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_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_XADD 0xc0 /* exclusive add */
25 #define BPF_MOV 0xb0 /* mov reg to reg */
26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28 /* change endianness of a register */
29 #define BPF_END 0xd0 /* flags for endianness conversion: */
30 #define BPF_TO_LE 0x00 /* convert to little-endian */
31 #define BPF_TO_BE 0x08 /* convert to big-endian */
32 #define BPF_FROM_LE BPF_TO_LE
33 #define BPF_FROM_BE BPF_TO_BE
36 #define BPF_JNE 0x50 /* jump != */
37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
43 #define BPF_CALL 0x80 /* function call */
44 #define BPF_EXIT 0x90 /* function return */
46 /* Register numbers */
62 /* BPF has 10 general purpose 64-bit registers and stack frame. */
63 #define MAX_BPF_REG __MAX_BPF_REG
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74 struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data[0]; /* Arbitrary size */
79 struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
84 /* BPF syscall commands, see bpf(2) man-page for details. */
99 BPF_PROG_GET_FD_BY_ID,
100 BPF_MAP_GET_FD_BY_ID,
101 BPF_OBJ_GET_INFO_BY_FD,
103 BPF_RAW_TRACEPOINT_OPEN,
105 BPF_BTF_GET_FD_BY_ID,
107 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
115 BPF_MAP_TYPE_PROG_ARRAY,
116 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
117 BPF_MAP_TYPE_PERCPU_HASH,
118 BPF_MAP_TYPE_PERCPU_ARRAY,
119 BPF_MAP_TYPE_STACK_TRACE,
120 BPF_MAP_TYPE_CGROUP_ARRAY,
121 BPF_MAP_TYPE_LRU_HASH,
122 BPF_MAP_TYPE_LRU_PERCPU_HASH,
123 BPF_MAP_TYPE_LPM_TRIE,
124 BPF_MAP_TYPE_ARRAY_OF_MAPS,
125 BPF_MAP_TYPE_HASH_OF_MAPS,
127 BPF_MAP_TYPE_SOCKMAP,
130 BPF_MAP_TYPE_SOCKHASH,
131 BPF_MAP_TYPE_CGROUP_STORAGE,
132 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
133 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
138 /* Note that tracing related programs such as
139 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
140 * are not subject to a stable API since kernel internal data
141 * structures can change from release to release and may
142 * therefore break existing tracing BPF programs. Tracing BPF
143 * programs correspond to /a/ specific kernel which is to be
144 * analyzed, and not /a/ specific kernel /and/ all future ones.
147 BPF_PROG_TYPE_UNSPEC,
148 BPF_PROG_TYPE_SOCKET_FILTER,
149 BPF_PROG_TYPE_KPROBE,
150 BPF_PROG_TYPE_SCHED_CLS,
151 BPF_PROG_TYPE_SCHED_ACT,
152 BPF_PROG_TYPE_TRACEPOINT,
154 BPF_PROG_TYPE_PERF_EVENT,
155 BPF_PROG_TYPE_CGROUP_SKB,
156 BPF_PROG_TYPE_CGROUP_SOCK,
157 BPF_PROG_TYPE_LWT_IN,
158 BPF_PROG_TYPE_LWT_OUT,
159 BPF_PROG_TYPE_LWT_XMIT,
160 BPF_PROG_TYPE_SOCK_OPS,
161 BPF_PROG_TYPE_SK_SKB,
162 BPF_PROG_TYPE_CGROUP_DEVICE,
163 BPF_PROG_TYPE_SK_MSG,
164 BPF_PROG_TYPE_RAW_TRACEPOINT,
165 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
166 BPF_PROG_TYPE_LWT_SEG6LOCAL,
167 BPF_PROG_TYPE_LIRC_MODE2,
168 BPF_PROG_TYPE_SK_REUSEPORT,
169 BPF_PROG_TYPE_FLOW_DISSECTOR,
170 BPF_PROG_TYPE_CGROUP_SYSCTL,
173 enum bpf_attach_type {
174 BPF_CGROUP_INET_INGRESS,
175 BPF_CGROUP_INET_EGRESS,
176 BPF_CGROUP_INET_SOCK_CREATE,
178 BPF_SK_SKB_STREAM_PARSER,
179 BPF_SK_SKB_STREAM_VERDICT,
182 BPF_CGROUP_INET4_BIND,
183 BPF_CGROUP_INET6_BIND,
184 BPF_CGROUP_INET4_CONNECT,
185 BPF_CGROUP_INET6_CONNECT,
186 BPF_CGROUP_INET4_POST_BIND,
187 BPF_CGROUP_INET6_POST_BIND,
188 BPF_CGROUP_UDP4_SENDMSG,
189 BPF_CGROUP_UDP6_SENDMSG,
193 __MAX_BPF_ATTACH_TYPE
196 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
198 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
200 * NONE(default): No further bpf programs allowed in the subtree.
202 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
203 * the program in this cgroup yields to sub-cgroup program.
205 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
206 * that cgroup program gets run in addition to the program in this cgroup.
208 * Only one program is allowed to be attached to a cgroup with
209 * NONE or BPF_F_ALLOW_OVERRIDE flag.
210 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
211 * release old program and attach the new one. Attach flags has to match.
213 * Multiple programs are allowed to be attached to a cgroup with
214 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
215 * (those that were attached first, run first)
216 * The programs of sub-cgroup are executed first, then programs of
217 * this cgroup and then programs of parent cgroup.
218 * When children program makes decision (like picking TCP CA or sock bind)
219 * parent program has a chance to override it.
221 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
222 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
224 * cgrp1 (MULTI progs A, B) ->
225 * cgrp2 (OVERRIDE prog C) ->
226 * cgrp3 (MULTI prog D) ->
227 * cgrp4 (OVERRIDE prog E) ->
228 * cgrp5 (NONE prog F)
229 * the event in cgrp5 triggers execution of F,D,A,B in that order.
230 * if prog F is detached, the execution is E,D,A,B
231 * if prog F and D are detached, the execution is E,A,B
232 * if prog F, E and D are detached, the execution is C,A,B
234 * All eligible programs are executed regardless of return code from
237 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
238 #define BPF_F_ALLOW_MULTI (1U << 1)
240 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
241 * verifier will perform strict alignment checking as if the kernel
242 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
243 * and NET_IP_ALIGN defined to 2.
245 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
247 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
248 * verifier will allow any alignment whatsoever. On platforms
249 * with strict alignment requirements for loads ands stores (such
250 * as sparc and mips) the verifier validates that all loads and
251 * stores provably follow this requirement. This flag turns that
252 * checking and enforcement off.
254 * It is mostly used for testing when we want to validate the
255 * context and memory access aspects of the verifier, but because
256 * of an unaligned access the alignment check would trigger before
257 * the one we are interested in.
259 #define BPF_F_ANY_ALIGNMENT (1U << 1)
261 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
264 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
265 * insn[0].imm: map fd map fd
266 * insn[1].imm: 0 offset into value
269 * ldimm64 rewrite: address of map address of map[0]+offset
270 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
272 #define BPF_PSEUDO_MAP_FD 1
273 #define BPF_PSEUDO_MAP_VALUE 2
275 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
276 * offset to another bpf function
278 #define BPF_PSEUDO_CALL 1
280 /* flags for BPF_MAP_UPDATE_ELEM command */
281 #define BPF_ANY 0 /* create new element or update existing */
282 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
283 #define BPF_EXIST 2 /* update existing element */
284 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
286 /* flags for BPF_MAP_CREATE command */
287 #define BPF_F_NO_PREALLOC (1U << 0)
288 /* Instead of having one common LRU list in the
289 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
290 * which can scale and perform better.
291 * Note, the LRU nodes (including free nodes) cannot be moved
292 * across different LRU lists.
294 #define BPF_F_NO_COMMON_LRU (1U << 1)
295 /* Specify numa node during map creation */
296 #define BPF_F_NUMA_NODE (1U << 2)
298 #define BPF_OBJ_NAME_LEN 16U
300 /* Flags for accessing BPF object from syscall side. */
301 #define BPF_F_RDONLY (1U << 3)
302 #define BPF_F_WRONLY (1U << 4)
304 /* Flag for stack_map, store build_id+offset instead of pointer */
305 #define BPF_F_STACK_BUILD_ID (1U << 5)
307 /* Zero-initialize hash function seed. This should only be used for testing. */
308 #define BPF_F_ZERO_SEED (1U << 6)
310 /* Flags for accessing BPF object from program side. */
311 #define BPF_F_RDONLY_PROG (1U << 7)
312 #define BPF_F_WRONLY_PROG (1U << 8)
314 /* flags for BPF_PROG_QUERY */
315 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
317 enum bpf_stack_build_id_status {
318 /* user space need an empty entry to identify end of a trace */
319 BPF_STACK_BUILD_ID_EMPTY = 0,
320 /* with valid build_id and offset */
321 BPF_STACK_BUILD_ID_VALID = 1,
322 /* couldn't get build_id, fallback to ip */
323 BPF_STACK_BUILD_ID_IP = 2,
326 #define BPF_BUILD_ID_SIZE 20
327 struct bpf_stack_build_id {
329 unsigned char build_id[BPF_BUILD_ID_SIZE];
337 struct { /* anonymous struct used by BPF_MAP_CREATE command */
338 __u32 map_type; /* one of enum bpf_map_type */
339 __u32 key_size; /* size of key in bytes */
340 __u32 value_size; /* size of value in bytes */
341 __u32 max_entries; /* max number of entries in a map */
342 __u32 map_flags; /* BPF_MAP_CREATE related
343 * flags defined above.
345 __u32 inner_map_fd; /* fd pointing to the inner map */
346 __u32 numa_node; /* numa node (effective only if
347 * BPF_F_NUMA_NODE is set).
349 char map_name[BPF_OBJ_NAME_LEN];
350 __u32 map_ifindex; /* ifindex of netdev to create on */
351 __u32 btf_fd; /* fd pointing to a BTF type data */
352 __u32 btf_key_type_id; /* BTF type_id of the key */
353 __u32 btf_value_type_id; /* BTF type_id of the value */
356 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
361 __aligned_u64 next_key;
366 struct { /* anonymous struct used by BPF_PROG_LOAD command */
367 __u32 prog_type; /* one of enum bpf_prog_type */
370 __aligned_u64 license;
371 __u32 log_level; /* verbosity level of verifier */
372 __u32 log_size; /* size of user buffer */
373 __aligned_u64 log_buf; /* user supplied buffer */
374 __u32 kern_version; /* not used */
376 char prog_name[BPF_OBJ_NAME_LEN];
377 __u32 prog_ifindex; /* ifindex of netdev to prep for */
378 /* For some prog types expected attach type must be known at
379 * load time to verify attach type specific parts of prog
380 * (context accesses, allowed helpers, etc).
382 __u32 expected_attach_type;
383 __u32 prog_btf_fd; /* fd pointing to BTF type data */
384 __u32 func_info_rec_size; /* userspace bpf_func_info size */
385 __aligned_u64 func_info; /* func info */
386 __u32 func_info_cnt; /* number of bpf_func_info records */
387 __u32 line_info_rec_size; /* userspace bpf_line_info size */
388 __aligned_u64 line_info; /* line info */
389 __u32 line_info_cnt; /* number of bpf_line_info records */
392 struct { /* anonymous struct used by BPF_OBJ_* commands */
393 __aligned_u64 pathname;
398 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
399 __u32 target_fd; /* container object to attach to */
400 __u32 attach_bpf_fd; /* eBPF program to attach */
405 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
408 __u32 data_size_in; /* input: len of data_in */
409 __u32 data_size_out; /* input/output: len of data_out
410 * returns ENOSPC if data_out
413 __aligned_u64 data_in;
414 __aligned_u64 data_out;
417 __u32 ctx_size_in; /* input: len of ctx_in */
418 __u32 ctx_size_out; /* input/output: len of ctx_out
419 * returns ENOSPC if ctx_out
422 __aligned_u64 ctx_in;
423 __aligned_u64 ctx_out;
426 struct { /* anonymous struct used by BPF_*_GET_*_ID */
437 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
443 struct { /* anonymous struct used by BPF_PROG_QUERY command */
444 __u32 target_fd; /* container object to query */
448 __aligned_u64 prog_ids;
457 struct { /* anonymous struct for BPF_BTF_LOAD */
459 __aligned_u64 btf_log_buf;
466 __u32 pid; /* input: pid */
467 __u32 fd; /* input: fd */
468 __u32 flags; /* input: flags */
469 __u32 buf_len; /* input/output: buf len */
470 __aligned_u64 buf; /* input/output:
471 * tp_name for tracepoint
473 * filename for uprobe
475 __u32 prog_id; /* output: prod_id */
476 __u32 fd_type; /* output: BPF_FD_TYPE_* */
477 __u64 probe_offset; /* output: probe_offset */
478 __u64 probe_addr; /* output: probe_addr */
480 } __attribute__((aligned(8)));
482 /* The description below is an attempt at providing documentation to eBPF
483 * developers about the multiple available eBPF helper functions. It can be
484 * parsed and used to produce a manual page. The workflow is the following,
485 * and requires the rst2man utility:
487 * $ ./scripts/bpf_helpers_doc.py \
488 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
489 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
490 * $ man /tmp/bpf-helpers.7
492 * Note that in order to produce this external documentation, some RST
493 * formatting is used in the descriptions to get "bold" and "italics" in
494 * manual pages. Also note that the few trailing white spaces are
495 * intentional, removing them would break paragraphs for rst2man.
497 * Start of BPF helper function descriptions:
499 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
501 * Perform a lookup in *map* for an entry associated to *key*.
503 * Map value associated to *key*, or **NULL** if no entry was
506 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
508 * Add or update the value of the entry associated to *key* in
509 * *map* with *value*. *flags* is one of:
512 * The entry for *key* must not exist in the map.
514 * The entry for *key* must already exist in the map.
516 * No condition on the existence of the entry for *key*.
518 * Flag value **BPF_NOEXIST** cannot be used for maps of types
519 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
520 * elements always exist), the helper would return an error.
522 * 0 on success, or a negative error in case of failure.
524 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
526 * Delete entry with *key* from *map*.
528 * 0 on success, or a negative error in case of failure.
530 * int bpf_probe_read(void *dst, u32 size, const void *src)
532 * For tracing programs, safely attempt to read *size* bytes from
533 * address *src* and store the data in *dst*.
535 * 0 on success, or a negative error in case of failure.
537 * u64 bpf_ktime_get_ns(void)
539 * Return the time elapsed since system boot, in nanoseconds.
543 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
545 * This helper is a "printk()-like" facility for debugging. It
546 * prints a message defined by format *fmt* (of size *fmt_size*)
547 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
548 * available. It can take up to three additional **u64**
549 * arguments (as an eBPF helpers, the total number of arguments is
552 * Each time the helper is called, it appends a line to the trace.
553 * The format of the trace is customizable, and the exact output
554 * one will get depends on the options set in
555 * *\/sys/kernel/debug/tracing/trace_options* (see also the
556 * *README* file under the same directory). However, it usually
557 * defaults to something like:
561 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
565 * * ``telnet`` is the name of the current task.
566 * * ``470`` is the PID of the current task.
567 * * ``001`` is the CPU number on which the task is
569 * * In ``.N..``, each character refers to a set of
570 * options (whether irqs are enabled, scheduling
571 * options, whether hard/softirqs are running, level of
572 * preempt_disabled respectively). **N** means that
573 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
575 * * ``419421.045894`` is a timestamp.
576 * * ``0x00000001`` is a fake value used by BPF for the
577 * instruction pointer register.
578 * * ``<formatted msg>`` is the message formatted with
581 * The conversion specifiers supported by *fmt* are similar, but
582 * more limited than for printk(). They are **%d**, **%i**,
583 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
584 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
585 * of field, padding with zeroes, etc.) is available, and the
586 * helper will return **-EINVAL** (but print nothing) if it
587 * encounters an unknown specifier.
589 * Also, note that **bpf_trace_printk**\ () is slow, and should
590 * only be used for debugging purposes. For this reason, a notice
591 * bloc (spanning several lines) is printed to kernel logs and
592 * states that the helper should not be used "for production use"
593 * the first time this helper is used (or more precisely, when
594 * **trace_printk**\ () buffers are allocated). For passing values
595 * to user space, perf events should be preferred.
597 * The number of bytes written to the buffer, or a negative error
598 * in case of failure.
600 * u32 bpf_get_prandom_u32(void)
602 * Get a pseudo-random number.
604 * From a security point of view, this helper uses its own
605 * pseudo-random internal state, and cannot be used to infer the
606 * seed of other random functions in the kernel. However, it is
607 * essential to note that the generator used by the helper is not
608 * cryptographically secure.
610 * A random 32-bit unsigned value.
612 * u32 bpf_get_smp_processor_id(void)
614 * Get the SMP (symmetric multiprocessing) processor id. Note that
615 * all programs run with preemption disabled, which means that the
616 * SMP processor id is stable during all the execution of the
619 * The SMP id of the processor running the program.
621 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
623 * Store *len* bytes from address *from* into the packet
624 * associated to *skb*, at *offset*. *flags* are a combination of
625 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
626 * checksum for the packet after storing the bytes) and
627 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
628 * **->swhash** and *skb*\ **->l4hash** to 0).
630 * A call to this helper is susceptible to change the underlaying
631 * packet buffer. Therefore, at load time, all checks on pointers
632 * previously done by the verifier are invalidated and must be
633 * performed again, if the helper is used in combination with
634 * direct packet access.
636 * 0 on success, or a negative error in case of failure.
638 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
640 * Recompute the layer 3 (e.g. IP) checksum for the packet
641 * associated to *skb*. Computation is incremental, so the helper
642 * must know the former value of the header field that was
643 * modified (*from*), the new value of this field (*to*), and the
644 * number of bytes (2 or 4) for this field, stored in *size*.
645 * Alternatively, it is possible to store the difference between
646 * the previous and the new values of the header field in *to*, by
647 * setting *from* and *size* to 0. For both methods, *offset*
648 * indicates the location of the IP checksum within the packet.
650 * This helper works in combination with **bpf_csum_diff**\ (),
651 * which does not update the checksum in-place, but offers more
652 * flexibility and can handle sizes larger than 2 or 4 for the
653 * checksum to update.
655 * A call to this helper is susceptible to change the underlaying
656 * packet buffer. Therefore, at load time, all checks on pointers
657 * previously done by the verifier are invalidated and must be
658 * performed again, if the helper is used in combination with
659 * direct packet access.
661 * 0 on success, or a negative error in case of failure.
663 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
665 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
666 * packet associated to *skb*. Computation is incremental, so the
667 * helper must know the former value of the header field that was
668 * modified (*from*), the new value of this field (*to*), and the
669 * number of bytes (2 or 4) for this field, stored on the lowest
670 * four bits of *flags*. Alternatively, it is possible to store
671 * the difference between the previous and the new values of the
672 * header field in *to*, by setting *from* and the four lowest
673 * bits of *flags* to 0. For both methods, *offset* indicates the
674 * location of the IP checksum within the packet. In addition to
675 * the size of the field, *flags* can be added (bitwise OR) actual
676 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
677 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
678 * for updates resulting in a null checksum the value is set to
679 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
680 * the checksum is to be computed against a pseudo-header.
682 * This helper works in combination with **bpf_csum_diff**\ (),
683 * which does not update the checksum in-place, but offers more
684 * flexibility and can handle sizes larger than 2 or 4 for the
685 * checksum to update.
687 * A call to this helper is susceptible to change the underlaying
688 * packet buffer. Therefore, at load time, all checks on pointers
689 * previously done by the verifier are invalidated and must be
690 * performed again, if the helper is used in combination with
691 * direct packet access.
693 * 0 on success, or a negative error in case of failure.
695 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
697 * This special helper is used to trigger a "tail call", or in
698 * other words, to jump into another eBPF program. The same stack
699 * frame is used (but values on stack and in registers for the
700 * caller are not accessible to the callee). This mechanism allows
701 * for program chaining, either for raising the maximum number of
702 * available eBPF instructions, or to execute given programs in
703 * conditional blocks. For security reasons, there is an upper
704 * limit to the number of successive tail calls that can be
707 * Upon call of this helper, the program attempts to jump into a
708 * program referenced at index *index* in *prog_array_map*, a
709 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
710 * *ctx*, a pointer to the context.
712 * If the call succeeds, the kernel immediately runs the first
713 * instruction of the new program. This is not a function call,
714 * and it never returns to the previous program. If the call
715 * fails, then the helper has no effect, and the caller continues
716 * to run its subsequent instructions. A call can fail if the
717 * destination program for the jump does not exist (i.e. *index*
718 * is superior to the number of entries in *prog_array_map*), or
719 * if the maximum number of tail calls has been reached for this
720 * chain of programs. This limit is defined in the kernel by the
721 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
722 * which is currently set to 32.
724 * 0 on success, or a negative error in case of failure.
726 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
728 * Clone and redirect the packet associated to *skb* to another
729 * net device of index *ifindex*. Both ingress and egress
730 * interfaces can be used for redirection. The **BPF_F_INGRESS**
731 * value in *flags* is used to make the distinction (ingress path
732 * is selected if the flag is present, egress path otherwise).
733 * This is the only flag supported for now.
735 * In comparison with **bpf_redirect**\ () helper,
736 * **bpf_clone_redirect**\ () has the associated cost of
737 * duplicating the packet buffer, but this can be executed out of
738 * the eBPF program. Conversely, **bpf_redirect**\ () is more
739 * efficient, but it is handled through an action code where the
740 * redirection happens only after the eBPF program has returned.
742 * A call to this helper is susceptible to change the underlaying
743 * packet buffer. Therefore, at load time, all checks on pointers
744 * previously done by the verifier are invalidated and must be
745 * performed again, if the helper is used in combination with
746 * direct packet access.
748 * 0 on success, or a negative error in case of failure.
750 * u64 bpf_get_current_pid_tgid(void)
752 * A 64-bit integer containing the current tgid and pid, and
754 * *current_task*\ **->tgid << 32 \|**
755 * *current_task*\ **->pid**.
757 * u64 bpf_get_current_uid_gid(void)
759 * A 64-bit integer containing the current GID and UID, and
760 * created as such: *current_gid* **<< 32 \|** *current_uid*.
762 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
764 * Copy the **comm** attribute of the current task into *buf* of
765 * *size_of_buf*. The **comm** attribute contains the name of
766 * the executable (excluding the path) for the current task. The
767 * *size_of_buf* must be strictly positive. On success, the
768 * helper makes sure that the *buf* is NUL-terminated. On failure,
769 * it is filled with zeroes.
771 * 0 on success, or a negative error in case of failure.
773 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
775 * Retrieve the classid for the current task, i.e. for the net_cls
776 * cgroup to which *skb* belongs.
778 * This helper can be used on TC egress path, but not on ingress.
780 * The net_cls cgroup provides an interface to tag network packets
781 * based on a user-provided identifier for all traffic coming from
782 * the tasks belonging to the related cgroup. See also the related
783 * kernel documentation, available from the Linux sources in file
784 * *Documentation/cgroup-v1/net_cls.txt*.
786 * The Linux kernel has two versions for cgroups: there are
787 * cgroups v1 and cgroups v2. Both are available to users, who can
788 * use a mixture of them, but note that the net_cls cgroup is for
789 * cgroup v1 only. This makes it incompatible with BPF programs
790 * run on cgroups, which is a cgroup-v2-only feature (a socket can
791 * only hold data for one version of cgroups at a time).
793 * This helper is only available is the kernel was compiled with
794 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
795 * "**y**" or to "**m**".
797 * The classid, or 0 for the default unconfigured classid.
799 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
801 * Push a *vlan_tci* (VLAN tag control information) of protocol
802 * *vlan_proto* to the packet associated to *skb*, then update
803 * the checksum. Note that if *vlan_proto* is different from
804 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
805 * be **ETH_P_8021Q**.
807 * A call to this helper is susceptible to change the underlaying
808 * packet buffer. Therefore, at load time, all checks on pointers
809 * previously done by the verifier are invalidated and must be
810 * performed again, if the helper is used in combination with
811 * direct packet access.
813 * 0 on success, or a negative error in case of failure.
815 * int bpf_skb_vlan_pop(struct sk_buff *skb)
817 * Pop a VLAN header from the packet associated to *skb*.
819 * A call to this helper is susceptible to change the underlaying
820 * packet buffer. Therefore, at load time, all checks on pointers
821 * previously done by the verifier are invalidated and must be
822 * performed again, if the helper is used in combination with
823 * direct packet access.
825 * 0 on success, or a negative error in case of failure.
827 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
829 * Get tunnel metadata. This helper takes a pointer *key* to an
830 * empty **struct bpf_tunnel_key** of **size**, that will be
831 * filled with tunnel metadata for the packet associated to *skb*.
832 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
833 * indicates that the tunnel is based on IPv6 protocol instead of
836 * The **struct bpf_tunnel_key** is an object that generalizes the
837 * principal parameters used by various tunneling protocols into a
838 * single struct. This way, it can be used to easily make a
839 * decision based on the contents of the encapsulation header,
840 * "summarized" in this struct. In particular, it holds the IP
841 * address of the remote end (IPv4 or IPv6, depending on the case)
842 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
843 * this struct exposes the *key*\ **->tunnel_id**, which is
844 * generally mapped to a VNI (Virtual Network Identifier), making
845 * it programmable together with the **bpf_skb_set_tunnel_key**\
848 * Let's imagine that the following code is part of a program
849 * attached to the TC ingress interface, on one end of a GRE
850 * tunnel, and is supposed to filter out all messages coming from
851 * remote ends with IPv4 address other than 10.0.0.1:
856 * struct bpf_tunnel_key key = {};
858 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
860 * return TC_ACT_SHOT; // drop packet
862 * if (key.remote_ipv4 != 0x0a000001)
863 * return TC_ACT_SHOT; // drop packet
865 * return TC_ACT_OK; // accept packet
867 * This interface can also be used with all encapsulation devices
868 * that can operate in "collect metadata" mode: instead of having
869 * one network device per specific configuration, the "collect
870 * metadata" mode only requires a single device where the
871 * configuration can be extracted from this helper.
873 * This can be used together with various tunnels such as VXLan,
874 * Geneve, GRE or IP in IP (IPIP).
876 * 0 on success, or a negative error in case of failure.
878 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
880 * Populate tunnel metadata for packet associated to *skb.* The
881 * tunnel metadata is set to the contents of *key*, of *size*. The
882 * *flags* can be set to a combination of the following values:
884 * **BPF_F_TUNINFO_IPV6**
885 * Indicate that the tunnel is based on IPv6 protocol
887 * **BPF_F_ZERO_CSUM_TX**
888 * For IPv4 packets, add a flag to tunnel metadata
889 * indicating that checksum computation should be skipped
890 * and checksum set to zeroes.
891 * **BPF_F_DONT_FRAGMENT**
892 * Add a flag to tunnel metadata indicating that the
893 * packet should not be fragmented.
894 * **BPF_F_SEQ_NUMBER**
895 * Add a flag to tunnel metadata indicating that a
896 * sequence number should be added to tunnel header before
897 * sending the packet. This flag was added for GRE
898 * encapsulation, but might be used with other protocols
899 * as well in the future.
901 * Here is a typical usage on the transmit path:
905 * struct bpf_tunnel_key key;
907 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
908 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
910 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
911 * helper for additional information.
913 * 0 on success, or a negative error in case of failure.
915 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
917 * Read the value of a perf event counter. This helper relies on a
918 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
919 * the perf event counter is selected when *map* is updated with
920 * perf event file descriptors. The *map* is an array whose size
921 * is the number of available CPUs, and each cell contains a value
922 * relative to one CPU. The value to retrieve is indicated by
923 * *flags*, that contains the index of the CPU to look up, masked
924 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
925 * **BPF_F_CURRENT_CPU** to indicate that the value for the
926 * current CPU should be retrieved.
928 * Note that before Linux 4.13, only hardware perf event can be
931 * Also, be aware that the newer helper
932 * **bpf_perf_event_read_value**\ () is recommended over
933 * **bpf_perf_event_read**\ () in general. The latter has some ABI
934 * quirks where error and counter value are used as a return code
935 * (which is wrong to do since ranges may overlap). This issue is
936 * fixed with **bpf_perf_event_read_value**\ (), which at the same
937 * time provides more features over the **bpf_perf_event_read**\
938 * () interface. Please refer to the description of
939 * **bpf_perf_event_read_value**\ () for details.
941 * The value of the perf event counter read from the map, or a
942 * negative error code in case of failure.
944 * int bpf_redirect(u32 ifindex, u64 flags)
946 * Redirect the packet to another net device of index *ifindex*.
947 * This helper is somewhat similar to **bpf_clone_redirect**\
948 * (), except that the packet is not cloned, which provides
949 * increased performance.
951 * Except for XDP, both ingress and egress interfaces can be used
952 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
953 * to make the distinction (ingress path is selected if the flag
954 * is present, egress path otherwise). Currently, XDP only
955 * supports redirection to the egress interface, and accepts no
958 * The same effect can be attained with the more generic
959 * **bpf_redirect_map**\ (), which requires specific maps to be
960 * used but offers better performance.
962 * For XDP, the helper returns **XDP_REDIRECT** on success or
963 * **XDP_ABORTED** on error. For other program types, the values
964 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
967 * u32 bpf_get_route_realm(struct sk_buff *skb)
969 * Retrieve the realm or the route, that is to say the
970 * **tclassid** field of the destination for the *skb*. The
971 * indentifier retrieved is a user-provided tag, similar to the
972 * one used with the net_cls cgroup (see description for
973 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
974 * held by a route (a destination entry), not by a task.
976 * Retrieving this identifier works with the clsact TC egress hook
977 * (see also **tc-bpf(8)**), or alternatively on conventional
978 * classful egress qdiscs, but not on TC ingress path. In case of
979 * clsact TC egress hook, this has the advantage that, internally,
980 * the destination entry has not been dropped yet in the transmit
981 * path. Therefore, the destination entry does not need to be
982 * artificially held via **netif_keep_dst**\ () for a classful
983 * qdisc until the *skb* is freed.
985 * This helper is available only if the kernel was compiled with
986 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
988 * The realm of the route for the packet associated to *skb*, or 0
991 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
993 * Write raw *data* blob into a special BPF perf event held by
994 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
995 * event must have the following attributes: **PERF_SAMPLE_RAW**
996 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
997 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
999 * The *flags* are used to indicate the index in *map* for which
1000 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1001 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1002 * to indicate that the index of the current CPU core should be
1005 * The value to write, of *size*, is passed through eBPF stack and
1006 * pointed by *data*.
1008 * The context of the program *ctx* needs also be passed to the
1011 * On user space, a program willing to read the values needs to
1012 * call **perf_event_open**\ () on the perf event (either for
1013 * one or for all CPUs) and to store the file descriptor into the
1014 * *map*. This must be done before the eBPF program can send data
1015 * into it. An example is available in file
1016 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1017 * tree (the eBPF program counterpart is in
1018 * *samples/bpf/trace_output_kern.c*).
1020 * **bpf_perf_event_output**\ () achieves better performance
1021 * than **bpf_trace_printk**\ () for sharing data with user
1022 * space, and is much better suitable for streaming data from eBPF
1025 * Note that this helper is not restricted to tracing use cases
1026 * and can be used with programs attached to TC or XDP as well,
1027 * where it allows for passing data to user space listeners. Data
1030 * * Only custom structs,
1031 * * Only the packet payload, or
1032 * * A combination of both.
1034 * 0 on success, or a negative error in case of failure.
1036 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1038 * This helper was provided as an easy way to load data from a
1039 * packet. It can be used to load *len* bytes from *offset* from
1040 * the packet associated to *skb*, into the buffer pointed by
1043 * Since Linux 4.7, usage of this helper has mostly been replaced
1044 * by "direct packet access", enabling packet data to be
1045 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1046 * pointing respectively to the first byte of packet data and to
1047 * the byte after the last byte of packet data. However, it
1048 * remains useful if one wishes to read large quantities of data
1049 * at once from a packet into the eBPF stack.
1051 * 0 on success, or a negative error in case of failure.
1053 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1055 * Walk a user or a kernel stack and return its id. To achieve
1056 * this, the helper needs *ctx*, which is a pointer to the context
1057 * on which the tracing program is executed, and a pointer to a
1058 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1060 * The last argument, *flags*, holds the number of stack frames to
1061 * skip (from 0 to 255), masked with
1062 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1063 * a combination of the following flags:
1065 * **BPF_F_USER_STACK**
1066 * Collect a user space stack instead of a kernel stack.
1067 * **BPF_F_FAST_STACK_CMP**
1068 * Compare stacks by hash only.
1069 * **BPF_F_REUSE_STACKID**
1070 * If two different stacks hash into the same *stackid*,
1071 * discard the old one.
1073 * The stack id retrieved is a 32 bit long integer handle which
1074 * can be further combined with other data (including other stack
1075 * ids) and used as a key into maps. This can be useful for
1076 * generating a variety of graphs (such as flame graphs or off-cpu
1079 * For walking a stack, this helper is an improvement over
1080 * **bpf_probe_read**\ (), which can be used with unrolled loops
1081 * but is not efficient and consumes a lot of eBPF instructions.
1082 * Instead, **bpf_get_stackid**\ () can collect up to
1083 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1084 * this limit can be controlled with the **sysctl** program, and
1085 * that it should be manually increased in order to profile long
1086 * user stacks (such as stacks for Java programs). To do so, use:
1090 * # sysctl kernel.perf_event_max_stack=<new value>
1092 * The positive or null stack id on success, or a negative error
1093 * in case of failure.
1095 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1097 * Compute a checksum difference, from the raw buffer pointed by
1098 * *from*, of length *from_size* (that must be a multiple of 4),
1099 * towards the raw buffer pointed by *to*, of size *to_size*
1100 * (same remark). An optional *seed* can be added to the value
1101 * (this can be cascaded, the seed may come from a previous call
1104 * This is flexible enough to be used in several ways:
1106 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1107 * checksum, it can be used when pushing new data.
1108 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1109 * checksum, it can be used when removing data from a packet.
1110 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1111 * can be used to compute a diff. Note that *from_size* and
1112 * *to_size* do not need to be equal.
1114 * This helper can be used in combination with
1115 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1116 * which one can feed in the difference computed with
1117 * **bpf_csum_diff**\ ().
1119 * The checksum result, or a negative error code in case of
1122 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1124 * Retrieve tunnel options metadata for the packet associated to
1125 * *skb*, and store the raw tunnel option data to the buffer *opt*
1128 * This helper can be used with encapsulation devices that can
1129 * operate in "collect metadata" mode (please refer to the related
1130 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1131 * more details). A particular example where this can be used is
1132 * in combination with the Geneve encapsulation protocol, where it
1133 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1134 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1135 * the eBPF program. This allows for full customization of these
1138 * The size of the option data retrieved.
1140 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1142 * Set tunnel options metadata for the packet associated to *skb*
1143 * to the option data contained in the raw buffer *opt* of *size*.
1145 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1146 * helper for additional information.
1148 * 0 on success, or a negative error in case of failure.
1150 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1152 * Change the protocol of the *skb* to *proto*. Currently
1153 * supported are transition from IPv4 to IPv6, and from IPv6 to
1154 * IPv4. The helper takes care of the groundwork for the
1155 * transition, including resizing the socket buffer. The eBPF
1156 * program is expected to fill the new headers, if any, via
1157 * **skb_store_bytes**\ () and to recompute the checksums with
1158 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1159 * (). The main case for this helper is to perform NAT64
1160 * operations out of an eBPF program.
1162 * Internally, the GSO type is marked as dodgy so that headers are
1163 * checked and segments are recalculated by the GSO/GRO engine.
1164 * The size for GSO target is adapted as well.
1166 * All values for *flags* are reserved for future usage, and must
1169 * A call to this helper is susceptible to change the underlaying
1170 * packet buffer. Therefore, at load time, all checks on pointers
1171 * previously done by the verifier are invalidated and must be
1172 * performed again, if the helper is used in combination with
1173 * direct packet access.
1175 * 0 on success, or a negative error in case of failure.
1177 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1179 * Change the packet type for the packet associated to *skb*. This
1180 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1181 * the eBPF program does not have a write access to *skb*\
1182 * **->pkt_type** beside this helper. Using a helper here allows
1183 * for graceful handling of errors.
1185 * The major use case is to change incoming *skb*s to
1186 * **PACKET_HOST** in a programmatic way instead of having to
1187 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1190 * Note that *type* only allows certain values. At this time, they
1195 * **PACKET_BROADCAST**
1196 * Send packet to all.
1197 * **PACKET_MULTICAST**
1198 * Send packet to group.
1199 * **PACKET_OTHERHOST**
1200 * Send packet to someone else.
1202 * 0 on success, or a negative error in case of failure.
1204 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1206 * Check whether *skb* is a descendant of the cgroup2 held by
1207 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1209 * The return value depends on the result of the test, and can be:
1211 * * 0, if the *skb* failed the cgroup2 descendant test.
1212 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1213 * * A negative error code, if an error occurred.
1215 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1217 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1218 * not set, in particular if the hash was cleared due to mangling,
1219 * recompute this hash. Later accesses to the hash can be done
1220 * directly with *skb*\ **->hash**.
1222 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1223 * prototype with **bpf_skb_change_proto**\ (), or calling
1224 * **bpf_skb_store_bytes**\ () with the
1225 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1226 * the hash and to trigger a new computation for the next call to
1227 * **bpf_get_hash_recalc**\ ().
1231 * u64 bpf_get_current_task(void)
1233 * A pointer to the current task struct.
1235 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1237 * Attempt in a safe way to write *len* bytes from the buffer
1238 * *src* to *dst* in memory. It only works for threads that are in
1239 * user context, and *dst* must be a valid user space address.
1241 * This helper should not be used to implement any kind of
1242 * security mechanism because of TOC-TOU attacks, but rather to
1243 * debug, divert, and manipulate execution of semi-cooperative
1246 * Keep in mind that this feature is meant for experiments, and it
1247 * has a risk of crashing the system and running programs.
1248 * Therefore, when an eBPF program using this helper is attached,
1249 * a warning including PID and process name is printed to kernel
1252 * 0 on success, or a negative error in case of failure.
1254 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1256 * Check whether the probe is being run is the context of a given
1257 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1258 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1260 * The return value depends on the result of the test, and can be:
1262 * * 0, if the *skb* task belongs to the cgroup2.
1263 * * 1, if the *skb* task does not belong to the cgroup2.
1264 * * A negative error code, if an error occurred.
1266 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1268 * Resize (trim or grow) the packet associated to *skb* to the
1269 * new *len*. The *flags* are reserved for future usage, and must
1272 * The basic idea is that the helper performs the needed work to
1273 * change the size of the packet, then the eBPF program rewrites
1274 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1275 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1276 * and others. This helper is a slow path utility intended for
1277 * replies with control messages. And because it is targeted for
1278 * slow path, the helper itself can afford to be slow: it
1279 * implicitly linearizes, unclones and drops offloads from the
1282 * A call to this helper is susceptible to change the underlaying
1283 * packet buffer. Therefore, at load time, all checks on pointers
1284 * previously done by the verifier are invalidated and must be
1285 * performed again, if the helper is used in combination with
1286 * direct packet access.
1288 * 0 on success, or a negative error in case of failure.
1290 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1292 * Pull in non-linear data in case the *skb* is non-linear and not
1293 * all of *len* are part of the linear section. Make *len* bytes
1294 * from *skb* readable and writable. If a zero value is passed for
1295 * *len*, then the whole length of the *skb* is pulled.
1297 * This helper is only needed for reading and writing with direct
1300 * For direct packet access, testing that offsets to access
1301 * are within packet boundaries (test on *skb*\ **->data_end**) is
1302 * susceptible to fail if offsets are invalid, or if the requested
1303 * data is in non-linear parts of the *skb*. On failure the
1304 * program can just bail out, or in the case of a non-linear
1305 * buffer, use a helper to make the data available. The
1306 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1307 * the data. Another one consists in using **bpf_skb_pull_data**
1308 * to pull in once the non-linear parts, then retesting and
1309 * eventually access the data.
1311 * At the same time, this also makes sure the *skb* is uncloned,
1312 * which is a necessary condition for direct write. As this needs
1313 * to be an invariant for the write part only, the verifier
1314 * detects writes and adds a prologue that is calling
1315 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1316 * the very beginning in case it is indeed cloned.
1318 * A call to this helper is susceptible to change the underlaying
1319 * packet buffer. Therefore, at load time, all checks on pointers
1320 * previously done by the verifier are invalidated and must be
1321 * performed again, if the helper is used in combination with
1322 * direct packet access.
1324 * 0 on success, or a negative error in case of failure.
1326 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1328 * Add the checksum *csum* into *skb*\ **->csum** in case the
1329 * driver has supplied a checksum for the entire packet into that
1330 * field. Return an error otherwise. This helper is intended to be
1331 * used in combination with **bpf_csum_diff**\ (), in particular
1332 * when the checksum needs to be updated after data has been
1333 * written into the packet through direct packet access.
1335 * The checksum on success, or a negative error code in case of
1338 * void bpf_set_hash_invalid(struct sk_buff *skb)
1340 * Invalidate the current *skb*\ **->hash**. It can be used after
1341 * mangling on headers through direct packet access, in order to
1342 * indicate that the hash is outdated and to trigger a
1343 * recalculation the next time the kernel tries to access this
1344 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1346 * int bpf_get_numa_node_id(void)
1348 * Return the id of the current NUMA node. The primary use case
1349 * for this helper is the selection of sockets for the local NUMA
1350 * node, when the program is attached to sockets using the
1351 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1352 * but the helper is also available to other eBPF program types,
1353 * similarly to **bpf_get_smp_processor_id**\ ().
1355 * The id of current NUMA node.
1357 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1359 * Grows headroom of packet associated to *skb* and adjusts the
1360 * offset of the MAC header accordingly, adding *len* bytes of
1361 * space. It automatically extends and reallocates memory as
1364 * This helper can be used on a layer 3 *skb* to push a MAC header
1365 * for redirection into a layer 2 device.
1367 * All values for *flags* are reserved for future usage, and must
1370 * A call to this helper is susceptible to change the underlaying
1371 * packet buffer. Therefore, at load time, all checks on pointers
1372 * previously done by the verifier are invalidated and must be
1373 * performed again, if the helper is used in combination with
1374 * direct packet access.
1376 * 0 on success, or a negative error in case of failure.
1378 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1380 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1381 * it is possible to use a negative value for *delta*. This helper
1382 * can be used to prepare the packet for pushing or popping
1385 * A call to this helper is susceptible to change the underlaying
1386 * packet buffer. Therefore, at load time, all checks on pointers
1387 * previously done by the verifier are invalidated and must be
1388 * performed again, if the helper is used in combination with
1389 * direct packet access.
1391 * 0 on success, or a negative error in case of failure.
1393 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1395 * Copy a NUL terminated string from an unsafe address
1396 * *unsafe_ptr* to *dst*. The *size* should include the
1397 * terminating NUL byte. In case the string length is smaller than
1398 * *size*, the target is not padded with further NUL bytes. If the
1399 * string length is larger than *size*, just *size*-1 bytes are
1400 * copied and the last byte is set to NUL.
1402 * On success, the length of the copied string is returned. This
1403 * makes this helper useful in tracing programs for reading
1404 * strings, and more importantly to get its length at runtime. See
1405 * the following snippet:
1409 * SEC("kprobe/sys_open")
1410 * void bpf_sys_open(struct pt_regs *ctx)
1412 * char buf[PATHLEN]; // PATHLEN is defined to 256
1413 * int res = bpf_probe_read_str(buf, sizeof(buf),
1416 * // Consume buf, for example push it to
1417 * // userspace via bpf_perf_event_output(); we
1418 * // can use res (the string length) as event
1419 * // size, after checking its boundaries.
1422 * In comparison, using **bpf_probe_read()** helper here instead
1423 * to read the string would require to estimate the length at
1424 * compile time, and would often result in copying more memory
1427 * Another useful use case is when parsing individual process
1428 * arguments or individual environment variables navigating
1429 * *current*\ **->mm->arg_start** and *current*\
1430 * **->mm->env_start**: using this helper and the return value,
1431 * one can quickly iterate at the right offset of the memory area.
1433 * On success, the strictly positive length of the string,
1434 * including the trailing NUL character. On error, a negative
1437 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1439 * If the **struct sk_buff** pointed by *skb* has a known socket,
1440 * retrieve the cookie (generated by the kernel) of this socket.
1441 * If no cookie has been set yet, generate a new cookie. Once
1442 * generated, the socket cookie remains stable for the life of the
1443 * socket. This helper can be useful for monitoring per socket
1444 * networking traffic statistics as it provides a unique socket
1445 * identifier per namespace.
1447 * A 8-byte long non-decreasing number on success, or 0 if the
1448 * socket field is missing inside *skb*.
1450 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1452 * Equivalent to bpf_get_socket_cookie() helper that accepts
1453 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1455 * A 8-byte long non-decreasing number.
1457 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1459 * Equivalent to bpf_get_socket_cookie() helper that accepts
1460 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1462 * A 8-byte long non-decreasing number.
1464 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1466 * The owner UID of the socket associated to *skb*. If the socket
1467 * is **NULL**, or if it is not a full socket (i.e. if it is a
1468 * time-wait or a request socket instead), **overflowuid** value
1469 * is returned (note that **overflowuid** might also be the actual
1470 * UID value for the socket).
1472 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1474 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1479 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1481 * Emulate a call to **setsockopt()** on the socket associated to
1482 * *bpf_socket*, which must be a full socket. The *level* at
1483 * which the option resides and the name *optname* of the option
1484 * must be specified, see **setsockopt(2)** for more information.
1485 * The option value of length *optlen* is pointed by *optval*.
1487 * This helper actually implements a subset of **setsockopt()**.
1488 * It supports the following *level*\ s:
1490 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1491 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1492 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1493 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1494 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1495 * **TCP_BPF_SNDCWND_CLAMP**.
1496 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1497 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1499 * 0 on success, or a negative error in case of failure.
1501 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1503 * Grow or shrink the room for data in the packet associated to
1504 * *skb* by *len_diff*, and according to the selected *mode*.
1506 * There are two supported modes at this time:
1508 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1509 * (room space is added or removed below the layer 2 header).
1511 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1512 * (room space is added or removed below the layer 3 header).
1514 * The following flags are supported at this time:
1516 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1517 * Adjusting mss in this way is not allowed for datagrams.
1519 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 **:
1520 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 **:
1521 * Any new space is reserved to hold a tunnel header.
1522 * Configure skb offsets and other fields accordingly.
1524 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE **:
1525 * * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **:
1526 * Use with ENCAP_L3 flags to further specify the tunnel type.
1528 * * **BPF_F_ADJ_ROOM_ENCAP_L2(len) **:
1529 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1530 * type; **len** is the length of the inner MAC header.
1532 * A call to this helper is susceptible to change the underlaying
1533 * packet buffer. Therefore, at load time, all checks on pointers
1534 * previously done by the verifier are invalidated and must be
1535 * performed again, if the helper is used in combination with
1536 * direct packet access.
1538 * 0 on success, or a negative error in case of failure.
1540 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1542 * Redirect the packet to the endpoint referenced by *map* at
1543 * index *key*. Depending on its type, this *map* can contain
1544 * references to net devices (for forwarding packets through other
1545 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1546 * but this is only implemented for native XDP (with driver
1547 * support) as of this writing).
1549 * All values for *flags* are reserved for future usage, and must
1552 * When used to redirect packets to net devices, this helper
1553 * provides a high performance increase over **bpf_redirect**\ ().
1554 * This is due to various implementation details of the underlying
1555 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1556 * () tries to send packet as a "bulk" to the device.
1558 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1560 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1562 * Redirect the packet to the socket referenced by *map* (of type
1563 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1564 * egress interfaces can be used for redirection. The
1565 * **BPF_F_INGRESS** value in *flags* is used to make the
1566 * distinction (ingress path is selected if the flag is present,
1567 * egress path otherwise). This is the only flag supported for now.
1569 * **SK_PASS** on success, or **SK_DROP** on error.
1571 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1573 * Add an entry to, or update a *map* referencing sockets. The
1574 * *skops* is used as a new value for the entry associated to
1575 * *key*. *flags* is one of:
1578 * The entry for *key* must not exist in the map.
1580 * The entry for *key* must already exist in the map.
1582 * No condition on the existence of the entry for *key*.
1584 * If the *map* has eBPF programs (parser and verdict), those will
1585 * be inherited by the socket being added. If the socket is
1586 * already attached to eBPF programs, this results in an error.
1588 * 0 on success, or a negative error in case of failure.
1590 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1592 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1593 * *delta* (which can be positive or negative). Note that this
1594 * operation modifies the address stored in *xdp_md*\ **->data**,
1595 * so the latter must be loaded only after the helper has been
1598 * The use of *xdp_md*\ **->data_meta** is optional and programs
1599 * are not required to use it. The rationale is that when the
1600 * packet is processed with XDP (e.g. as DoS filter), it is
1601 * possible to push further meta data along with it before passing
1602 * to the stack, and to give the guarantee that an ingress eBPF
1603 * program attached as a TC classifier on the same device can pick
1604 * this up for further post-processing. Since TC works with socket
1605 * buffers, it remains possible to set from XDP the **mark** or
1606 * **priority** pointers, or other pointers for the socket buffer.
1607 * Having this scratch space generic and programmable allows for
1608 * more flexibility as the user is free to store whatever meta
1611 * A call to this helper is susceptible to change the underlaying
1612 * packet buffer. Therefore, at load time, all checks on pointers
1613 * previously done by the verifier are invalidated and must be
1614 * performed again, if the helper is used in combination with
1615 * direct packet access.
1617 * 0 on success, or a negative error in case of failure.
1619 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1621 * Read the value of a perf event counter, and store it into *buf*
1622 * of size *buf_size*. This helper relies on a *map* of type
1623 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1624 * counter is selected when *map* is updated with perf event file
1625 * descriptors. The *map* is an array whose size is the number of
1626 * available CPUs, and each cell contains a value relative to one
1627 * CPU. The value to retrieve is indicated by *flags*, that
1628 * contains the index of the CPU to look up, masked with
1629 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1630 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1631 * current CPU should be retrieved.
1633 * This helper behaves in a way close to
1634 * **bpf_perf_event_read**\ () helper, save that instead of
1635 * just returning the value observed, it fills the *buf*
1636 * structure. This allows for additional data to be retrieved: in
1637 * particular, the enabled and running times (in *buf*\
1638 * **->enabled** and *buf*\ **->running**, respectively) are
1639 * copied. In general, **bpf_perf_event_read_value**\ () is
1640 * recommended over **bpf_perf_event_read**\ (), which has some
1641 * ABI issues and provides fewer functionalities.
1643 * These values are interesting, because hardware PMU (Performance
1644 * Monitoring Unit) counters are limited resources. When there are
1645 * more PMU based perf events opened than available counters,
1646 * kernel will multiplex these events so each event gets certain
1647 * percentage (but not all) of the PMU time. In case that
1648 * multiplexing happens, the number of samples or counter value
1649 * will not reflect the case compared to when no multiplexing
1650 * occurs. This makes comparison between different runs difficult.
1651 * Typically, the counter value should be normalized before
1652 * comparing to other experiments. The usual normalization is done
1657 * normalized_counter = counter * t_enabled / t_running
1659 * Where t_enabled is the time enabled for event and t_running is
1660 * the time running for event since last normalization. The
1661 * enabled and running times are accumulated since the perf event
1662 * open. To achieve scaling factor between two invocations of an
1663 * eBPF program, users can can use CPU id as the key (which is
1664 * typical for perf array usage model) to remember the previous
1665 * value and do the calculation inside the eBPF program.
1667 * 0 on success, or a negative error in case of failure.
1669 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1671 * For en eBPF program attached to a perf event, retrieve the
1672 * value of the event counter associated to *ctx* and store it in
1673 * the structure pointed by *buf* and of size *buf_size*. Enabled
1674 * and running times are also stored in the structure (see
1675 * description of helper **bpf_perf_event_read_value**\ () for
1678 * 0 on success, or a negative error in case of failure.
1680 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1682 * Emulate a call to **getsockopt()** on the socket associated to
1683 * *bpf_socket*, which must be a full socket. The *level* at
1684 * which the option resides and the name *optname* of the option
1685 * must be specified, see **getsockopt(2)** for more information.
1686 * The retrieved value is stored in the structure pointed by
1687 * *opval* and of length *optlen*.
1689 * This helper actually implements a subset of **getsockopt()**.
1690 * It supports the following *level*\ s:
1692 * * **IPPROTO_TCP**, which supports *optname*
1693 * **TCP_CONGESTION**.
1694 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1695 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1697 * 0 on success, or a negative error in case of failure.
1699 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1701 * Used for error injection, this helper uses kprobes to override
1702 * the return value of the probed function, and to set it to *rc*.
1703 * The first argument is the context *regs* on which the kprobe
1706 * This helper works by setting setting the PC (program counter)
1707 * to an override function which is run in place of the original
1708 * probed function. This means the probed function is not run at
1709 * all. The replacement function just returns with the required
1712 * This helper has security implications, and thus is subject to
1713 * restrictions. It is only available if the kernel was compiled
1714 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1715 * option, and in this case it only works on functions tagged with
1716 * **ALLOW_ERROR_INJECTION** in the kernel code.
1718 * Also, the helper is only available for the architectures having
1719 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1720 * x86 architecture is the only one to support this feature.
1724 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1726 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1727 * for the full TCP socket associated to *bpf_sock_ops* to
1730 * The primary use of this field is to determine if there should
1731 * be calls to eBPF programs of type
1732 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1733 * code. A program of the same type can change its value, per
1734 * connection and as necessary, when the connection is
1735 * established. This field is directly accessible for reading, but
1736 * this helper must be used for updates in order to return an
1737 * error if an eBPF program tries to set a callback that is not
1738 * supported in the current kernel.
1740 * The supported callback values that *argval* can combine are:
1742 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1743 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1744 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1746 * Here are some examples of where one could call such eBPF
1750 * * When a packet is retransmitted.
1751 * * When the connection terminates.
1752 * * When a packet is sent.
1753 * * When a packet is received.
1755 * Code **-EINVAL** if the socket is not a full TCP socket;
1756 * otherwise, a positive number containing the bits that could not
1757 * be set is returned (which comes down to 0 if all bits were set
1760 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1762 * This helper is used in programs implementing policies at the
1763 * socket level. If the message *msg* is allowed to pass (i.e. if
1764 * the verdict eBPF program returns **SK_PASS**), redirect it to
1765 * the socket referenced by *map* (of type
1766 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1767 * egress interfaces can be used for redirection. The
1768 * **BPF_F_INGRESS** value in *flags* is used to make the
1769 * distinction (ingress path is selected if the flag is present,
1770 * egress path otherwise). This is the only flag supported for now.
1772 * **SK_PASS** on success, or **SK_DROP** on error.
1774 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1776 * For socket policies, apply the verdict of the eBPF program to
1777 * the next *bytes* (number of bytes) of message *msg*.
1779 * For example, this helper can be used in the following cases:
1781 * * A single **sendmsg**\ () or **sendfile**\ () system call
1782 * contains multiple logical messages that the eBPF program is
1783 * supposed to read and for which it should apply a verdict.
1784 * * An eBPF program only cares to read the first *bytes* of a
1785 * *msg*. If the message has a large payload, then setting up
1786 * and calling the eBPF program repeatedly for all bytes, even
1787 * though the verdict is already known, would create unnecessary
1790 * When called from within an eBPF program, the helper sets a
1791 * counter internal to the BPF infrastructure, that is used to
1792 * apply the last verdict to the next *bytes*. If *bytes* is
1793 * smaller than the current data being processed from a
1794 * **sendmsg**\ () or **sendfile**\ () system call, the first
1795 * *bytes* will be sent and the eBPF program will be re-run with
1796 * the pointer for start of data pointing to byte number *bytes*
1797 * **+ 1**. If *bytes* is larger than the current data being
1798 * processed, then the eBPF verdict will be applied to multiple
1799 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1802 * Note that if a socket closes with the internal counter holding
1803 * a non-zero value, this is not a problem because data is not
1804 * being buffered for *bytes* and is sent as it is received.
1808 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1810 * For socket policies, prevent the execution of the verdict eBPF
1811 * program for message *msg* until *bytes* (byte number) have been
1814 * This can be used when one needs a specific number of bytes
1815 * before a verdict can be assigned, even if the data spans
1816 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1817 * case would be a user calling **sendmsg**\ () repeatedly with
1818 * 1-byte long message segments. Obviously, this is bad for
1819 * performance, but it is still valid. If the eBPF program needs
1820 * *bytes* bytes to validate a header, this helper can be used to
1821 * prevent the eBPF program to be called again until *bytes* have
1826 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1828 * For socket policies, pull in non-linear data from user space
1829 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1830 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1833 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1834 * *msg* it can only parse data that the (**data**, **data_end**)
1835 * pointers have already consumed. For **sendmsg**\ () hooks this
1836 * is likely the first scatterlist element. But for calls relying
1837 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1838 * be the range (**0**, **0**) because the data is shared with
1839 * user space and by default the objective is to avoid allowing
1840 * user space to modify data while (or after) eBPF verdict is
1841 * being decided. This helper can be used to pull in data and to
1842 * set the start and end pointer to given values. Data will be
1843 * copied if necessary (i.e. if data was not linear and if start
1844 * and end pointers do not point to the same chunk).
1846 * A call to this helper is susceptible to change the underlaying
1847 * packet buffer. Therefore, at load time, all checks on pointers
1848 * previously done by the verifier are invalidated and must be
1849 * performed again, if the helper is used in combination with
1850 * direct packet access.
1852 * All values for *flags* are reserved for future usage, and must
1855 * 0 on success, or a negative error in case of failure.
1857 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1859 * Bind the socket associated to *ctx* to the address pointed by
1860 * *addr*, of length *addr_len*. This allows for making outgoing
1861 * connection from the desired IP address, which can be useful for
1862 * example when all processes inside a cgroup should use one
1863 * single IP address on a host that has multiple IP configured.
1865 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1866 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1867 * **AF_INET6**). Looking for a free port to bind to can be
1868 * expensive, therefore binding to port is not permitted by the
1869 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1870 * must be set to zero.
1872 * 0 on success, or a negative error in case of failure.
1874 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1876 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1877 * only possible to shrink the packet as of this writing,
1878 * therefore *delta* must be a negative integer.
1880 * A call to this helper is susceptible to change the underlaying
1881 * packet buffer. Therefore, at load time, all checks on pointers
1882 * previously done by the verifier are invalidated and must be
1883 * performed again, if the helper is used in combination with
1884 * direct packet access.
1886 * 0 on success, or a negative error in case of failure.
1888 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1890 * Retrieve the XFRM state (IP transform framework, see also
1891 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1893 * The retrieved value is stored in the **struct bpf_xfrm_state**
1894 * pointed by *xfrm_state* and of length *size*.
1896 * All values for *flags* are reserved for future usage, and must
1899 * This helper is available only if the kernel was compiled with
1900 * **CONFIG_XFRM** configuration option.
1902 * 0 on success, or a negative error in case of failure.
1904 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1906 * Return a user or a kernel stack in bpf program provided buffer.
1907 * To achieve this, the helper needs *ctx*, which is a pointer
1908 * to the context on which the tracing program is executed.
1909 * To store the stacktrace, the bpf program provides *buf* with
1910 * a nonnegative *size*.
1912 * The last argument, *flags*, holds the number of stack frames to
1913 * skip (from 0 to 255), masked with
1914 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1915 * the following flags:
1917 * **BPF_F_USER_STACK**
1918 * Collect a user space stack instead of a kernel stack.
1919 * **BPF_F_USER_BUILD_ID**
1920 * Collect buildid+offset instead of ips for user stack,
1921 * only valid if **BPF_F_USER_STACK** is also specified.
1923 * **bpf_get_stack**\ () can collect up to
1924 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1925 * to sufficient large buffer size. Note that
1926 * this limit can be controlled with the **sysctl** program, and
1927 * that it should be manually increased in order to profile long
1928 * user stacks (such as stacks for Java programs). To do so, use:
1932 * # sysctl kernel.perf_event_max_stack=<new value>
1934 * A non-negative value equal to or less than *size* on success,
1935 * or a negative error in case of failure.
1937 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1939 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1940 * it provides an easy way to load *len* bytes from *offset*
1941 * from the packet associated to *skb*, into the buffer pointed
1942 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1943 * a fifth argument *start_header* exists in order to select a
1944 * base offset to start from. *start_header* can be one of:
1946 * **BPF_HDR_START_MAC**
1947 * Base offset to load data from is *skb*'s mac header.
1948 * **BPF_HDR_START_NET**
1949 * Base offset to load data from is *skb*'s network header.
1951 * In general, "direct packet access" is the preferred method to
1952 * access packet data, however, this helper is in particular useful
1953 * in socket filters where *skb*\ **->data** does not always point
1954 * to the start of the mac header and where "direct packet access"
1957 * 0 on success, or a negative error in case of failure.
1959 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1961 * Do FIB lookup in kernel tables using parameters in *params*.
1962 * If lookup is successful and result shows packet is to be
1963 * forwarded, the neighbor tables are searched for the nexthop.
1964 * If successful (ie., FIB lookup shows forwarding and nexthop
1965 * is resolved), the nexthop address is returned in ipv4_dst
1966 * or ipv6_dst based on family, smac is set to mac address of
1967 * egress device, dmac is set to nexthop mac address, rt_metric
1968 * is set to metric from route (IPv4/IPv6 only), and ifindex
1969 * is set to the device index of the nexthop from the FIB lookup.
1971 * *plen* argument is the size of the passed in struct.
1972 * *flags* argument can be a combination of one or more of the
1975 * **BPF_FIB_LOOKUP_DIRECT**
1976 * Do a direct table lookup vs full lookup using FIB
1978 * **BPF_FIB_LOOKUP_OUTPUT**
1979 * Perform lookup from an egress perspective (default is
1982 * *ctx* is either **struct xdp_md** for XDP programs or
1983 * **struct sk_buff** tc cls_act programs.
1985 * * < 0 if any input argument is invalid
1986 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1987 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1988 * packet is not forwarded or needs assist from full stack
1990 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1992 * Add an entry to, or update a sockhash *map* referencing sockets.
1993 * The *skops* is used as a new value for the entry associated to
1994 * *key*. *flags* is one of:
1997 * The entry for *key* must not exist in the map.
1999 * The entry for *key* must already exist in the map.
2001 * No condition on the existence of the entry for *key*.
2003 * If the *map* has eBPF programs (parser and verdict), those will
2004 * be inherited by the socket being added. If the socket is
2005 * already attached to eBPF programs, this results in an error.
2007 * 0 on success, or a negative error in case of failure.
2009 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2011 * This helper is used in programs implementing policies at the
2012 * socket level. If the message *msg* is allowed to pass (i.e. if
2013 * the verdict eBPF program returns **SK_PASS**), redirect it to
2014 * the socket referenced by *map* (of type
2015 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2016 * egress interfaces can be used for redirection. The
2017 * **BPF_F_INGRESS** value in *flags* is used to make the
2018 * distinction (ingress path is selected if the flag is present,
2019 * egress path otherwise). This is the only flag supported for now.
2021 * **SK_PASS** on success, or **SK_DROP** on error.
2023 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2025 * This helper is used in programs implementing policies at the
2026 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2027 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2028 * to the socket referenced by *map* (of type
2029 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2030 * egress interfaces can be used for redirection. The
2031 * **BPF_F_INGRESS** value in *flags* is used to make the
2032 * distinction (ingress path is selected if the flag is present,
2033 * egress otherwise). This is the only flag supported for now.
2035 * **SK_PASS** on success, or **SK_DROP** on error.
2037 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2039 * Encapsulate the packet associated to *skb* within a Layer 3
2040 * protocol header. This header is provided in the buffer at
2041 * address *hdr*, with *len* its size in bytes. *type* indicates
2042 * the protocol of the header and can be one of:
2044 * **BPF_LWT_ENCAP_SEG6**
2045 * IPv6 encapsulation with Segment Routing Header
2046 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2047 * the IPv6 header is computed by the kernel.
2048 * **BPF_LWT_ENCAP_SEG6_INLINE**
2049 * Only works if *skb* contains an IPv6 packet. Insert a
2050 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2052 * **BPF_LWT_ENCAP_IP**
2053 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2054 * must be IPv4 or IPv6, followed by zero or more
2055 * additional headers, up to LWT_BPF_MAX_HEADROOM total
2056 * bytes in all prepended headers. Please note that
2057 * if skb_is_gso(skb) is true, no more than two headers
2058 * can be prepended, and the inner header, if present,
2059 * should be either GRE or UDP/GUE.
2061 * BPF_LWT_ENCAP_SEG6*** types can be called by bpf programs of
2062 * type BPF_PROG_TYPE_LWT_IN; BPF_LWT_ENCAP_IP type can be called
2063 * by bpf programs of types BPF_PROG_TYPE_LWT_IN and
2064 * BPF_PROG_TYPE_LWT_XMIT.
2066 * A call to this helper is susceptible to change the underlaying
2067 * packet buffer. Therefore, at load time, all checks on pointers
2068 * previously done by the verifier are invalidated and must be
2069 * performed again, if the helper is used in combination with
2070 * direct packet access.
2072 * 0 on success, or a negative error in case of failure.
2074 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2076 * Store *len* bytes from address *from* into the packet
2077 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2078 * inside the outermost IPv6 Segment Routing Header can be
2079 * modified through this helper.
2081 * A call to this helper is susceptible to change the underlaying
2082 * packet buffer. Therefore, at load time, all checks on pointers
2083 * previously done by the verifier are invalidated and must be
2084 * performed again, if the helper is used in combination with
2085 * direct packet access.
2087 * 0 on success, or a negative error in case of failure.
2089 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2091 * Adjust the size allocated to TLVs in the outermost IPv6
2092 * Segment Routing Header contained in the packet associated to
2093 * *skb*, at position *offset* by *delta* bytes. Only offsets
2094 * after the segments are accepted. *delta* can be as well
2095 * positive (growing) as negative (shrinking).
2097 * A call to this helper is susceptible to change the underlaying
2098 * packet buffer. Therefore, at load time, all checks on pointers
2099 * previously done by the verifier are invalidated and must be
2100 * performed again, if the helper is used in combination with
2101 * direct packet access.
2103 * 0 on success, or a negative error in case of failure.
2105 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2107 * Apply an IPv6 Segment Routing action of type *action* to the
2108 * packet associated to *skb*. Each action takes a parameter
2109 * contained at address *param*, and of length *param_len* bytes.
2110 * *action* can be one of:
2112 * **SEG6_LOCAL_ACTION_END_X**
2113 * End.X action: Endpoint with Layer-3 cross-connect.
2114 * Type of *param*: **struct in6_addr**.
2115 * **SEG6_LOCAL_ACTION_END_T**
2116 * End.T action: Endpoint with specific IPv6 table lookup.
2117 * Type of *param*: **int**.
2118 * **SEG6_LOCAL_ACTION_END_B6**
2119 * End.B6 action: Endpoint bound to an SRv6 policy.
2120 * Type of param: **struct ipv6_sr_hdr**.
2121 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2122 * End.B6.Encap action: Endpoint bound to an SRv6
2123 * encapsulation policy.
2124 * Type of param: **struct ipv6_sr_hdr**.
2126 * A call to this helper is susceptible to change the underlaying
2127 * packet buffer. Therefore, at load time, all checks on pointers
2128 * previously done by the verifier are invalidated and must be
2129 * performed again, if the helper is used in combination with
2130 * direct packet access.
2132 * 0 on success, or a negative error in case of failure.
2134 * int bpf_rc_repeat(void *ctx)
2136 * This helper is used in programs implementing IR decoding, to
2137 * report a successfully decoded repeat key message. This delays
2138 * the generation of a key up event for previously generated
2141 * Some IR protocols like NEC have a special IR message for
2142 * repeating last button, for when a button is held down.
2144 * The *ctx* should point to the lirc sample as passed into
2147 * This helper is only available is the kernel was compiled with
2148 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2153 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2155 * This helper is used in programs implementing IR decoding, to
2156 * report a successfully decoded key press with *scancode*,
2157 * *toggle* value in the given *protocol*. The scancode will be
2158 * translated to a keycode using the rc keymap, and reported as
2159 * an input key down event. After a period a key up event is
2160 * generated. This period can be extended by calling either
2161 * **bpf_rc_keydown**\ () again with the same values, or calling
2162 * **bpf_rc_repeat**\ ().
2164 * Some protocols include a toggle bit, in case the button was
2165 * released and pressed again between consecutive scancodes.
2167 * The *ctx* should point to the lirc sample as passed into
2170 * The *protocol* is the decoded protocol number (see
2171 * **enum rc_proto** for some predefined values).
2173 * This helper is only available is the kernel was compiled with
2174 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2179 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2181 * Return the cgroup v2 id of the socket associated with the *skb*.
2182 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2183 * helper for cgroup v1 by providing a tag resp. identifier that
2184 * can be matched on or used for map lookups e.g. to implement
2185 * policy. The cgroup v2 id of a given path in the hierarchy is
2186 * exposed in user space through the f_handle API in order to get
2187 * to the same 64-bit id.
2189 * This helper can be used on TC egress path, but not on ingress,
2190 * and is available only if the kernel was compiled with the
2191 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2193 * The id is returned or 0 in case the id could not be retrieved.
2195 * u64 bpf_get_current_cgroup_id(void)
2197 * A 64-bit integer containing the current cgroup id based
2198 * on the cgroup within which the current task is running.
2200 * void *bpf_get_local_storage(void *map, u64 flags)
2202 * Get the pointer to the local storage area.
2203 * The type and the size of the local storage is defined
2204 * by the *map* argument.
2205 * The *flags* meaning is specific for each map type,
2206 * and has to be 0 for cgroup local storage.
2208 * Depending on the BPF program type, a local storage area
2209 * can be shared between multiple instances of the BPF program,
2210 * running simultaneously.
2212 * A user should care about the synchronization by himself.
2213 * For example, by using the **BPF_STX_XADD** instruction to alter
2216 * A pointer to the local storage area.
2218 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2220 * Select a **SO_REUSEPORT** socket from a
2221 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2222 * It checks the selected socket is matching the incoming
2223 * request in the socket buffer.
2225 * 0 on success, or a negative error in case of failure.
2227 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2229 * Return id of cgroup v2 that is ancestor of cgroup associated
2230 * with the *skb* at the *ancestor_level*. The root cgroup is at
2231 * *ancestor_level* zero and each step down the hierarchy
2232 * increments the level. If *ancestor_level* == level of cgroup
2233 * associated with *skb*, then return value will be same as that
2234 * of **bpf_skb_cgroup_id**\ ().
2236 * The helper is useful to implement policies based on cgroups
2237 * that are upper in hierarchy than immediate cgroup associated
2240 * The format of returned id and helper limitations are same as in
2241 * **bpf_skb_cgroup_id**\ ().
2243 * The id is returned or 0 in case the id could not be retrieved.
2245 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2247 * Look for TCP socket matching *tuple*, optionally in a child
2248 * network namespace *netns*. The return value must be checked,
2249 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2251 * The *ctx* should point to the context of the program, such as
2252 * the skb or socket (depending on the hook in use). This is used
2253 * to determine the base network namespace for the lookup.
2255 * *tuple_size* must be one of:
2257 * **sizeof**\ (*tuple*\ **->ipv4**)
2258 * Look for an IPv4 socket.
2259 * **sizeof**\ (*tuple*\ **->ipv6**)
2260 * Look for an IPv6 socket.
2262 * If the *netns* is a negative signed 32-bit integer, then the
2263 * socket lookup table in the netns associated with the *ctx* will
2264 * will be used. For the TC hooks, this is the netns of the device
2265 * in the skb. For socket hooks, this is the netns of the socket.
2266 * If *netns* is any other signed 32-bit value greater than or
2267 * equal to zero then it specifies the ID of the netns relative to
2268 * the netns associated with the *ctx*. *netns* values beyond the
2269 * range of 32-bit integers are reserved for future use.
2271 * All values for *flags* are reserved for future usage, and must
2274 * This helper is available only if the kernel was compiled with
2275 * **CONFIG_NET** configuration option.
2277 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2278 * For sockets with reuseport option, the **struct bpf_sock**
2279 * result is from **reuse->socks**\ [] using the hash of the tuple.
2281 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2283 * Look for UDP socket matching *tuple*, optionally in a child
2284 * network namespace *netns*. The return value must be checked,
2285 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2287 * The *ctx* should point to the context of the program, such as
2288 * the skb or socket (depending on the hook in use). This is used
2289 * to determine the base network namespace for the lookup.
2291 * *tuple_size* must be one of:
2293 * **sizeof**\ (*tuple*\ **->ipv4**)
2294 * Look for an IPv4 socket.
2295 * **sizeof**\ (*tuple*\ **->ipv6**)
2296 * Look for an IPv6 socket.
2298 * If the *netns* is a negative signed 32-bit integer, then the
2299 * socket lookup table in the netns associated with the *ctx* will
2300 * will be used. For the TC hooks, this is the netns of the device
2301 * in the skb. For socket hooks, this is the netns of the socket.
2302 * If *netns* is any other signed 32-bit value greater than or
2303 * equal to zero then it specifies the ID of the netns relative to
2304 * the netns associated with the *ctx*. *netns* values beyond the
2305 * range of 32-bit integers are reserved for future use.
2307 * All values for *flags* are reserved for future usage, and must
2310 * This helper is available only if the kernel was compiled with
2311 * **CONFIG_NET** configuration option.
2313 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2314 * For sockets with reuseport option, the **struct bpf_sock**
2315 * result is from **reuse->socks**\ [] using the hash of the tuple.
2317 * int bpf_sk_release(struct bpf_sock *sock)
2319 * Release the reference held by *sock*. *sock* must be a
2320 * non-**NULL** pointer that was returned from
2321 * **bpf_sk_lookup_xxx**\ ().
2323 * 0 on success, or a negative error in case of failure.
2325 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2327 * Push an element *value* in *map*. *flags* is one of:
2330 * If the queue/stack is full, the oldest element is
2331 * removed to make room for this.
2333 * 0 on success, or a negative error in case of failure.
2335 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2337 * Pop an element from *map*.
2339 * 0 on success, or a negative error in case of failure.
2341 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2343 * Get an element from *map* without removing it.
2345 * 0 on success, or a negative error in case of failure.
2347 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2349 * For socket policies, insert *len* bytes into *msg* at offset
2352 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2353 * *msg* it may want to insert metadata or options into the *msg*.
2354 * This can later be read and used by any of the lower layer BPF
2357 * This helper may fail if under memory pressure (a malloc
2358 * fails) in these cases BPF programs will get an appropriate
2359 * error and BPF programs will need to handle them.
2361 * 0 on success, or a negative error in case of failure.
2363 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2365 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2366 * This may result in **ENOMEM** errors under certain situations if
2367 * an allocation and copy are required due to a full ring buffer.
2368 * However, the helper will try to avoid doing the allocation
2369 * if possible. Other errors can occur if input parameters are
2370 * invalid either due to *start* byte not being valid part of *msg*
2371 * payload and/or *pop* value being to large.
2373 * 0 on success, or a negative error in case of failure.
2375 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2377 * This helper is used in programs implementing IR decoding, to
2378 * report a successfully decoded pointer movement.
2380 * The *ctx* should point to the lirc sample as passed into
2383 * This helper is only available is the kernel was compiled with
2384 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2389 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2391 * Acquire a spinlock represented by the pointer *lock*, which is
2392 * stored as part of a value of a map. Taking the lock allows to
2393 * safely update the rest of the fields in that value. The
2394 * spinlock can (and must) later be released with a call to
2395 * **bpf_spin_unlock**\ (\ *lock*\ ).
2397 * Spinlocks in BPF programs come with a number of restrictions
2400 * * **bpf_spin_lock** objects are only allowed inside maps of
2401 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2402 * list could be extended in the future).
2403 * * BTF description of the map is mandatory.
2404 * * The BPF program can take ONE lock at a time, since taking two
2405 * or more could cause dead locks.
2406 * * Only one **struct bpf_spin_lock** is allowed per map element.
2407 * * When the lock is taken, calls (either BPF to BPF or helpers)
2409 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2410 * allowed inside a spinlock-ed region.
2411 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2412 * the lock, on all execution paths, before it returns.
2413 * * The BPF program can access **struct bpf_spin_lock** only via
2414 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2415 * helpers. Loading or storing data into the **struct
2416 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2417 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2418 * of the map value must be a struct and have **struct
2419 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2420 * Nested lock inside another struct is not allowed.
2421 * * The **struct bpf_spin_lock** *lock* field in a map value must
2422 * be aligned on a multiple of 4 bytes in that value.
2423 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2424 * the **bpf_spin_lock** field to user space.
2425 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2426 * a BPF program, do not update the **bpf_spin_lock** field.
2427 * * **bpf_spin_lock** cannot be on the stack or inside a
2428 * networking packet (it can only be inside of a map values).
2429 * * **bpf_spin_lock** is available to root only.
2430 * * Tracing programs and socket filter programs cannot use
2431 * **bpf_spin_lock**\ () due to insufficient preemption checks
2432 * (but this may change in the future).
2433 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2437 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2439 * Release the *lock* previously locked by a call to
2440 * **bpf_spin_lock**\ (\ *lock*\ ).
2444 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2446 * This helper gets a **struct bpf_sock** pointer such
2447 * that all the fields in this **bpf_sock** can be accessed.
2449 * A **struct bpf_sock** pointer on success, or **NULL** in
2452 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2454 * This helper gets a **struct bpf_tcp_sock** pointer from a
2455 * **struct bpf_sock** pointer.
2457 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2460 * int bpf_skb_ecn_set_ce(struct sk_buf *skb)
2462 * Set ECN (Explicit Congestion Notification) field of IP header
2463 * to **CE** (Congestion Encountered) if current value is **ECT**
2464 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2467 * 1 if the **CE** flag is set (either by the current helper call
2468 * or because it was already present), 0 if it is not set.
2470 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2472 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2473 * **bpf_sk_release**\ () is unnecessary and not allowed.
2475 * A **struct bpf_sock** pointer on success, or **NULL** in
2478 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2480 * Look for TCP socket matching *tuple*, optionally in a child
2481 * network namespace *netns*. The return value must be checked,
2482 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2484 * This function is identical to bpf_sk_lookup_tcp, except that it
2485 * also returns timewait or request sockets. Use bpf_sk_fullsock
2486 * or bpf_tcp_socket to access the full structure.
2488 * This helper is available only if the kernel was compiled with
2489 * **CONFIG_NET** configuration option.
2491 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2492 * For sockets with reuseport option, the **struct bpf_sock**
2493 * result is from **reuse->socks**\ [] using the hash of the tuple.
2495 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2497 * Check whether iph and th contain a valid SYN cookie ACK for
2498 * the listening socket in sk.
2500 * iph points to the start of the IPv4 or IPv6 header, while
2501 * iph_len contains sizeof(struct iphdr) or sizeof(struct ip6hdr).
2503 * th points to the start of the TCP header, while th_len contains
2504 * sizeof(struct tcphdr).
2507 * 0 if iph and th are a valid SYN cookie ACK, or a negative error
2510 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2512 * Get name of sysctl in /proc/sys/ and copy it into provided by
2513 * program buffer *buf* of size *buf_len*.
2515 * The buffer is always NUL terminated, unless it's zero-sized.
2517 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2518 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2519 * only (e.g. "tcp_mem").
2521 * Number of character copied (not including the trailing NUL).
2523 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2524 * truncated name in this case).
2526 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2528 * Get current value of sysctl as it is presented in /proc/sys
2529 * (incl. newline, etc), and copy it as a string into provided
2530 * by program buffer *buf* of size *buf_len*.
2532 * The whole value is copied, no matter what file position user
2533 * space issued e.g. sys_read at.
2535 * The buffer is always NUL terminated, unless it's zero-sized.
2537 * Number of character copied (not including the trailing NUL).
2539 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2540 * truncated name in this case).
2542 * **-EINVAL** if current value was unavailable, e.g. because
2543 * sysctl is uninitialized and read returns -EIO for it.
2545 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2547 * Get new value being written by user space to sysctl (before
2548 * the actual write happens) and copy it as a string into
2549 * provided by program buffer *buf* of size *buf_len*.
2551 * User space may write new value at file position > 0.
2553 * The buffer is always NUL terminated, unless it's zero-sized.
2555 * Number of character copied (not including the trailing NUL).
2557 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2558 * truncated name in this case).
2560 * **-EINVAL** if sysctl is being read.
2562 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2564 * Override new value being written by user space to sysctl with
2565 * value provided by program in buffer *buf* of size *buf_len*.
2567 * *buf* should contain a string in same form as provided by user
2568 * space on sysctl write.
2570 * User space may write new value at file position > 0. To override
2571 * the whole sysctl value file position should be set to zero.
2575 * **-E2BIG** if the *buf_len* is too big.
2577 * **-EINVAL** if sysctl is being read.
2579 #define __BPF_FUNC_MAPPER(FN) \
2581 FN(map_lookup_elem), \
2582 FN(map_update_elem), \
2583 FN(map_delete_elem), \
2587 FN(get_prandom_u32), \
2588 FN(get_smp_processor_id), \
2589 FN(skb_store_bytes), \
2590 FN(l3_csum_replace), \
2591 FN(l4_csum_replace), \
2593 FN(clone_redirect), \
2594 FN(get_current_pid_tgid), \
2595 FN(get_current_uid_gid), \
2596 FN(get_current_comm), \
2597 FN(get_cgroup_classid), \
2598 FN(skb_vlan_push), \
2600 FN(skb_get_tunnel_key), \
2601 FN(skb_set_tunnel_key), \
2602 FN(perf_event_read), \
2604 FN(get_route_realm), \
2605 FN(perf_event_output), \
2606 FN(skb_load_bytes), \
2609 FN(skb_get_tunnel_opt), \
2610 FN(skb_set_tunnel_opt), \
2611 FN(skb_change_proto), \
2612 FN(skb_change_type), \
2613 FN(skb_under_cgroup), \
2614 FN(get_hash_recalc), \
2615 FN(get_current_task), \
2616 FN(probe_write_user), \
2617 FN(current_task_under_cgroup), \
2618 FN(skb_change_tail), \
2619 FN(skb_pull_data), \
2621 FN(set_hash_invalid), \
2622 FN(get_numa_node_id), \
2623 FN(skb_change_head), \
2624 FN(xdp_adjust_head), \
2625 FN(probe_read_str), \
2626 FN(get_socket_cookie), \
2627 FN(get_socket_uid), \
2630 FN(skb_adjust_room), \
2632 FN(sk_redirect_map), \
2633 FN(sock_map_update), \
2634 FN(xdp_adjust_meta), \
2635 FN(perf_event_read_value), \
2636 FN(perf_prog_read_value), \
2638 FN(override_return), \
2639 FN(sock_ops_cb_flags_set), \
2640 FN(msg_redirect_map), \
2641 FN(msg_apply_bytes), \
2642 FN(msg_cork_bytes), \
2643 FN(msg_pull_data), \
2645 FN(xdp_adjust_tail), \
2646 FN(skb_get_xfrm_state), \
2648 FN(skb_load_bytes_relative), \
2650 FN(sock_hash_update), \
2651 FN(msg_redirect_hash), \
2652 FN(sk_redirect_hash), \
2653 FN(lwt_push_encap), \
2654 FN(lwt_seg6_store_bytes), \
2655 FN(lwt_seg6_adjust_srh), \
2656 FN(lwt_seg6_action), \
2659 FN(skb_cgroup_id), \
2660 FN(get_current_cgroup_id), \
2661 FN(get_local_storage), \
2662 FN(sk_select_reuseport), \
2663 FN(skb_ancestor_cgroup_id), \
2664 FN(sk_lookup_tcp), \
2665 FN(sk_lookup_udp), \
2667 FN(map_push_elem), \
2669 FN(map_peek_elem), \
2670 FN(msg_push_data), \
2672 FN(rc_pointer_rel), \
2677 FN(skb_ecn_set_ce), \
2678 FN(get_listener_sock), \
2679 FN(skc_lookup_tcp), \
2680 FN(tcp_check_syncookie), \
2681 FN(sysctl_get_name), \
2682 FN(sysctl_get_current_value), \
2683 FN(sysctl_get_new_value), \
2684 FN(sysctl_set_new_value),
2686 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2687 * function eBPF program intends to call
2689 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2691 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2694 #undef __BPF_ENUM_FN
2696 /* All flags used by eBPF helper functions, placed here. */
2698 /* BPF_FUNC_skb_store_bytes flags. */
2699 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2700 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2702 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2703 * First 4 bits are for passing the header field size.
2705 #define BPF_F_HDR_FIELD_MASK 0xfULL
2707 /* BPF_FUNC_l4_csum_replace flags. */
2708 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2709 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2710 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2712 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2713 #define BPF_F_INGRESS (1ULL << 0)
2715 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2716 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2718 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2719 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2720 #define BPF_F_USER_STACK (1ULL << 8)
2721 /* flags used by BPF_FUNC_get_stackid only. */
2722 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2723 #define BPF_F_REUSE_STACKID (1ULL << 10)
2724 /* flags used by BPF_FUNC_get_stack only. */
2725 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2727 /* BPF_FUNC_skb_set_tunnel_key flags. */
2728 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2729 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2730 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2732 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2733 * BPF_FUNC_perf_event_read_value flags.
2735 #define BPF_F_INDEX_MASK 0xffffffffULL
2736 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2737 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2738 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2740 /* Current network namespace */
2741 #define BPF_F_CURRENT_NETNS (-1L)
2743 /* BPF_FUNC_skb_adjust_room flags. */
2744 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
2746 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
2747 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
2749 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
2750 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
2751 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
2752 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
2753 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
2754 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
2755 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
2757 /* BPF_FUNC_sysctl_get_name flags. */
2758 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
2760 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2761 enum bpf_adj_room_mode {
2766 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2767 enum bpf_hdr_start_off {
2772 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2773 enum bpf_lwt_encap_mode {
2775 BPF_LWT_ENCAP_SEG6_INLINE,
2779 #define __bpf_md_ptr(type, name) \
2783 } __attribute__((aligned(8)))
2785 /* user accessible mirror of in-kernel sk_buff.
2786 * new fields can only be added to the end of this structure
2792 __u32 queue_mapping;
2798 __u32 ingress_ifindex;
2808 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2810 __u32 remote_ip4; /* Stored in network byte order */
2811 __u32 local_ip4; /* Stored in network byte order */
2812 __u32 remote_ip6[4]; /* Stored in network byte order */
2813 __u32 local_ip6[4]; /* Stored in network byte order */
2814 __u32 remote_port; /* Stored in network byte order */
2815 __u32 local_port; /* stored in host byte order */
2819 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
2823 __bpf_md_ptr(struct bpf_sock *, sk);
2826 struct bpf_tunnel_key {
2830 __u32 remote_ipv6[4];
2834 __u16 tunnel_ext; /* Padding, future use. */
2838 /* user accessible mirror of in-kernel xfrm_state.
2839 * new fields can only be added to the end of this structure
2841 struct bpf_xfrm_state {
2843 __u32 spi; /* Stored in network byte order */
2845 __u16 ext; /* Padding, future use. */
2847 __u32 remote_ipv4; /* Stored in network byte order */
2848 __u32 remote_ipv6[4]; /* Stored in network byte order */
2852 /* Generic BPF return codes which all BPF program types may support.
2853 * The values are binary compatible with their TC_ACT_* counter-part to
2854 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2857 * XDP is handled seprately, see XDP_*.
2865 /* >127 are reserved for prog type specific return codes.
2867 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
2868 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
2869 * changed and should be routed based on its new L3 header.
2870 * (This is an L3 redirect, as opposed to L2 redirect
2871 * represented by BPF_REDIRECT above).
2873 BPF_LWT_REROUTE = 128,
2883 /* IP address also allows 1 and 2 bytes access */
2886 __u32 src_port; /* host byte order */
2887 __u32 dst_port; /* network byte order */
2893 struct bpf_tcp_sock {
2894 __u32 snd_cwnd; /* Sending congestion window */
2895 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
2897 __u32 snd_ssthresh; /* Slow start size threshold */
2898 __u32 rcv_nxt; /* What we want to receive next */
2899 __u32 snd_nxt; /* Next sequence we send */
2900 __u32 snd_una; /* First byte we want an ack for */
2901 __u32 mss_cache; /* Cached effective mss, not including SACKS */
2902 __u32 ecn_flags; /* ECN status bits. */
2903 __u32 rate_delivered; /* saved rate sample: packets delivered */
2904 __u32 rate_interval_us; /* saved rate sample: time elapsed */
2905 __u32 packets_out; /* Packets which are "in flight" */
2906 __u32 retrans_out; /* Retransmitted packets out */
2907 __u32 total_retrans; /* Total retransmits for entire connection */
2908 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
2909 * total number of segments in.
2911 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
2912 * total number of data segments in.
2914 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
2915 * The total number of segments sent.
2917 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
2918 * total number of data segments sent.
2920 __u32 lost_out; /* Lost packets */
2921 __u32 sacked_out; /* SACK'd packets */
2922 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
2923 * sum(delta(rcv_nxt)), or how many bytes
2926 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
2927 * sum(delta(snd_una)), or how many bytes
2932 struct bpf_sock_tuple {
2949 #define XDP_PACKET_HEADROOM 256
2951 /* User return codes for XDP prog type.
2952 * A valid XDP program must return one of these defined values. All other
2953 * return codes are reserved for future use. Unknown return codes will
2954 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2964 /* user accessible metadata for XDP packet hook
2965 * new fields must be added to the end of this structure
2971 /* Below access go through struct xdp_rxq_info */
2972 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2973 __u32 rx_queue_index; /* rxq->queue_index */
2981 /* user accessible metadata for SK_MSG packet hook, new fields must
2982 * be added to the end of this structure
2985 __bpf_md_ptr(void *, data);
2986 __bpf_md_ptr(void *, data_end);
2989 __u32 remote_ip4; /* Stored in network byte order */
2990 __u32 local_ip4; /* Stored in network byte order */
2991 __u32 remote_ip6[4]; /* Stored in network byte order */
2992 __u32 local_ip6[4]; /* Stored in network byte order */
2993 __u32 remote_port; /* Stored in network byte order */
2994 __u32 local_port; /* stored in host byte order */
2995 __u32 size; /* Total size of sk_msg */
2998 struct sk_reuseport_md {
3000 * Start of directly accessible data. It begins from
3001 * the tcp/udp header.
3003 __bpf_md_ptr(void *, data);
3004 /* End of directly accessible data */
3005 __bpf_md_ptr(void *, data_end);
3007 * Total length of packet (starting from the tcp/udp header).
3008 * Note that the directly accessible bytes (data_end - data)
3009 * could be less than this "len". Those bytes could be
3010 * indirectly read by a helper "bpf_skb_load_bytes()".
3014 * Eth protocol in the mac header (network byte order). e.g.
3015 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3018 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3019 __u32 bind_inany; /* Is sock bound to an INANY address? */
3020 __u32 hash; /* A hash of the packet 4 tuples */
3023 #define BPF_TAG_SIZE 8
3025 struct bpf_prog_info {
3028 __u8 tag[BPF_TAG_SIZE];
3029 __u32 jited_prog_len;
3030 __u32 xlated_prog_len;
3031 __aligned_u64 jited_prog_insns;
3032 __aligned_u64 xlated_prog_insns;
3033 __u64 load_time; /* ns since boottime */
3034 __u32 created_by_uid;
3036 __aligned_u64 map_ids;
3037 char name[BPF_OBJ_NAME_LEN];
3039 __u32 gpl_compatible:1;
3042 __u32 nr_jited_ksyms;
3043 __u32 nr_jited_func_lens;
3044 __aligned_u64 jited_ksyms;
3045 __aligned_u64 jited_func_lens;
3047 __u32 func_info_rec_size;
3048 __aligned_u64 func_info;
3051 __aligned_u64 line_info;
3052 __aligned_u64 jited_line_info;
3053 __u32 nr_jited_line_info;
3054 __u32 line_info_rec_size;
3055 __u32 jited_line_info_rec_size;
3057 __aligned_u64 prog_tags;
3060 } __attribute__((aligned(8)));
3062 struct bpf_map_info {
3069 char name[BPF_OBJ_NAME_LEN];
3075 __u32 btf_key_type_id;
3076 __u32 btf_value_type_id;
3077 } __attribute__((aligned(8)));
3079 struct bpf_btf_info {
3083 } __attribute__((aligned(8)));
3085 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3086 * by user and intended to be used by socket (e.g. to bind to, depends on
3087 * attach attach type).
3089 struct bpf_sock_addr {
3090 __u32 user_family; /* Allows 4-byte read, but no write. */
3091 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3092 * Stored in network byte order.
3094 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
3095 * Stored in network byte order.
3097 __u32 user_port; /* Allows 4-byte read and write.
3098 * Stored in network byte order
3100 __u32 family; /* Allows 4-byte read, but no write */
3101 __u32 type; /* Allows 4-byte read, but no write */
3102 __u32 protocol; /* Allows 4-byte read, but no write */
3103 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
3104 * Stored in network byte order.
3106 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
3107 * Stored in network byte order.
3111 /* User bpf_sock_ops struct to access socket values and specify request ops
3112 * and their replies.
3113 * Some of this fields are in network (bigendian) byte order and may need
3114 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3115 * New fields can only be added at the end of this structure
3117 struct bpf_sock_ops {
3120 __u32 args[4]; /* Optionally passed to bpf program */
3121 __u32 reply; /* Returned by bpf program */
3122 __u32 replylong[4]; /* Optionally returned by bpf prog */
3125 __u32 remote_ip4; /* Stored in network byte order */
3126 __u32 local_ip4; /* Stored in network byte order */
3127 __u32 remote_ip6[4]; /* Stored in network byte order */
3128 __u32 local_ip6[4]; /* Stored in network byte order */
3129 __u32 remote_port; /* Stored in network byte order */
3130 __u32 local_port; /* stored in host byte order */
3131 __u32 is_fullsock; /* Some TCP fields are only valid if
3132 * there is a full socket. If not, the
3133 * fields read as zero.
3136 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
3137 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3146 __u32 rate_delivered;
3147 __u32 rate_interval_us;
3150 __u32 total_retrans;
3154 __u32 data_segs_out;
3158 __u64 bytes_received;
3162 /* Definitions for bpf_sock_ops_cb_flags */
3163 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
3164 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
3165 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3166 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
3167 * supported cb flags
3170 /* List of known BPF sock_ops operators.
3171 * New entries can only be added at the end
3175 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3176 * -1 if default value should be used
3178 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3179 * window (in packets) or -1 if default
3180 * value should be used
3182 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3183 * active connection is initialized
3185 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3186 * active connection is
3189 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3190 * passive connection is
3193 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3196 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3197 * based on the path and may be
3198 * dependent on the congestion control
3199 * algorithm. In general it indicates
3200 * a congestion threshold. RTTs above
3201 * this indicate congestion
3203 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3204 * Arg1: value of icsk_retransmits
3205 * Arg2: value of icsk_rto
3206 * Arg3: whether RTO has expired
3208 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3209 * Arg1: sequence number of 1st byte
3211 * Arg3: return value of
3212 * tcp_transmit_skb (0 => success)
3214 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3218 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3219 * socket transition to LISTEN state.
3223 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3224 * changes between the TCP and BPF versions. Ideally this should never happen.
3225 * If it does, we need to add code to convert them before calling
3226 * the BPF sock_ops function.
3229 BPF_TCP_ESTABLISHED = 1,
3239 BPF_TCP_CLOSING, /* Now a valid state */
3240 BPF_TCP_NEW_SYN_RECV,
3242 BPF_TCP_MAX_STATES /* Leave at the end! */
3245 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3246 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3248 struct bpf_perf_event_value {
3254 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3255 #define BPF_DEVCG_ACC_READ (1ULL << 1)
3256 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3258 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3259 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3261 struct bpf_cgroup_dev_ctx {
3262 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3268 struct bpf_raw_tracepoint_args {
3272 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3273 * OUTPUT: Do lookup from egress perspective; default is ingress
3275 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
3276 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
3279 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3280 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3281 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3282 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3283 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3284 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3285 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3286 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3287 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3290 struct bpf_fib_lookup {
3291 /* input: network family for lookup (AF_INET, AF_INET6)
3292 * output: network family of egress nexthop
3296 /* set if lookup is to consider L4 data - e.g., FIB rules */
3301 /* total length of packet from network header - used for MTU check */
3304 /* input: L3 device index for lookup
3305 * output: device index from FIB lookup
3310 /* inputs to lookup */
3311 __u8 tos; /* AF_INET */
3312 __be32 flowinfo; /* AF_INET6, flow_label + priority */
3314 /* output: metric of fib result (IPv4/IPv6 only) */
3320 __u32 ipv6_src[4]; /* in6_addr; network order */
3323 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3324 * network header. output: bpf_fib_lookup sets to gateway address
3325 * if FIB lookup returns gateway route
3329 __u32 ipv6_dst[4]; /* in6_addr; network order */
3333 __be16 h_vlan_proto;
3335 __u8 smac[6]; /* ETH_ALEN */
3336 __u8 dmac[6]; /* ETH_ALEN */
3339 enum bpf_task_fd_type {
3340 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3341 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3342 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3343 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3344 BPF_FD_TYPE_UPROBE, /* filename + offset */
3345 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3348 struct bpf_flow_keys {
3351 __u16 addr_proto; /* ETH_P_* of valid addrs */
3365 __u32 ipv6_src[4]; /* in6_addr; network order */
3366 __u32 ipv6_dst[4]; /* in6_addr; network order */
3371 struct bpf_func_info {
3376 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3377 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3379 struct bpf_line_info {
3381 __u32 file_name_off;
3386 struct bpf_spin_lock {
3391 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3392 * Allows 1,2,4-byte read, but no write.
3394 __u32 file_pos; /* Sysctl file position to read from, write to.
3395 * Allows 1,2,4-byte read an 4-byte write.
3399 #endif /* _UAPI__LINUX_BPF_H__ */