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,
112 BPF_MAP_TYPE_PROG_ARRAY,
113 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
114 BPF_MAP_TYPE_PERCPU_HASH,
115 BPF_MAP_TYPE_PERCPU_ARRAY,
116 BPF_MAP_TYPE_STACK_TRACE,
117 BPF_MAP_TYPE_CGROUP_ARRAY,
118 BPF_MAP_TYPE_LRU_HASH,
119 BPF_MAP_TYPE_LRU_PERCPU_HASH,
120 BPF_MAP_TYPE_LPM_TRIE,
121 BPF_MAP_TYPE_ARRAY_OF_MAPS,
122 BPF_MAP_TYPE_HASH_OF_MAPS,
124 BPF_MAP_TYPE_SOCKMAP,
127 BPF_MAP_TYPE_SOCKHASH,
128 BPF_MAP_TYPE_CGROUP_STORAGE,
129 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
133 BPF_PROG_TYPE_UNSPEC,
134 BPF_PROG_TYPE_SOCKET_FILTER,
135 BPF_PROG_TYPE_KPROBE,
136 BPF_PROG_TYPE_SCHED_CLS,
137 BPF_PROG_TYPE_SCHED_ACT,
138 BPF_PROG_TYPE_TRACEPOINT,
140 BPF_PROG_TYPE_PERF_EVENT,
141 BPF_PROG_TYPE_CGROUP_SKB,
142 BPF_PROG_TYPE_CGROUP_SOCK,
143 BPF_PROG_TYPE_LWT_IN,
144 BPF_PROG_TYPE_LWT_OUT,
145 BPF_PROG_TYPE_LWT_XMIT,
146 BPF_PROG_TYPE_SOCK_OPS,
147 BPF_PROG_TYPE_SK_SKB,
148 BPF_PROG_TYPE_CGROUP_DEVICE,
149 BPF_PROG_TYPE_SK_MSG,
150 BPF_PROG_TYPE_RAW_TRACEPOINT,
151 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
152 BPF_PROG_TYPE_LWT_SEG6LOCAL,
153 BPF_PROG_TYPE_LIRC_MODE2,
154 BPF_PROG_TYPE_SK_REUSEPORT,
157 enum bpf_attach_type {
158 BPF_CGROUP_INET_INGRESS,
159 BPF_CGROUP_INET_EGRESS,
160 BPF_CGROUP_INET_SOCK_CREATE,
162 BPF_SK_SKB_STREAM_PARSER,
163 BPF_SK_SKB_STREAM_VERDICT,
166 BPF_CGROUP_INET4_BIND,
167 BPF_CGROUP_INET6_BIND,
168 BPF_CGROUP_INET4_CONNECT,
169 BPF_CGROUP_INET6_CONNECT,
170 BPF_CGROUP_INET4_POST_BIND,
171 BPF_CGROUP_INET6_POST_BIND,
172 BPF_CGROUP_UDP4_SENDMSG,
173 BPF_CGROUP_UDP6_SENDMSG,
175 BPF_CGROUP_UDP4_RECVMSG = 19,
176 BPF_CGROUP_UDP6_RECVMSG,
177 __MAX_BPF_ATTACH_TYPE
180 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
182 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
184 * NONE(default): No further bpf programs allowed in the subtree.
186 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
187 * the program in this cgroup yields to sub-cgroup program.
189 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
190 * that cgroup program gets run in addition to the program in this cgroup.
192 * Only one program is allowed to be attached to a cgroup with
193 * NONE or BPF_F_ALLOW_OVERRIDE flag.
194 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
195 * release old program and attach the new one. Attach flags has to match.
197 * Multiple programs are allowed to be attached to a cgroup with
198 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
199 * (those that were attached first, run first)
200 * The programs of sub-cgroup are executed first, then programs of
201 * this cgroup and then programs of parent cgroup.
202 * When children program makes decision (like picking TCP CA or sock bind)
203 * parent program has a chance to override it.
205 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
206 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
208 * cgrp1 (MULTI progs A, B) ->
209 * cgrp2 (OVERRIDE prog C) ->
210 * cgrp3 (MULTI prog D) ->
211 * cgrp4 (OVERRIDE prog E) ->
212 * cgrp5 (NONE prog F)
213 * the event in cgrp5 triggers execution of F,D,A,B in that order.
214 * if prog F is detached, the execution is E,D,A,B
215 * if prog F and D are detached, the execution is E,A,B
216 * if prog F, E and D are detached, the execution is C,A,B
218 * All eligible programs are executed regardless of return code from
221 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
222 #define BPF_F_ALLOW_MULTI (1U << 1)
224 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
225 * verifier will perform strict alignment checking as if the kernel
226 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
227 * and NET_IP_ALIGN defined to 2.
229 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
231 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
232 * verifier will allow any alignment whatsoever. On platforms
233 * with strict alignment requirements for loads ands stores (such
234 * as sparc and mips) the verifier validates that all loads and
235 * stores provably follow this requirement. This flag turns that
236 * checking and enforcement off.
238 * It is mostly used for testing when we want to validate the
239 * context and memory access aspects of the verifier, but because
240 * of an unaligned access the alignment check would trigger before
241 * the one we are interested in.
243 #define BPF_F_ANY_ALIGNMENT (1U << 1)
245 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
246 #define BPF_PSEUDO_MAP_FD 1
248 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
249 * offset to another bpf function
251 #define BPF_PSEUDO_CALL 1
253 /* flags for BPF_MAP_UPDATE_ELEM command */
254 #define BPF_ANY 0 /* create new element or update existing */
255 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
256 #define BPF_EXIST 2 /* update existing element */
258 /* flags for BPF_MAP_CREATE command */
259 #define BPF_F_NO_PREALLOC (1U << 0)
260 /* Instead of having one common LRU list in the
261 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
262 * which can scale and perform better.
263 * Note, the LRU nodes (including free nodes) cannot be moved
264 * across different LRU lists.
266 #define BPF_F_NO_COMMON_LRU (1U << 1)
267 /* Specify numa node during map creation */
268 #define BPF_F_NUMA_NODE (1U << 2)
270 /* flags for BPF_PROG_QUERY */
271 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
273 #define BPF_OBJ_NAME_LEN 16U
275 /* Flags for accessing BPF object */
276 #define BPF_F_RDONLY (1U << 3)
277 #define BPF_F_WRONLY (1U << 4)
279 /* Flag for stack_map, store build_id+offset instead of pointer */
280 #define BPF_F_STACK_BUILD_ID (1U << 5)
282 enum bpf_stack_build_id_status {
283 /* user space need an empty entry to identify end of a trace */
284 BPF_STACK_BUILD_ID_EMPTY = 0,
285 /* with valid build_id and offset */
286 BPF_STACK_BUILD_ID_VALID = 1,
287 /* couldn't get build_id, fallback to ip */
288 BPF_STACK_BUILD_ID_IP = 2,
291 #define BPF_BUILD_ID_SIZE 20
292 struct bpf_stack_build_id {
294 unsigned char build_id[BPF_BUILD_ID_SIZE];
302 struct { /* anonymous struct used by BPF_MAP_CREATE command */
303 __u32 map_type; /* one of enum bpf_map_type */
304 __u32 key_size; /* size of key in bytes */
305 __u32 value_size; /* size of value in bytes */
306 __u32 max_entries; /* max number of entries in a map */
307 __u32 map_flags; /* BPF_MAP_CREATE related
308 * flags defined above.
310 __u32 inner_map_fd; /* fd pointing to the inner map */
311 __u32 numa_node; /* numa node (effective only if
312 * BPF_F_NUMA_NODE is set).
314 char map_name[BPF_OBJ_NAME_LEN];
315 __u32 map_ifindex; /* ifindex of netdev to create on */
316 __u32 btf_fd; /* fd pointing to a BTF type data */
317 __u32 btf_key_type_id; /* BTF type_id of the key */
318 __u32 btf_value_type_id; /* BTF type_id of the value */
321 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
326 __aligned_u64 next_key;
331 struct { /* anonymous struct used by BPF_PROG_LOAD command */
332 __u32 prog_type; /* one of enum bpf_prog_type */
335 __aligned_u64 license;
336 __u32 log_level; /* verbosity level of verifier */
337 __u32 log_size; /* size of user buffer */
338 __aligned_u64 log_buf; /* user supplied buffer */
339 __u32 kern_version; /* checked when prog_type=kprobe */
341 char prog_name[BPF_OBJ_NAME_LEN];
342 __u32 prog_ifindex; /* ifindex of netdev to prep for */
343 /* For some prog types expected attach type must be known at
344 * load time to verify attach type specific parts of prog
345 * (context accesses, allowed helpers, etc).
347 __u32 expected_attach_type;
350 struct { /* anonymous struct used by BPF_OBJ_* commands */
351 __aligned_u64 pathname;
356 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
357 __u32 target_fd; /* container object to attach to */
358 __u32 attach_bpf_fd; /* eBPF program to attach */
363 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
368 __aligned_u64 data_in;
369 __aligned_u64 data_out;
374 struct { /* anonymous struct used by BPF_*_GET_*_ID */
385 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
391 struct { /* anonymous struct used by BPF_PROG_QUERY command */
392 __u32 target_fd; /* container object to query */
396 __aligned_u64 prog_ids;
405 struct { /* anonymous struct for BPF_BTF_LOAD */
407 __aligned_u64 btf_log_buf;
414 __u32 pid; /* input: pid */
415 __u32 fd; /* input: fd */
416 __u32 flags; /* input: flags */
417 __u32 buf_len; /* input/output: buf len */
418 __aligned_u64 buf; /* input/output:
419 * tp_name for tracepoint
421 * filename for uprobe
423 __u32 prog_id; /* output: prod_id */
424 __u32 fd_type; /* output: BPF_FD_TYPE_* */
425 __u64 probe_offset; /* output: probe_offset */
426 __u64 probe_addr; /* output: probe_addr */
428 } __attribute__((aligned(8)));
430 /* The description below is an attempt at providing documentation to eBPF
431 * developers about the multiple available eBPF helper functions. It can be
432 * parsed and used to produce a manual page. The workflow is the following,
433 * and requires the rst2man utility:
435 * $ ./scripts/bpf_helpers_doc.py \
436 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
437 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
438 * $ man /tmp/bpf-helpers.7
440 * Note that in order to produce this external documentation, some RST
441 * formatting is used in the descriptions to get "bold" and "italics" in
442 * manual pages. Also note that the few trailing white spaces are
443 * intentional, removing them would break paragraphs for rst2man.
445 * Start of BPF helper function descriptions:
447 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
449 * Perform a lookup in *map* for an entry associated to *key*.
451 * Map value associated to *key*, or **NULL** if no entry was
454 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
456 * Add or update the value of the entry associated to *key* in
457 * *map* with *value*. *flags* is one of:
460 * The entry for *key* must not exist in the map.
462 * The entry for *key* must already exist in the map.
464 * No condition on the existence of the entry for *key*.
466 * Flag value **BPF_NOEXIST** cannot be used for maps of types
467 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
468 * elements always exist), the helper would return an error.
470 * 0 on success, or a negative error in case of failure.
472 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
474 * Delete entry with *key* from *map*.
476 * 0 on success, or a negative error in case of failure.
478 * int bpf_probe_read(void *dst, u32 size, const void *src)
480 * For tracing programs, safely attempt to read *size* bytes from
481 * address *src* and store the data in *dst*.
483 * 0 on success, or a negative error in case of failure.
485 * u64 bpf_ktime_get_ns(void)
487 * Return the time elapsed since system boot, in nanoseconds.
491 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
493 * This helper is a "printk()-like" facility for debugging. It
494 * prints a message defined by format *fmt* (of size *fmt_size*)
495 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
496 * available. It can take up to three additional **u64**
497 * arguments (as an eBPF helpers, the total number of arguments is
500 * Each time the helper is called, it appends a line to the trace.
501 * The format of the trace is customizable, and the exact output
502 * one will get depends on the options set in
503 * *\/sys/kernel/debug/tracing/trace_options* (see also the
504 * *README* file under the same directory). However, it usually
505 * defaults to something like:
509 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
513 * * ``telnet`` is the name of the current task.
514 * * ``470`` is the PID of the current task.
515 * * ``001`` is the CPU number on which the task is
517 * * In ``.N..``, each character refers to a set of
518 * options (whether irqs are enabled, scheduling
519 * options, whether hard/softirqs are running, level of
520 * preempt_disabled respectively). **N** means that
521 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
523 * * ``419421.045894`` is a timestamp.
524 * * ``0x00000001`` is a fake value used by BPF for the
525 * instruction pointer register.
526 * * ``<formatted msg>`` is the message formatted with
529 * The conversion specifiers supported by *fmt* are similar, but
530 * more limited than for printk(). They are **%d**, **%i**,
531 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
532 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
533 * of field, padding with zeroes, etc.) is available, and the
534 * helper will return **-EINVAL** (but print nothing) if it
535 * encounters an unknown specifier.
537 * Also, note that **bpf_trace_printk**\ () is slow, and should
538 * only be used for debugging purposes. For this reason, a notice
539 * bloc (spanning several lines) is printed to kernel logs and
540 * states that the helper should not be used "for production use"
541 * the first time this helper is used (or more precisely, when
542 * **trace_printk**\ () buffers are allocated). For passing values
543 * to user space, perf events should be preferred.
545 * The number of bytes written to the buffer, or a negative error
546 * in case of failure.
548 * u32 bpf_get_prandom_u32(void)
550 * Get a pseudo-random number.
552 * From a security point of view, this helper uses its own
553 * pseudo-random internal state, and cannot be used to infer the
554 * seed of other random functions in the kernel. However, it is
555 * essential to note that the generator used by the helper is not
556 * cryptographically secure.
558 * A random 32-bit unsigned value.
560 * u32 bpf_get_smp_processor_id(void)
562 * Get the SMP (symmetric multiprocessing) processor id. Note that
563 * all programs run with preemption disabled, which means that the
564 * SMP processor id is stable during all the execution of the
567 * The SMP id of the processor running the program.
569 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
571 * Store *len* bytes from address *from* into the packet
572 * associated to *skb*, at *offset*. *flags* are a combination of
573 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
574 * checksum for the packet after storing the bytes) and
575 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
576 * **->swhash** and *skb*\ **->l4hash** to 0).
578 * A call to this helper is susceptible to change the underlaying
579 * packet buffer. Therefore, at load time, all checks on pointers
580 * previously done by the verifier are invalidated and must be
581 * performed again, if the helper is used in combination with
582 * direct packet access.
584 * 0 on success, or a negative error in case of failure.
586 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
588 * Recompute the layer 3 (e.g. IP) checksum for the packet
589 * associated to *skb*. Computation is incremental, so the helper
590 * must know the former value of the header field that was
591 * modified (*from*), the new value of this field (*to*), and the
592 * number of bytes (2 or 4) for this field, stored in *size*.
593 * Alternatively, it is possible to store the difference between
594 * the previous and the new values of the header field in *to*, by
595 * setting *from* and *size* to 0. For both methods, *offset*
596 * indicates the location of the IP checksum within the packet.
598 * This helper works in combination with **bpf_csum_diff**\ (),
599 * which does not update the checksum in-place, but offers more
600 * flexibility and can handle sizes larger than 2 or 4 for the
601 * checksum to update.
603 * A call to this helper is susceptible to change the underlaying
604 * packet buffer. Therefore, at load time, all checks on pointers
605 * previously done by the verifier are invalidated and must be
606 * performed again, if the helper is used in combination with
607 * direct packet access.
609 * 0 on success, or a negative error in case of failure.
611 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
613 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
614 * packet associated to *skb*. Computation is incremental, so the
615 * helper must know the former value of the header field that was
616 * modified (*from*), the new value of this field (*to*), and the
617 * number of bytes (2 or 4) for this field, stored on the lowest
618 * four bits of *flags*. Alternatively, it is possible to store
619 * the difference between the previous and the new values of the
620 * header field in *to*, by setting *from* and the four lowest
621 * bits of *flags* to 0. For both methods, *offset* indicates the
622 * location of the IP checksum within the packet. In addition to
623 * the size of the field, *flags* can be added (bitwise OR) actual
624 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
625 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
626 * for updates resulting in a null checksum the value is set to
627 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
628 * the checksum is to be computed against a pseudo-header.
630 * This helper works in combination with **bpf_csum_diff**\ (),
631 * which does not update the checksum in-place, but offers more
632 * flexibility and can handle sizes larger than 2 or 4 for the
633 * checksum to update.
635 * A call to this helper is susceptible to change the underlaying
636 * packet buffer. Therefore, at load time, all checks on pointers
637 * previously done by the verifier are invalidated and must be
638 * performed again, if the helper is used in combination with
639 * direct packet access.
641 * 0 on success, or a negative error in case of failure.
643 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
645 * This special helper is used to trigger a "tail call", or in
646 * other words, to jump into another eBPF program. The same stack
647 * frame is used (but values on stack and in registers for the
648 * caller are not accessible to the callee). This mechanism allows
649 * for program chaining, either for raising the maximum number of
650 * available eBPF instructions, or to execute given programs in
651 * conditional blocks. For security reasons, there is an upper
652 * limit to the number of successive tail calls that can be
655 * Upon call of this helper, the program attempts to jump into a
656 * program referenced at index *index* in *prog_array_map*, a
657 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
658 * *ctx*, a pointer to the context.
660 * If the call succeeds, the kernel immediately runs the first
661 * instruction of the new program. This is not a function call,
662 * and it never returns to the previous program. If the call
663 * fails, then the helper has no effect, and the caller continues
664 * to run its subsequent instructions. A call can fail if the
665 * destination program for the jump does not exist (i.e. *index*
666 * is superior to the number of entries in *prog_array_map*), or
667 * if the maximum number of tail calls has been reached for this
668 * chain of programs. This limit is defined in the kernel by the
669 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
670 * which is currently set to 32.
672 * 0 on success, or a negative error in case of failure.
674 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
676 * Clone and redirect the packet associated to *skb* to another
677 * net device of index *ifindex*. Both ingress and egress
678 * interfaces can be used for redirection. The **BPF_F_INGRESS**
679 * value in *flags* is used to make the distinction (ingress path
680 * is selected if the flag is present, egress path otherwise).
681 * This is the only flag supported for now.
683 * In comparison with **bpf_redirect**\ () helper,
684 * **bpf_clone_redirect**\ () has the associated cost of
685 * duplicating the packet buffer, but this can be executed out of
686 * the eBPF program. Conversely, **bpf_redirect**\ () is more
687 * efficient, but it is handled through an action code where the
688 * redirection happens only after the eBPF program has returned.
690 * A call to this helper is susceptible to change the underlaying
691 * packet buffer. Therefore, at load time, all checks on pointers
692 * previously done by the verifier are invalidated and must be
693 * performed again, if the helper is used in combination with
694 * direct packet access.
696 * 0 on success, or a negative error in case of failure. Positive
697 * error indicates a potential drop or congestion in the target
698 * device. The particular positive error codes are not defined.
700 * u64 bpf_get_current_pid_tgid(void)
702 * A 64-bit integer containing the current tgid and pid, and
704 * *current_task*\ **->tgid << 32 \|**
705 * *current_task*\ **->pid**.
707 * u64 bpf_get_current_uid_gid(void)
709 * A 64-bit integer containing the current GID and UID, and
710 * created as such: *current_gid* **<< 32 \|** *current_uid*.
712 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
714 * Copy the **comm** attribute of the current task into *buf* of
715 * *size_of_buf*. The **comm** attribute contains the name of
716 * the executable (excluding the path) for the current task. The
717 * *size_of_buf* must be strictly positive. On success, the
718 * helper makes sure that the *buf* is NUL-terminated. On failure,
719 * it is filled with zeroes.
721 * 0 on success, or a negative error in case of failure.
723 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
725 * Retrieve the classid for the current task, i.e. for the net_cls
726 * cgroup to which *skb* belongs.
728 * This helper can be used on TC egress path, but not on ingress.
730 * The net_cls cgroup provides an interface to tag network packets
731 * based on a user-provided identifier for all traffic coming from
732 * the tasks belonging to the related cgroup. See also the related
733 * kernel documentation, available from the Linux sources in file
734 * *Documentation/cgroup-v1/net_cls.txt*.
736 * The Linux kernel has two versions for cgroups: there are
737 * cgroups v1 and cgroups v2. Both are available to users, who can
738 * use a mixture of them, but note that the net_cls cgroup is for
739 * cgroup v1 only. This makes it incompatible with BPF programs
740 * run on cgroups, which is a cgroup-v2-only feature (a socket can
741 * only hold data for one version of cgroups at a time).
743 * This helper is only available is the kernel was compiled with
744 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
745 * "**y**" or to "**m**".
747 * The classid, or 0 for the default unconfigured classid.
749 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
751 * Push a *vlan_tci* (VLAN tag control information) of protocol
752 * *vlan_proto* to the packet associated to *skb*, then update
753 * the checksum. Note that if *vlan_proto* is different from
754 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
755 * be **ETH_P_8021Q**.
757 * A call to this helper is susceptible to change the underlaying
758 * packet buffer. Therefore, at load time, all checks on pointers
759 * previously done by the verifier are invalidated and must be
760 * performed again, if the helper is used in combination with
761 * direct packet access.
763 * 0 on success, or a negative error in case of failure.
765 * int bpf_skb_vlan_pop(struct sk_buff *skb)
767 * Pop a VLAN header from the packet associated to *skb*.
769 * A call to this helper is susceptible to change the underlaying
770 * packet buffer. Therefore, at load time, all checks on pointers
771 * previously done by the verifier are invalidated and must be
772 * performed again, if the helper is used in combination with
773 * direct packet access.
775 * 0 on success, or a negative error in case of failure.
777 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
779 * Get tunnel metadata. This helper takes a pointer *key* to an
780 * empty **struct bpf_tunnel_key** of **size**, that will be
781 * filled with tunnel metadata for the packet associated to *skb*.
782 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
783 * indicates that the tunnel is based on IPv6 protocol instead of
786 * The **struct bpf_tunnel_key** is an object that generalizes the
787 * principal parameters used by various tunneling protocols into a
788 * single struct. This way, it can be used to easily make a
789 * decision based on the contents of the encapsulation header,
790 * "summarized" in this struct. In particular, it holds the IP
791 * address of the remote end (IPv4 or IPv6, depending on the case)
792 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
793 * this struct exposes the *key*\ **->tunnel_id**, which is
794 * generally mapped to a VNI (Virtual Network Identifier), making
795 * it programmable together with the **bpf_skb_set_tunnel_key**\
798 * Let's imagine that the following code is part of a program
799 * attached to the TC ingress interface, on one end of a GRE
800 * tunnel, and is supposed to filter out all messages coming from
801 * remote ends with IPv4 address other than 10.0.0.1:
806 * struct bpf_tunnel_key key = {};
808 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
810 * return TC_ACT_SHOT; // drop packet
812 * if (key.remote_ipv4 != 0x0a000001)
813 * return TC_ACT_SHOT; // drop packet
815 * return TC_ACT_OK; // accept packet
817 * This interface can also be used with all encapsulation devices
818 * that can operate in "collect metadata" mode: instead of having
819 * one network device per specific configuration, the "collect
820 * metadata" mode only requires a single device where the
821 * configuration can be extracted from this helper.
823 * This can be used together with various tunnels such as VXLan,
824 * Geneve, GRE or IP in IP (IPIP).
826 * 0 on success, or a negative error in case of failure.
828 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
830 * Populate tunnel metadata for packet associated to *skb.* The
831 * tunnel metadata is set to the contents of *key*, of *size*. The
832 * *flags* can be set to a combination of the following values:
834 * **BPF_F_TUNINFO_IPV6**
835 * Indicate that the tunnel is based on IPv6 protocol
837 * **BPF_F_ZERO_CSUM_TX**
838 * For IPv4 packets, add a flag to tunnel metadata
839 * indicating that checksum computation should be skipped
840 * and checksum set to zeroes.
841 * **BPF_F_DONT_FRAGMENT**
842 * Add a flag to tunnel metadata indicating that the
843 * packet should not be fragmented.
844 * **BPF_F_SEQ_NUMBER**
845 * Add a flag to tunnel metadata indicating that a
846 * sequence number should be added to tunnel header before
847 * sending the packet. This flag was added for GRE
848 * encapsulation, but might be used with other protocols
849 * as well in the future.
851 * Here is a typical usage on the transmit path:
855 * struct bpf_tunnel_key key;
857 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
858 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
860 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
861 * helper for additional information.
863 * 0 on success, or a negative error in case of failure.
865 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
867 * Read the value of a perf event counter. This helper relies on a
868 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
869 * the perf event counter is selected when *map* is updated with
870 * perf event file descriptors. The *map* is an array whose size
871 * is the number of available CPUs, and each cell contains a value
872 * relative to one CPU. The value to retrieve is indicated by
873 * *flags*, that contains the index of the CPU to look up, masked
874 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
875 * **BPF_F_CURRENT_CPU** to indicate that the value for the
876 * current CPU should be retrieved.
878 * Note that before Linux 4.13, only hardware perf event can be
881 * Also, be aware that the newer helper
882 * **bpf_perf_event_read_value**\ () is recommended over
883 * **bpf_perf_event_read**\ () in general. The latter has some ABI
884 * quirks where error and counter value are used as a return code
885 * (which is wrong to do since ranges may overlap). This issue is
886 * fixed with **bpf_perf_event_read_value**\ (), which at the same
887 * time provides more features over the **bpf_perf_event_read**\
888 * () interface. Please refer to the description of
889 * **bpf_perf_event_read_value**\ () for details.
891 * The value of the perf event counter read from the map, or a
892 * negative error code in case of failure.
894 * int bpf_redirect(u32 ifindex, u64 flags)
896 * Redirect the packet to another net device of index *ifindex*.
897 * This helper is somewhat similar to **bpf_clone_redirect**\
898 * (), except that the packet is not cloned, which provides
899 * increased performance.
901 * Except for XDP, both ingress and egress interfaces can be used
902 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
903 * to make the distinction (ingress path is selected if the flag
904 * is present, egress path otherwise). Currently, XDP only
905 * supports redirection to the egress interface, and accepts no
908 * The same effect can be attained with the more generic
909 * **bpf_redirect_map**\ (), which requires specific maps to be
910 * used but offers better performance.
912 * For XDP, the helper returns **XDP_REDIRECT** on success or
913 * **XDP_ABORTED** on error. For other program types, the values
914 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
917 * u32 bpf_get_route_realm(struct sk_buff *skb)
919 * Retrieve the realm or the route, that is to say the
920 * **tclassid** field of the destination for the *skb*. The
921 * indentifier retrieved is a user-provided tag, similar to the
922 * one used with the net_cls cgroup (see description for
923 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
924 * held by a route (a destination entry), not by a task.
926 * Retrieving this identifier works with the clsact TC egress hook
927 * (see also **tc-bpf(8)**), or alternatively on conventional
928 * classful egress qdiscs, but not on TC ingress path. In case of
929 * clsact TC egress hook, this has the advantage that, internally,
930 * the destination entry has not been dropped yet in the transmit
931 * path. Therefore, the destination entry does not need to be
932 * artificially held via **netif_keep_dst**\ () for a classful
933 * qdisc until the *skb* is freed.
935 * This helper is available only if the kernel was compiled with
936 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
938 * The realm of the route for the packet associated to *skb*, or 0
941 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
943 * Write raw *data* blob into a special BPF perf event held by
944 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
945 * event must have the following attributes: **PERF_SAMPLE_RAW**
946 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
947 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
949 * The *flags* are used to indicate the index in *map* for which
950 * the value must be put, masked with **BPF_F_INDEX_MASK**.
951 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
952 * to indicate that the index of the current CPU core should be
955 * The value to write, of *size*, is passed through eBPF stack and
958 * The context of the program *ctx* needs also be passed to the
961 * On user space, a program willing to read the values needs to
962 * call **perf_event_open**\ () on the perf event (either for
963 * one or for all CPUs) and to store the file descriptor into the
964 * *map*. This must be done before the eBPF program can send data
965 * into it. An example is available in file
966 * *samples/bpf/trace_output_user.c* in the Linux kernel source
967 * tree (the eBPF program counterpart is in
968 * *samples/bpf/trace_output_kern.c*).
970 * **bpf_perf_event_output**\ () achieves better performance
971 * than **bpf_trace_printk**\ () for sharing data with user
972 * space, and is much better suitable for streaming data from eBPF
975 * Note that this helper is not restricted to tracing use cases
976 * and can be used with programs attached to TC or XDP as well,
977 * where it allows for passing data to user space listeners. Data
980 * * Only custom structs,
981 * * Only the packet payload, or
982 * * A combination of both.
984 * 0 on success, or a negative error in case of failure.
986 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
988 * This helper was provided as an easy way to load data from a
989 * packet. It can be used to load *len* bytes from *offset* from
990 * the packet associated to *skb*, into the buffer pointed by
993 * Since Linux 4.7, usage of this helper has mostly been replaced
994 * by "direct packet access", enabling packet data to be
995 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
996 * pointing respectively to the first byte of packet data and to
997 * the byte after the last byte of packet data. However, it
998 * remains useful if one wishes to read large quantities of data
999 * at once from a packet into the eBPF stack.
1001 * 0 on success, or a negative error in case of failure.
1003 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1005 * Walk a user or a kernel stack and return its id. To achieve
1006 * this, the helper needs *ctx*, which is a pointer to the context
1007 * on which the tracing program is executed, and a pointer to a
1008 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1010 * The last argument, *flags*, holds the number of stack frames to
1011 * skip (from 0 to 255), masked with
1012 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1013 * a combination of the following flags:
1015 * **BPF_F_USER_STACK**
1016 * Collect a user space stack instead of a kernel stack.
1017 * **BPF_F_FAST_STACK_CMP**
1018 * Compare stacks by hash only.
1019 * **BPF_F_REUSE_STACKID**
1020 * If two different stacks hash into the same *stackid*,
1021 * discard the old one.
1023 * The stack id retrieved is a 32 bit long integer handle which
1024 * can be further combined with other data (including other stack
1025 * ids) and used as a key into maps. This can be useful for
1026 * generating a variety of graphs (such as flame graphs or off-cpu
1029 * For walking a stack, this helper is an improvement over
1030 * **bpf_probe_read**\ (), which can be used with unrolled loops
1031 * but is not efficient and consumes a lot of eBPF instructions.
1032 * Instead, **bpf_get_stackid**\ () can collect up to
1033 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1034 * this limit can be controlled with the **sysctl** program, and
1035 * that it should be manually increased in order to profile long
1036 * user stacks (such as stacks for Java programs). To do so, use:
1040 * # sysctl kernel.perf_event_max_stack=<new value>
1042 * The positive or null stack id on success, or a negative error
1043 * in case of failure.
1045 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1047 * Compute a checksum difference, from the raw buffer pointed by
1048 * *from*, of length *from_size* (that must be a multiple of 4),
1049 * towards the raw buffer pointed by *to*, of size *to_size*
1050 * (same remark). An optional *seed* can be added to the value
1051 * (this can be cascaded, the seed may come from a previous call
1054 * This is flexible enough to be used in several ways:
1056 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1057 * checksum, it can be used when pushing new data.
1058 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1059 * checksum, it can be used when removing data from a packet.
1060 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1061 * can be used to compute a diff. Note that *from_size* and
1062 * *to_size* do not need to be equal.
1064 * This helper can be used in combination with
1065 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1066 * which one can feed in the difference computed with
1067 * **bpf_csum_diff**\ ().
1069 * The checksum result, or a negative error code in case of
1072 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1074 * Retrieve tunnel options metadata for the packet associated to
1075 * *skb*, and store the raw tunnel option data to the buffer *opt*
1078 * This helper can be used with encapsulation devices that can
1079 * operate in "collect metadata" mode (please refer to the related
1080 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1081 * more details). A particular example where this can be used is
1082 * in combination with the Geneve encapsulation protocol, where it
1083 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1084 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1085 * the eBPF program. This allows for full customization of these
1088 * The size of the option data retrieved.
1090 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1092 * Set tunnel options metadata for the packet associated to *skb*
1093 * to the option data contained in the raw buffer *opt* of *size*.
1095 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1096 * helper for additional information.
1098 * 0 on success, or a negative error in case of failure.
1100 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1102 * Change the protocol of the *skb* to *proto*. Currently
1103 * supported are transition from IPv4 to IPv6, and from IPv6 to
1104 * IPv4. The helper takes care of the groundwork for the
1105 * transition, including resizing the socket buffer. The eBPF
1106 * program is expected to fill the new headers, if any, via
1107 * **skb_store_bytes**\ () and to recompute the checksums with
1108 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1109 * (). The main case for this helper is to perform NAT64
1110 * operations out of an eBPF program.
1112 * Internally, the GSO type is marked as dodgy so that headers are
1113 * checked and segments are recalculated by the GSO/GRO engine.
1114 * The size for GSO target is adapted as well.
1116 * All values for *flags* are reserved for future usage, and must
1119 * A call to this helper is susceptible to change the underlaying
1120 * packet buffer. Therefore, at load time, all checks on pointers
1121 * previously done by the verifier are invalidated and must be
1122 * performed again, if the helper is used in combination with
1123 * direct packet access.
1125 * 0 on success, or a negative error in case of failure.
1127 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1129 * Change the packet type for the packet associated to *skb*. This
1130 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1131 * the eBPF program does not have a write access to *skb*\
1132 * **->pkt_type** beside this helper. Using a helper here allows
1133 * for graceful handling of errors.
1135 * The major use case is to change incoming *skb*s to
1136 * **PACKET_HOST** in a programmatic way instead of having to
1137 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1140 * Note that *type* only allows certain values. At this time, they
1145 * **PACKET_BROADCAST**
1146 * Send packet to all.
1147 * **PACKET_MULTICAST**
1148 * Send packet to group.
1149 * **PACKET_OTHERHOST**
1150 * Send packet to someone else.
1152 * 0 on success, or a negative error in case of failure.
1154 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1156 * Check whether *skb* is a descendant of the cgroup2 held by
1157 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1159 * The return value depends on the result of the test, and can be:
1161 * * 0, if the *skb* failed the cgroup2 descendant test.
1162 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1163 * * A negative error code, if an error occurred.
1165 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1167 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1168 * not set, in particular if the hash was cleared due to mangling,
1169 * recompute this hash. Later accesses to the hash can be done
1170 * directly with *skb*\ **->hash**.
1172 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1173 * prototype with **bpf_skb_change_proto**\ (), or calling
1174 * **bpf_skb_store_bytes**\ () with the
1175 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1176 * the hash and to trigger a new computation for the next call to
1177 * **bpf_get_hash_recalc**\ ().
1181 * u64 bpf_get_current_task(void)
1183 * A pointer to the current task struct.
1185 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1187 * Attempt in a safe way to write *len* bytes from the buffer
1188 * *src* to *dst* in memory. It only works for threads that are in
1189 * user context, and *dst* must be a valid user space address.
1191 * This helper should not be used to implement any kind of
1192 * security mechanism because of TOC-TOU attacks, but rather to
1193 * debug, divert, and manipulate execution of semi-cooperative
1196 * Keep in mind that this feature is meant for experiments, and it
1197 * has a risk of crashing the system and running programs.
1198 * Therefore, when an eBPF program using this helper is attached,
1199 * a warning including PID and process name is printed to kernel
1202 * 0 on success, or a negative error in case of failure.
1204 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1206 * Check whether the probe is being run is the context of a given
1207 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1208 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1210 * The return value depends on the result of the test, and can be:
1212 * * 1, if current task belongs to the cgroup2.
1213 * * 0, if current task does not belong to the cgroup2.
1214 * * A negative error code, if an error occurred.
1216 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1218 * Resize (trim or grow) the packet associated to *skb* to the
1219 * new *len*. The *flags* are reserved for future usage, and must
1222 * The basic idea is that the helper performs the needed work to
1223 * change the size of the packet, then the eBPF program rewrites
1224 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1225 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1226 * and others. This helper is a slow path utility intended for
1227 * replies with control messages. And because it is targeted for
1228 * slow path, the helper itself can afford to be slow: it
1229 * implicitly linearizes, unclones and drops offloads from the
1232 * A call to this helper is susceptible to change the underlaying
1233 * packet buffer. Therefore, at load time, all checks on pointers
1234 * previously done by the verifier are invalidated and must be
1235 * performed again, if the helper is used in combination with
1236 * direct packet access.
1238 * 0 on success, or a negative error in case of failure.
1240 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1242 * Pull in non-linear data in case the *skb* is non-linear and not
1243 * all of *len* are part of the linear section. Make *len* bytes
1244 * from *skb* readable and writable. If a zero value is passed for
1245 * *len*, then the whole length of the *skb* is pulled.
1247 * This helper is only needed for reading and writing with direct
1250 * For direct packet access, testing that offsets to access
1251 * are within packet boundaries (test on *skb*\ **->data_end**) is
1252 * susceptible to fail if offsets are invalid, or if the requested
1253 * data is in non-linear parts of the *skb*. On failure the
1254 * program can just bail out, or in the case of a non-linear
1255 * buffer, use a helper to make the data available. The
1256 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1257 * the data. Another one consists in using **bpf_skb_pull_data**
1258 * to pull in once the non-linear parts, then retesting and
1259 * eventually access the data.
1261 * At the same time, this also makes sure the *skb* is uncloned,
1262 * which is a necessary condition for direct write. As this needs
1263 * to be an invariant for the write part only, the verifier
1264 * detects writes and adds a prologue that is calling
1265 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1266 * the very beginning in case it is indeed cloned.
1268 * A call to this helper is susceptible to change the underlaying
1269 * packet buffer. Therefore, at load time, all checks on pointers
1270 * previously done by the verifier are invalidated and must be
1271 * performed again, if the helper is used in combination with
1272 * direct packet access.
1274 * 0 on success, or a negative error in case of failure.
1276 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1278 * Add the checksum *csum* into *skb*\ **->csum** in case the
1279 * driver has supplied a checksum for the entire packet into that
1280 * field. Return an error otherwise. This helper is intended to be
1281 * used in combination with **bpf_csum_diff**\ (), in particular
1282 * when the checksum needs to be updated after data has been
1283 * written into the packet through direct packet access.
1285 * The checksum on success, or a negative error code in case of
1288 * void bpf_set_hash_invalid(struct sk_buff *skb)
1290 * Invalidate the current *skb*\ **->hash**. It can be used after
1291 * mangling on headers through direct packet access, in order to
1292 * indicate that the hash is outdated and to trigger a
1293 * recalculation the next time the kernel tries to access this
1294 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1296 * int bpf_get_numa_node_id(void)
1298 * Return the id of the current NUMA node. The primary use case
1299 * for this helper is the selection of sockets for the local NUMA
1300 * node, when the program is attached to sockets using the
1301 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1302 * but the helper is also available to other eBPF program types,
1303 * similarly to **bpf_get_smp_processor_id**\ ().
1305 * The id of current NUMA node.
1307 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1309 * Grows headroom of packet associated to *skb* and adjusts the
1310 * offset of the MAC header accordingly, adding *len* bytes of
1311 * space. It automatically extends and reallocates memory as
1314 * This helper can be used on a layer 3 *skb* to push a MAC header
1315 * for redirection into a layer 2 device.
1317 * All values for *flags* are reserved for future usage, and must
1320 * A call to this helper is susceptible to change the underlaying
1321 * packet buffer. Therefore, at load time, all checks on pointers
1322 * previously done by the verifier are invalidated and must be
1323 * performed again, if the helper is used in combination with
1324 * direct packet access.
1326 * 0 on success, or a negative error in case of failure.
1328 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1330 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1331 * it is possible to use a negative value for *delta*. This helper
1332 * can be used to prepare the packet for pushing or popping
1335 * A call to this helper is susceptible to change the underlaying
1336 * packet buffer. Therefore, at load time, all checks on pointers
1337 * previously done by the verifier are invalidated and must be
1338 * performed again, if the helper is used in combination with
1339 * direct packet access.
1341 * 0 on success, or a negative error in case of failure.
1343 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1345 * Copy a NUL terminated string from an unsafe address
1346 * *unsafe_ptr* to *dst*. The *size* should include the
1347 * terminating NUL byte. In case the string length is smaller than
1348 * *size*, the target is not padded with further NUL bytes. If the
1349 * string length is larger than *size*, just *size*-1 bytes are
1350 * copied and the last byte is set to NUL.
1352 * On success, the length of the copied string is returned. This
1353 * makes this helper useful in tracing programs for reading
1354 * strings, and more importantly to get its length at runtime. See
1355 * the following snippet:
1359 * SEC("kprobe/sys_open")
1360 * void bpf_sys_open(struct pt_regs *ctx)
1362 * char buf[PATHLEN]; // PATHLEN is defined to 256
1363 * int res = bpf_probe_read_str(buf, sizeof(buf),
1366 * // Consume buf, for example push it to
1367 * // userspace via bpf_perf_event_output(); we
1368 * // can use res (the string length) as event
1369 * // size, after checking its boundaries.
1372 * In comparison, using **bpf_probe_read()** helper here instead
1373 * to read the string would require to estimate the length at
1374 * compile time, and would often result in copying more memory
1377 * Another useful use case is when parsing individual process
1378 * arguments or individual environment variables navigating
1379 * *current*\ **->mm->arg_start** and *current*\
1380 * **->mm->env_start**: using this helper and the return value,
1381 * one can quickly iterate at the right offset of the memory area.
1383 * On success, the strictly positive length of the string,
1384 * including the trailing NUL character. On error, a negative
1387 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1389 * If the **struct sk_buff** pointed by *skb* has a known socket,
1390 * retrieve the cookie (generated by the kernel) of this socket.
1391 * If no cookie has been set yet, generate a new cookie. Once
1392 * generated, the socket cookie remains stable for the life of the
1393 * socket. This helper can be useful for monitoring per socket
1394 * networking traffic statistics as it provides a unique socket
1395 * identifier per namespace.
1397 * A 8-byte long non-decreasing number on success, or 0 if the
1398 * socket field is missing inside *skb*.
1400 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1402 * Equivalent to bpf_get_socket_cookie() helper that accepts
1403 * *skb*, but gets socket from **struct bpf_sock_addr** contex.
1405 * A 8-byte long non-decreasing number.
1407 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1409 * Equivalent to bpf_get_socket_cookie() helper that accepts
1410 * *skb*, but gets socket from **struct bpf_sock_ops** contex.
1412 * A 8-byte long non-decreasing number.
1414 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1416 * The owner UID of the socket associated to *skb*. If the socket
1417 * is **NULL**, or if it is not a full socket (i.e. if it is a
1418 * time-wait or a request socket instead), **overflowuid** value
1419 * is returned (note that **overflowuid** might also be the actual
1420 * UID value for the socket).
1422 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1424 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1429 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1431 * Emulate a call to **setsockopt()** on the socket associated to
1432 * *bpf_socket*, which must be a full socket. The *level* at
1433 * which the option resides and the name *optname* of the option
1434 * must be specified, see **setsockopt(2)** for more information.
1435 * The option value of length *optlen* is pointed by *optval*.
1437 * This helper actually implements a subset of **setsockopt()**.
1438 * It supports the following *level*\ s:
1440 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1441 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1442 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1443 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1444 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1445 * **TCP_BPF_SNDCWND_CLAMP**.
1446 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1447 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1449 * 0 on success, or a negative error in case of failure.
1451 * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags)
1453 * Grow or shrink the room for data in the packet associated to
1454 * *skb* by *len_diff*, and according to the selected *mode*.
1456 * There is a single supported mode at this time:
1458 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1459 * (room space is added or removed below the layer 3 header).
1461 * All values for *flags* are reserved for future usage, and must
1464 * A call to this helper is susceptible to change the underlaying
1465 * packet buffer. Therefore, at load time, all checks on pointers
1466 * previously done by the verifier are invalidated and must be
1467 * performed again, if the helper is used in combination with
1468 * direct packet access.
1470 * 0 on success, or a negative error in case of failure.
1472 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1474 * Redirect the packet to the endpoint referenced by *map* at
1475 * index *key*. Depending on its type, this *map* can contain
1476 * references to net devices (for forwarding packets through other
1477 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1478 * but this is only implemented for native XDP (with driver
1479 * support) as of this writing).
1481 * All values for *flags* are reserved for future usage, and must
1484 * When used to redirect packets to net devices, this helper
1485 * provides a high performance increase over **bpf_redirect**\ ().
1486 * This is due to various implementation details of the underlying
1487 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1488 * () tries to send packet as a "bulk" to the device.
1490 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1492 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1494 * Redirect the packet to the socket referenced by *map* (of type
1495 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1496 * egress interfaces can be used for redirection. The
1497 * **BPF_F_INGRESS** value in *flags* is used to make the
1498 * distinction (ingress path is selected if the flag is present,
1499 * egress path otherwise). This is the only flag supported for now.
1501 * **SK_PASS** on success, or **SK_DROP** on error.
1503 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1505 * Add an entry to, or update a *map* referencing sockets. The
1506 * *skops* is used as a new value for the entry associated to
1507 * *key*. *flags* is one of:
1510 * The entry for *key* must not exist in the map.
1512 * The entry for *key* must already exist in the map.
1514 * No condition on the existence of the entry for *key*.
1516 * If the *map* has eBPF programs (parser and verdict), those will
1517 * be inherited by the socket being added. If the socket is
1518 * already attached to eBPF programs, this results in an error.
1520 * 0 on success, or a negative error in case of failure.
1522 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1524 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1525 * *delta* (which can be positive or negative). Note that this
1526 * operation modifies the address stored in *xdp_md*\ **->data**,
1527 * so the latter must be loaded only after the helper has been
1530 * The use of *xdp_md*\ **->data_meta** is optional and programs
1531 * are not required to use it. The rationale is that when the
1532 * packet is processed with XDP (e.g. as DoS filter), it is
1533 * possible to push further meta data along with it before passing
1534 * to the stack, and to give the guarantee that an ingress eBPF
1535 * program attached as a TC classifier on the same device can pick
1536 * this up for further post-processing. Since TC works with socket
1537 * buffers, it remains possible to set from XDP the **mark** or
1538 * **priority** pointers, or other pointers for the socket buffer.
1539 * Having this scratch space generic and programmable allows for
1540 * more flexibility as the user is free to store whatever meta
1543 * A call to this helper is susceptible to change the underlaying
1544 * packet buffer. Therefore, at load time, all checks on pointers
1545 * previously done by the verifier are invalidated and must be
1546 * performed again, if the helper is used in combination with
1547 * direct packet access.
1549 * 0 on success, or a negative error in case of failure.
1551 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1553 * Read the value of a perf event counter, and store it into *buf*
1554 * of size *buf_size*. This helper relies on a *map* of type
1555 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1556 * counter is selected when *map* is updated with perf event file
1557 * descriptors. The *map* is an array whose size is the number of
1558 * available CPUs, and each cell contains a value relative to one
1559 * CPU. The value to retrieve is indicated by *flags*, that
1560 * contains the index of the CPU to look up, masked with
1561 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1562 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1563 * current CPU should be retrieved.
1565 * This helper behaves in a way close to
1566 * **bpf_perf_event_read**\ () helper, save that instead of
1567 * just returning the value observed, it fills the *buf*
1568 * structure. This allows for additional data to be retrieved: in
1569 * particular, the enabled and running times (in *buf*\
1570 * **->enabled** and *buf*\ **->running**, respectively) are
1571 * copied. In general, **bpf_perf_event_read_value**\ () is
1572 * recommended over **bpf_perf_event_read**\ (), which has some
1573 * ABI issues and provides fewer functionalities.
1575 * These values are interesting, because hardware PMU (Performance
1576 * Monitoring Unit) counters are limited resources. When there are
1577 * more PMU based perf events opened than available counters,
1578 * kernel will multiplex these events so each event gets certain
1579 * percentage (but not all) of the PMU time. In case that
1580 * multiplexing happens, the number of samples or counter value
1581 * will not reflect the case compared to when no multiplexing
1582 * occurs. This makes comparison between different runs difficult.
1583 * Typically, the counter value should be normalized before
1584 * comparing to other experiments. The usual normalization is done
1589 * normalized_counter = counter * t_enabled / t_running
1591 * Where t_enabled is the time enabled for event and t_running is
1592 * the time running for event since last normalization. The
1593 * enabled and running times are accumulated since the perf event
1594 * open. To achieve scaling factor between two invocations of an
1595 * eBPF program, users can can use CPU id as the key (which is
1596 * typical for perf array usage model) to remember the previous
1597 * value and do the calculation inside the eBPF program.
1599 * 0 on success, or a negative error in case of failure.
1601 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1603 * For en eBPF program attached to a perf event, retrieve the
1604 * value of the event counter associated to *ctx* and store it in
1605 * the structure pointed by *buf* and of size *buf_size*. Enabled
1606 * and running times are also stored in the structure (see
1607 * description of helper **bpf_perf_event_read_value**\ () for
1610 * 0 on success, or a negative error in case of failure.
1612 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1614 * Emulate a call to **getsockopt()** on the socket associated to
1615 * *bpf_socket*, which must be a full socket. The *level* at
1616 * which the option resides and the name *optname* of the option
1617 * must be specified, see **getsockopt(2)** for more information.
1618 * The retrieved value is stored in the structure pointed by
1619 * *opval* and of length *optlen*.
1621 * This helper actually implements a subset of **getsockopt()**.
1622 * It supports the following *level*\ s:
1624 * * **IPPROTO_TCP**, which supports *optname*
1625 * **TCP_CONGESTION**.
1626 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1627 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1629 * 0 on success, or a negative error in case of failure.
1631 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1633 * Used for error injection, this helper uses kprobes to override
1634 * the return value of the probed function, and to set it to *rc*.
1635 * The first argument is the context *regs* on which the kprobe
1638 * This helper works by setting setting the PC (program counter)
1639 * to an override function which is run in place of the original
1640 * probed function. This means the probed function is not run at
1641 * all. The replacement function just returns with the required
1644 * This helper has security implications, and thus is subject to
1645 * restrictions. It is only available if the kernel was compiled
1646 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1647 * option, and in this case it only works on functions tagged with
1648 * **ALLOW_ERROR_INJECTION** in the kernel code.
1650 * Also, the helper is only available for the architectures having
1651 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1652 * x86 architecture is the only one to support this feature.
1656 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1658 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1659 * for the full TCP socket associated to *bpf_sock_ops* to
1662 * The primary use of this field is to determine if there should
1663 * be calls to eBPF programs of type
1664 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1665 * code. A program of the same type can change its value, per
1666 * connection and as necessary, when the connection is
1667 * established. This field is directly accessible for reading, but
1668 * this helper must be used for updates in order to return an
1669 * error if an eBPF program tries to set a callback that is not
1670 * supported in the current kernel.
1672 * The supported callback values that *argval* can combine are:
1674 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1675 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1676 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1678 * Here are some examples of where one could call such eBPF
1682 * * When a packet is retransmitted.
1683 * * When the connection terminates.
1684 * * When a packet is sent.
1685 * * When a packet is received.
1687 * Code **-EINVAL** if the socket is not a full TCP socket;
1688 * otherwise, a positive number containing the bits that could not
1689 * be set is returned (which comes down to 0 if all bits were set
1692 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1694 * This helper is used in programs implementing policies at the
1695 * socket level. If the message *msg* is allowed to pass (i.e. if
1696 * the verdict eBPF program returns **SK_PASS**), redirect it to
1697 * the socket referenced by *map* (of type
1698 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1699 * egress interfaces can be used for redirection. The
1700 * **BPF_F_INGRESS** value in *flags* is used to make the
1701 * distinction (ingress path is selected if the flag is present,
1702 * egress path otherwise). This is the only flag supported for now.
1704 * **SK_PASS** on success, or **SK_DROP** on error.
1706 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1708 * For socket policies, apply the verdict of the eBPF program to
1709 * the next *bytes* (number of bytes) of message *msg*.
1711 * For example, this helper can be used in the following cases:
1713 * * A single **sendmsg**\ () or **sendfile**\ () system call
1714 * contains multiple logical messages that the eBPF program is
1715 * supposed to read and for which it should apply a verdict.
1716 * * An eBPF program only cares to read the first *bytes* of a
1717 * *msg*. If the message has a large payload, then setting up
1718 * and calling the eBPF program repeatedly for all bytes, even
1719 * though the verdict is already known, would create unnecessary
1722 * When called from within an eBPF program, the helper sets a
1723 * counter internal to the BPF infrastructure, that is used to
1724 * apply the last verdict to the next *bytes*. If *bytes* is
1725 * smaller than the current data being processed from a
1726 * **sendmsg**\ () or **sendfile**\ () system call, the first
1727 * *bytes* will be sent and the eBPF program will be re-run with
1728 * the pointer for start of data pointing to byte number *bytes*
1729 * **+ 1**. If *bytes* is larger than the current data being
1730 * processed, then the eBPF verdict will be applied to multiple
1731 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1734 * Note that if a socket closes with the internal counter holding
1735 * a non-zero value, this is not a problem because data is not
1736 * being buffered for *bytes* and is sent as it is received.
1740 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1742 * For socket policies, prevent the execution of the verdict eBPF
1743 * program for message *msg* until *bytes* (byte number) have been
1746 * This can be used when one needs a specific number of bytes
1747 * before a verdict can be assigned, even if the data spans
1748 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1749 * case would be a user calling **sendmsg**\ () repeatedly with
1750 * 1-byte long message segments. Obviously, this is bad for
1751 * performance, but it is still valid. If the eBPF program needs
1752 * *bytes* bytes to validate a header, this helper can be used to
1753 * prevent the eBPF program to be called again until *bytes* have
1758 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1760 * For socket policies, pull in non-linear data from user space
1761 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1762 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1765 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1766 * *msg* it can only parse data that the (**data**, **data_end**)
1767 * pointers have already consumed. For **sendmsg**\ () hooks this
1768 * is likely the first scatterlist element. But for calls relying
1769 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1770 * be the range (**0**, **0**) because the data is shared with
1771 * user space and by default the objective is to avoid allowing
1772 * user space to modify data while (or after) eBPF verdict is
1773 * being decided. This helper can be used to pull in data and to
1774 * set the start and end pointer to given values. Data will be
1775 * copied if necessary (i.e. if data was not linear and if start
1776 * and end pointers do not point to the same chunk).
1778 * A call to this helper is susceptible to change the underlaying
1779 * packet buffer. Therefore, at load time, all checks on pointers
1780 * previously done by the verifier are invalidated and must be
1781 * performed again, if the helper is used in combination with
1782 * direct packet access.
1784 * All values for *flags* are reserved for future usage, and must
1787 * 0 on success, or a negative error in case of failure.
1789 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1791 * Bind the socket associated to *ctx* to the address pointed by
1792 * *addr*, of length *addr_len*. This allows for making outgoing
1793 * connection from the desired IP address, which can be useful for
1794 * example when all processes inside a cgroup should use one
1795 * single IP address on a host that has multiple IP configured.
1797 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1798 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1799 * **AF_INET6**). Looking for a free port to bind to can be
1800 * expensive, therefore binding to port is not permitted by the
1801 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1802 * must be set to zero.
1804 * 0 on success, or a negative error in case of failure.
1806 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1808 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1809 * only possible to shrink the packet as of this writing,
1810 * therefore *delta* must be a negative integer.
1812 * A call to this helper is susceptible to change the underlaying
1813 * packet buffer. Therefore, at load time, all checks on pointers
1814 * previously done by the verifier are invalidated and must be
1815 * performed again, if the helper is used in combination with
1816 * direct packet access.
1818 * 0 on success, or a negative error in case of failure.
1820 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1822 * Retrieve the XFRM state (IP transform framework, see also
1823 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1825 * The retrieved value is stored in the **struct bpf_xfrm_state**
1826 * pointed by *xfrm_state* and of length *size*.
1828 * All values for *flags* are reserved for future usage, and must
1831 * This helper is available only if the kernel was compiled with
1832 * **CONFIG_XFRM** configuration option.
1834 * 0 on success, or a negative error in case of failure.
1836 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1838 * Return a user or a kernel stack in bpf program provided buffer.
1839 * To achieve this, the helper needs *ctx*, which is a pointer
1840 * to the context on which the tracing program is executed.
1841 * To store the stacktrace, the bpf program provides *buf* with
1842 * a nonnegative *size*.
1844 * The last argument, *flags*, holds the number of stack frames to
1845 * skip (from 0 to 255), masked with
1846 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1847 * the following flags:
1849 * **BPF_F_USER_STACK**
1850 * Collect a user space stack instead of a kernel stack.
1851 * **BPF_F_USER_BUILD_ID**
1852 * Collect buildid+offset instead of ips for user stack,
1853 * only valid if **BPF_F_USER_STACK** is also specified.
1855 * **bpf_get_stack**\ () can collect up to
1856 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1857 * to sufficient large buffer size. Note that
1858 * this limit can be controlled with the **sysctl** program, and
1859 * that it should be manually increased in order to profile long
1860 * user stacks (such as stacks for Java programs). To do so, use:
1864 * # sysctl kernel.perf_event_max_stack=<new value>
1866 * A non-negative value equal to or less than *size* on success,
1867 * or a negative error in case of failure.
1869 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1871 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1872 * it provides an easy way to load *len* bytes from *offset*
1873 * from the packet associated to *skb*, into the buffer pointed
1874 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1875 * a fifth argument *start_header* exists in order to select a
1876 * base offset to start from. *start_header* can be one of:
1878 * **BPF_HDR_START_MAC**
1879 * Base offset to load data from is *skb*'s mac header.
1880 * **BPF_HDR_START_NET**
1881 * Base offset to load data from is *skb*'s network header.
1883 * In general, "direct packet access" is the preferred method to
1884 * access packet data, however, this helper is in particular useful
1885 * in socket filters where *skb*\ **->data** does not always point
1886 * to the start of the mac header and where "direct packet access"
1889 * 0 on success, or a negative error in case of failure.
1891 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1893 * Do FIB lookup in kernel tables using parameters in *params*.
1894 * If lookup is successful and result shows packet is to be
1895 * forwarded, the neighbor tables are searched for the nexthop.
1896 * If successful (ie., FIB lookup shows forwarding and nexthop
1897 * is resolved), the nexthop address is returned in ipv4_dst
1898 * or ipv6_dst based on family, smac is set to mac address of
1899 * egress device, dmac is set to nexthop mac address, rt_metric
1900 * is set to metric from route (IPv4/IPv6 only), and ifindex
1901 * is set to the device index of the nexthop from the FIB lookup.
1903 * *plen* argument is the size of the passed in struct.
1904 * *flags* argument can be a combination of one or more of the
1907 * **BPF_FIB_LOOKUP_DIRECT**
1908 * Do a direct table lookup vs full lookup using FIB
1910 * **BPF_FIB_LOOKUP_OUTPUT**
1911 * Perform lookup from an egress perspective (default is
1914 * *ctx* is either **struct xdp_md** for XDP programs or
1915 * **struct sk_buff** tc cls_act programs.
1917 * * < 0 if any input argument is invalid
1918 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1919 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1920 * packet is not forwarded or needs assist from full stack
1922 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1924 * Add an entry to, or update a sockhash *map* referencing sockets.
1925 * The *skops* is used as a new value for the entry associated to
1926 * *key*. *flags* is one of:
1929 * The entry for *key* must not exist in the map.
1931 * The entry for *key* must already exist in the map.
1933 * No condition on the existence of the entry for *key*.
1935 * If the *map* has eBPF programs (parser and verdict), those will
1936 * be inherited by the socket being added. If the socket is
1937 * already attached to eBPF programs, this results in an error.
1939 * 0 on success, or a negative error in case of failure.
1941 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1943 * This helper is used in programs implementing policies at the
1944 * socket level. If the message *msg* is allowed to pass (i.e. if
1945 * the verdict eBPF program returns **SK_PASS**), redirect it to
1946 * the socket referenced by *map* (of type
1947 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1948 * egress interfaces can be used for redirection. The
1949 * **BPF_F_INGRESS** value in *flags* is used to make the
1950 * distinction (ingress path is selected if the flag is present,
1951 * egress path otherwise). This is the only flag supported for now.
1953 * **SK_PASS** on success, or **SK_DROP** on error.
1955 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1957 * This helper is used in programs implementing policies at the
1958 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1959 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1960 * to the socket referenced by *map* (of type
1961 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1962 * egress interfaces can be used for redirection. The
1963 * **BPF_F_INGRESS** value in *flags* is used to make the
1964 * distinction (ingress path is selected if the flag is present,
1965 * egress otherwise). This is the only flag supported for now.
1967 * **SK_PASS** on success, or **SK_DROP** on error.
1969 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
1971 * Encapsulate the packet associated to *skb* within a Layer 3
1972 * protocol header. This header is provided in the buffer at
1973 * address *hdr*, with *len* its size in bytes. *type* indicates
1974 * the protocol of the header and can be one of:
1976 * **BPF_LWT_ENCAP_SEG6**
1977 * IPv6 encapsulation with Segment Routing Header
1978 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
1979 * the IPv6 header is computed by the kernel.
1980 * **BPF_LWT_ENCAP_SEG6_INLINE**
1981 * Only works if *skb* contains an IPv6 packet. Insert a
1982 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
1985 * A call to this helper is susceptible to change the underlaying
1986 * packet buffer. Therefore, at load time, all checks on pointers
1987 * previously done by the verifier are invalidated and must be
1988 * performed again, if the helper is used in combination with
1989 * direct packet access.
1991 * 0 on success, or a negative error in case of failure.
1993 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
1995 * Store *len* bytes from address *from* into the packet
1996 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
1997 * inside the outermost IPv6 Segment Routing Header can be
1998 * modified through this helper.
2000 * A call to this helper is susceptible to change the underlaying
2001 * packet buffer. Therefore, at load time, all checks on pointers
2002 * previously done by the verifier are invalidated and must be
2003 * performed again, if the helper is used in combination with
2004 * direct packet access.
2006 * 0 on success, or a negative error in case of failure.
2008 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2010 * Adjust the size allocated to TLVs in the outermost IPv6
2011 * Segment Routing Header contained in the packet associated to
2012 * *skb*, at position *offset* by *delta* bytes. Only offsets
2013 * after the segments are accepted. *delta* can be as well
2014 * positive (growing) as negative (shrinking).
2016 * A call to this helper is susceptible to change the underlaying
2017 * packet buffer. Therefore, at load time, all checks on pointers
2018 * previously done by the verifier are invalidated and must be
2019 * performed again, if the helper is used in combination with
2020 * direct packet access.
2022 * 0 on success, or a negative error in case of failure.
2024 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2026 * Apply an IPv6 Segment Routing action of type *action* to the
2027 * packet associated to *skb*. Each action takes a parameter
2028 * contained at address *param*, and of length *param_len* bytes.
2029 * *action* can be one of:
2031 * **SEG6_LOCAL_ACTION_END_X**
2032 * End.X action: Endpoint with Layer-3 cross-connect.
2033 * Type of *param*: **struct in6_addr**.
2034 * **SEG6_LOCAL_ACTION_END_T**
2035 * End.T action: Endpoint with specific IPv6 table lookup.
2036 * Type of *param*: **int**.
2037 * **SEG6_LOCAL_ACTION_END_B6**
2038 * End.B6 action: Endpoint bound to an SRv6 policy.
2039 * Type of param: **struct ipv6_sr_hdr**.
2040 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2041 * End.B6.Encap action: Endpoint bound to an SRv6
2042 * encapsulation policy.
2043 * Type of param: **struct ipv6_sr_hdr**.
2045 * A call to this helper is susceptible to change the underlaying
2046 * packet buffer. Therefore, at load time, all checks on pointers
2047 * previously done by the verifier are invalidated and must be
2048 * performed again, if the helper is used in combination with
2049 * direct packet access.
2051 * 0 on success, or a negative error in case of failure.
2053 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2055 * This helper is used in programs implementing IR decoding, to
2056 * report a successfully decoded key press with *scancode*,
2057 * *toggle* value in the given *protocol*. The scancode will be
2058 * translated to a keycode using the rc keymap, and reported as
2059 * an input key down event. After a period a key up event is
2060 * generated. This period can be extended by calling either
2061 * **bpf_rc_keydown** () again with the same values, or calling
2062 * **bpf_rc_repeat** ().
2064 * Some protocols include a toggle bit, in case the button was
2065 * released and pressed again between consecutive scancodes.
2067 * The *ctx* should point to the lirc sample as passed into
2070 * The *protocol* is the decoded protocol number (see
2071 * **enum rc_proto** for some predefined values).
2073 * This helper is only available is the kernel was compiled with
2074 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2079 * int bpf_rc_repeat(void *ctx)
2081 * This helper is used in programs implementing IR decoding, to
2082 * report a successfully decoded repeat key message. This delays
2083 * the generation of a key up event for previously generated
2086 * Some IR protocols like NEC have a special IR message for
2087 * repeating last button, for when a button is held down.
2089 * The *ctx* should point to the lirc sample as passed into
2092 * This helper is only available is the kernel was compiled with
2093 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2098 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2100 * Return the cgroup v2 id of the socket associated with the *skb*.
2101 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2102 * helper for cgroup v1 by providing a tag resp. identifier that
2103 * can be matched on or used for map lookups e.g. to implement
2104 * policy. The cgroup v2 id of a given path in the hierarchy is
2105 * exposed in user space through the f_handle API in order to get
2106 * to the same 64-bit id.
2108 * This helper can be used on TC egress path, but not on ingress,
2109 * and is available only if the kernel was compiled with the
2110 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2112 * The id is returned or 0 in case the id could not be retrieved.
2114 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2116 * Return id of cgroup v2 that is ancestor of cgroup associated
2117 * with the *skb* at the *ancestor_level*. The root cgroup is at
2118 * *ancestor_level* zero and each step down the hierarchy
2119 * increments the level. If *ancestor_level* == level of cgroup
2120 * associated with *skb*, then return value will be same as that
2121 * of **bpf_skb_cgroup_id**\ ().
2123 * The helper is useful to implement policies based on cgroups
2124 * that are upper in hierarchy than immediate cgroup associated
2127 * The format of returned id and helper limitations are same as in
2128 * **bpf_skb_cgroup_id**\ ().
2130 * The id is returned or 0 in case the id could not be retrieved.
2132 * u64 bpf_get_current_cgroup_id(void)
2134 * A 64-bit integer containing the current cgroup id based
2135 * on the cgroup within which the current task is running.
2137 * void* get_local_storage(void *map, u64 flags)
2139 * Get the pointer to the local storage area.
2140 * The type and the size of the local storage is defined
2141 * by the *map* argument.
2142 * The *flags* meaning is specific for each map type,
2143 * and has to be 0 for cgroup local storage.
2145 * Depending on the bpf program type, a local storage area
2146 * can be shared between multiple instances of the bpf program,
2147 * running simultaneously.
2149 * A user should care about the synchronization by himself.
2150 * For example, by using the BPF_STX_XADD instruction to alter
2153 * Pointer to the local storage area.
2155 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2157 * Select a SO_REUSEPORT sk from a BPF_MAP_TYPE_REUSEPORT_ARRAY map
2158 * It checks the selected sk is matching the incoming
2159 * request in the skb.
2161 * 0 on success, or a negative error in case of failure.
2163 #define __BPF_FUNC_MAPPER(FN) \
2165 FN(map_lookup_elem), \
2166 FN(map_update_elem), \
2167 FN(map_delete_elem), \
2171 FN(get_prandom_u32), \
2172 FN(get_smp_processor_id), \
2173 FN(skb_store_bytes), \
2174 FN(l3_csum_replace), \
2175 FN(l4_csum_replace), \
2177 FN(clone_redirect), \
2178 FN(get_current_pid_tgid), \
2179 FN(get_current_uid_gid), \
2180 FN(get_current_comm), \
2181 FN(get_cgroup_classid), \
2182 FN(skb_vlan_push), \
2184 FN(skb_get_tunnel_key), \
2185 FN(skb_set_tunnel_key), \
2186 FN(perf_event_read), \
2188 FN(get_route_realm), \
2189 FN(perf_event_output), \
2190 FN(skb_load_bytes), \
2193 FN(skb_get_tunnel_opt), \
2194 FN(skb_set_tunnel_opt), \
2195 FN(skb_change_proto), \
2196 FN(skb_change_type), \
2197 FN(skb_under_cgroup), \
2198 FN(get_hash_recalc), \
2199 FN(get_current_task), \
2200 FN(probe_write_user), \
2201 FN(current_task_under_cgroup), \
2202 FN(skb_change_tail), \
2203 FN(skb_pull_data), \
2205 FN(set_hash_invalid), \
2206 FN(get_numa_node_id), \
2207 FN(skb_change_head), \
2208 FN(xdp_adjust_head), \
2209 FN(probe_read_str), \
2210 FN(get_socket_cookie), \
2211 FN(get_socket_uid), \
2214 FN(skb_adjust_room), \
2216 FN(sk_redirect_map), \
2217 FN(sock_map_update), \
2218 FN(xdp_adjust_meta), \
2219 FN(perf_event_read_value), \
2220 FN(perf_prog_read_value), \
2222 FN(override_return), \
2223 FN(sock_ops_cb_flags_set), \
2224 FN(msg_redirect_map), \
2225 FN(msg_apply_bytes), \
2226 FN(msg_cork_bytes), \
2227 FN(msg_pull_data), \
2229 FN(xdp_adjust_tail), \
2230 FN(skb_get_xfrm_state), \
2232 FN(skb_load_bytes_relative), \
2234 FN(sock_hash_update), \
2235 FN(msg_redirect_hash), \
2236 FN(sk_redirect_hash), \
2237 FN(lwt_push_encap), \
2238 FN(lwt_seg6_store_bytes), \
2239 FN(lwt_seg6_adjust_srh), \
2240 FN(lwt_seg6_action), \
2243 FN(skb_cgroup_id), \
2244 FN(get_current_cgroup_id), \
2245 FN(get_local_storage), \
2246 FN(sk_select_reuseport), \
2247 FN(skb_ancestor_cgroup_id),
2249 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2250 * function eBPF program intends to call
2252 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2254 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2257 #undef __BPF_ENUM_FN
2259 /* All flags used by eBPF helper functions, placed here. */
2261 /* BPF_FUNC_skb_store_bytes flags. */
2262 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2263 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2265 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2266 * First 4 bits are for passing the header field size.
2268 #define BPF_F_HDR_FIELD_MASK 0xfULL
2270 /* BPF_FUNC_l4_csum_replace flags. */
2271 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2272 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2273 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2275 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2276 #define BPF_F_INGRESS (1ULL << 0)
2278 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2279 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2281 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2282 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2283 #define BPF_F_USER_STACK (1ULL << 8)
2284 /* flags used by BPF_FUNC_get_stackid only. */
2285 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2286 #define BPF_F_REUSE_STACKID (1ULL << 10)
2287 /* flags used by BPF_FUNC_get_stack only. */
2288 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2290 /* BPF_FUNC_skb_set_tunnel_key flags. */
2291 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2292 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2293 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2295 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2296 * BPF_FUNC_perf_event_read_value flags.
2298 #define BPF_F_INDEX_MASK 0xffffffffULL
2299 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2300 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2301 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2303 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2304 enum bpf_adj_room_mode {
2308 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2309 enum bpf_hdr_start_off {
2314 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2315 enum bpf_lwt_encap_mode {
2317 BPF_LWT_ENCAP_SEG6_INLINE
2320 /* user accessible mirror of in-kernel sk_buff.
2321 * new fields can only be added to the end of this structure
2327 __u32 queue_mapping;
2333 __u32 ingress_ifindex;
2343 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2345 __u32 remote_ip4; /* Stored in network byte order */
2346 __u32 local_ip4; /* Stored in network byte order */
2347 __u32 remote_ip6[4]; /* Stored in network byte order */
2348 __u32 local_ip6[4]; /* Stored in network byte order */
2349 __u32 remote_port; /* Stored in network byte order */
2350 __u32 local_port; /* stored in host byte order */
2356 struct bpf_tunnel_key {
2360 __u32 remote_ipv6[4];
2364 __u16 tunnel_ext; /* Padding, future use. */
2368 /* user accessible mirror of in-kernel xfrm_state.
2369 * new fields can only be added to the end of this structure
2371 struct bpf_xfrm_state {
2373 __u32 spi; /* Stored in network byte order */
2375 __u16 ext; /* Padding, future use. */
2377 __u32 remote_ipv4; /* Stored in network byte order */
2378 __u32 remote_ipv6[4]; /* Stored in network byte order */
2382 /* Generic BPF return codes which all BPF program types may support.
2383 * The values are binary compatible with their TC_ACT_* counter-part to
2384 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2387 * XDP is handled seprately, see XDP_*.
2395 /* >127 are reserved for prog type specific return codes */
2405 __u32 src_ip4; /* Allows 1,2,4-byte read.
2406 * Stored in network byte order.
2408 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2409 * Stored in network byte order.
2411 __u32 src_port; /* Allows 4-byte read.
2412 * Stored in host byte order
2416 #define XDP_PACKET_HEADROOM 256
2418 /* User return codes for XDP prog type.
2419 * A valid XDP program must return one of these defined values. All other
2420 * return codes are reserved for future use. Unknown return codes will
2421 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2431 /* user accessible metadata for XDP packet hook
2432 * new fields must be added to the end of this structure
2438 /* Below access go through struct xdp_rxq_info */
2439 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2440 __u32 rx_queue_index; /* rxq->queue_index */
2448 /* user accessible metadata for SK_MSG packet hook, new fields must
2449 * be added to the end of this structure
2456 __u32 remote_ip4; /* Stored in network byte order */
2457 __u32 local_ip4; /* Stored in network byte order */
2458 __u32 remote_ip6[4]; /* Stored in network byte order */
2459 __u32 local_ip6[4]; /* Stored in network byte order */
2460 __u32 remote_port; /* Stored in network byte order */
2461 __u32 local_port; /* stored in host byte order */
2464 struct sk_reuseport_md {
2466 * Start of directly accessible data. It begins from
2467 * the tcp/udp header.
2470 void *data_end; /* End of directly accessible data */
2472 * Total length of packet (starting from the tcp/udp header).
2473 * Note that the directly accessible bytes (data_end - data)
2474 * could be less than this "len". Those bytes could be
2475 * indirectly read by a helper "bpf_skb_load_bytes()".
2479 * Eth protocol in the mac header (network byte order). e.g.
2480 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2483 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2484 __u32 bind_inany; /* Is sock bound to an INANY address? */
2485 __u32 hash; /* A hash of the packet 4 tuples */
2488 #define BPF_TAG_SIZE 8
2490 struct bpf_prog_info {
2493 __u8 tag[BPF_TAG_SIZE];
2494 __u32 jited_prog_len;
2495 __u32 xlated_prog_len;
2496 __aligned_u64 jited_prog_insns;
2497 __aligned_u64 xlated_prog_insns;
2498 __u64 load_time; /* ns since boottime */
2499 __u32 created_by_uid;
2501 __aligned_u64 map_ids;
2502 char name[BPF_OBJ_NAME_LEN];
2504 __u32 gpl_compatible:1;
2505 __u32 :31; /* alignment pad */
2508 __u32 nr_jited_ksyms;
2509 __u32 nr_jited_func_lens;
2510 __aligned_u64 jited_ksyms;
2511 __aligned_u64 jited_func_lens;
2512 } __attribute__((aligned(8)));
2514 struct bpf_map_info {
2521 char name[BPF_OBJ_NAME_LEN];
2527 __u32 btf_key_type_id;
2528 __u32 btf_value_type_id;
2529 } __attribute__((aligned(8)));
2531 struct bpf_btf_info {
2535 } __attribute__((aligned(8)));
2537 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2538 * by user and intended to be used by socket (e.g. to bind to, depends on
2539 * attach attach type).
2541 struct bpf_sock_addr {
2542 __u32 user_family; /* Allows 4-byte read, but no write. */
2543 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2544 * Stored in network byte order.
2546 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2547 * Stored in network byte order.
2549 __u32 user_port; /* Allows 4-byte read and write.
2550 * Stored in network byte order
2552 __u32 family; /* Allows 4-byte read, but no write */
2553 __u32 type; /* Allows 4-byte read, but no write */
2554 __u32 protocol; /* Allows 4-byte read, but no write */
2555 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2556 * Stored in network byte order.
2558 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2559 * Stored in network byte order.
2563 /* User bpf_sock_ops struct to access socket values and specify request ops
2564 * and their replies.
2565 * Some of this fields are in network (bigendian) byte order and may need
2566 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2567 * New fields can only be added at the end of this structure
2569 struct bpf_sock_ops {
2572 __u32 args[4]; /* Optionally passed to bpf program */
2573 __u32 reply; /* Returned by bpf program */
2574 __u32 replylong[4]; /* Optionally returned by bpf prog */
2577 __u32 remote_ip4; /* Stored in network byte order */
2578 __u32 local_ip4; /* Stored in network byte order */
2579 __u32 remote_ip6[4]; /* Stored in network byte order */
2580 __u32 local_ip6[4]; /* Stored in network byte order */
2581 __u32 remote_port; /* Stored in network byte order */
2582 __u32 local_port; /* stored in host byte order */
2583 __u32 is_fullsock; /* Some TCP fields are only valid if
2584 * there is a full socket. If not, the
2585 * fields read as zero.
2588 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2589 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2598 __u32 rate_delivered;
2599 __u32 rate_interval_us;
2602 __u32 total_retrans;
2606 __u32 data_segs_out;
2610 __u64 bytes_received;
2614 /* Definitions for bpf_sock_ops_cb_flags */
2615 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2616 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2617 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2618 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2619 * supported cb flags
2622 /* List of known BPF sock_ops operators.
2623 * New entries can only be added at the end
2627 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2628 * -1 if default value should be used
2630 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2631 * window (in packets) or -1 if default
2632 * value should be used
2634 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2635 * active connection is initialized
2637 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2638 * active connection is
2641 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2642 * passive connection is
2645 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2648 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2649 * based on the path and may be
2650 * dependent on the congestion control
2651 * algorithm. In general it indicates
2652 * a congestion threshold. RTTs above
2653 * this indicate congestion
2655 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2656 * Arg1: value of icsk_retransmits
2657 * Arg2: value of icsk_rto
2658 * Arg3: whether RTO has expired
2660 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2661 * Arg1: sequence number of 1st byte
2663 * Arg3: return value of
2664 * tcp_transmit_skb (0 => success)
2666 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2670 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
2671 * socket transition to LISTEN state.
2675 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2676 * changes between the TCP and BPF versions. Ideally this should never happen.
2677 * If it does, we need to add code to convert them before calling
2678 * the BPF sock_ops function.
2681 BPF_TCP_ESTABLISHED = 1,
2691 BPF_TCP_CLOSING, /* Now a valid state */
2692 BPF_TCP_NEW_SYN_RECV,
2694 BPF_TCP_MAX_STATES /* Leave at the end! */
2697 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
2698 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
2700 struct bpf_perf_event_value {
2706 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2707 #define BPF_DEVCG_ACC_READ (1ULL << 1)
2708 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2710 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2711 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2713 struct bpf_cgroup_dev_ctx {
2714 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2720 struct bpf_raw_tracepoint_args {
2724 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
2725 * OUTPUT: Do lookup from egress perspective; default is ingress
2727 #define BPF_FIB_LOOKUP_DIRECT (1U << 0)
2728 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
2731 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
2732 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
2733 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
2734 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
2735 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
2736 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
2737 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
2738 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
2739 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
2742 struct bpf_fib_lookup {
2743 /* input: network family for lookup (AF_INET, AF_INET6)
2744 * output: network family of egress nexthop
2748 /* set if lookup is to consider L4 data - e.g., FIB rules */
2753 /* total length of packet from network header - used for MTU check */
2756 /* input: L3 device index for lookup
2757 * output: device index from FIB lookup
2762 /* inputs to lookup */
2763 __u8 tos; /* AF_INET */
2764 __be32 flowinfo; /* AF_INET6, flow_label + priority */
2766 /* output: metric of fib result (IPv4/IPv6 only) */
2772 __u32 ipv6_src[4]; /* in6_addr; network order */
2775 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2776 * network header. output: bpf_fib_lookup sets to gateway address
2777 * if FIB lookup returns gateway route
2781 __u32 ipv6_dst[4]; /* in6_addr; network order */
2785 __be16 h_vlan_proto;
2787 __u8 smac[6]; /* ETH_ALEN */
2788 __u8 dmac[6]; /* ETH_ALEN */
2791 enum bpf_task_fd_type {
2792 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
2793 BPF_FD_TYPE_TRACEPOINT, /* tp name */
2794 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
2795 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
2796 BPF_FD_TYPE_UPROBE, /* filename + offset */
2797 BPF_FD_TYPE_URETPROBE, /* filename + offset */
2800 #endif /* _UAPI__LINUX_BPF_H__ */