be16 h_vlan_proto;
2909 __u8 smac[6]; /* ETH_ALEN */
2910 __u8 dmac[6]; /* ETH_ALEN */
2913 enum bpf_task_fd_type {
2914 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
2915 BPF_FD_TYPE_TRACEPOINT, /* tp name */
2916 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
2917 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
2918 BPF_FD_TYPE_UPROBE, /* filename + offset */
2919 BPF_FD_TYPE_URETPROBE, /* filename + offset */
2922 struct bpf_flow_keys {
2925 __u16 addr_proto; /* ETH_P_* of valid addrs */
2939 __u32 ipv6_src[4]; /* in6_addr; network order */
2940 __u32 ipv6_dst[4]; /* in6_addr; network order */
2945 #endif /* _UAPI__LINUX_BPF_H__ */