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,
116 BPF_MAP_TYPE_PROG_ARRAY,
117 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
118 BPF_MAP_TYPE_PERCPU_HASH,
119 BPF_MAP_TYPE_PERCPU_ARRAY,
120 BPF_MAP_TYPE_STACK_TRACE,
121 BPF_MAP_TYPE_CGROUP_ARRAY,
122 BPF_MAP_TYPE_LRU_HASH,
123 BPF_MAP_TYPE_LRU_PERCPU_HASH,
124 BPF_MAP_TYPE_LPM_TRIE,
125 BPF_MAP_TYPE_ARRAY_OF_MAPS,
126 BPF_MAP_TYPE_HASH_OF_MAPS,
128 BPF_MAP_TYPE_SOCKMAP,
131 BPF_MAP_TYPE_SOCKHASH,
132 BPF_MAP_TYPE_CGROUP_STORAGE,
133 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
134 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
137 BPF_MAP_TYPE_SK_STORAGE,
138 BPF_MAP_TYPE_DEVMAP_HASH,
141 /* Note that tracing related programs such as
142 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
143 * are not subject to a stable API since kernel internal data
144 * structures can change from release to release and may
145 * therefore break existing tracing BPF programs. Tracing BPF
146 * programs correspond to /a/ specific kernel which is to be
147 * analyzed, and not /a/ specific kernel /and/ all future ones.
150 BPF_PROG_TYPE_UNSPEC,
151 BPF_PROG_TYPE_SOCKET_FILTER,
152 BPF_PROG_TYPE_KPROBE,
153 BPF_PROG_TYPE_SCHED_CLS,
154 BPF_PROG_TYPE_SCHED_ACT,
155 BPF_PROG_TYPE_TRACEPOINT,
157 BPF_PROG_TYPE_PERF_EVENT,
158 BPF_PROG_TYPE_CGROUP_SKB,
159 BPF_PROG_TYPE_CGROUP_SOCK,
160 BPF_PROG_TYPE_LWT_IN,
161 BPF_PROG_TYPE_LWT_OUT,
162 BPF_PROG_TYPE_LWT_XMIT,
163 BPF_PROG_TYPE_SOCK_OPS,
164 BPF_PROG_TYPE_SK_SKB,
165 BPF_PROG_TYPE_CGROUP_DEVICE,
166 BPF_PROG_TYPE_SK_MSG,
167 BPF_PROG_TYPE_RAW_TRACEPOINT,
168 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
169 BPF_PROG_TYPE_LWT_SEG6LOCAL,
170 BPF_PROG_TYPE_LIRC_MODE2,
171 BPF_PROG_TYPE_SK_REUSEPORT,
172 BPF_PROG_TYPE_FLOW_DISSECTOR,
173 BPF_PROG_TYPE_CGROUP_SYSCTL,
174 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
175 BPF_PROG_TYPE_CGROUP_SOCKOPT,
176 BPF_PROG_TYPE_TRACING,
179 enum bpf_attach_type {
180 BPF_CGROUP_INET_INGRESS,
181 BPF_CGROUP_INET_EGRESS,
182 BPF_CGROUP_INET_SOCK_CREATE,
184 BPF_SK_SKB_STREAM_PARSER,
185 BPF_SK_SKB_STREAM_VERDICT,
188 BPF_CGROUP_INET4_BIND,
189 BPF_CGROUP_INET6_BIND,
190 BPF_CGROUP_INET4_CONNECT,
191 BPF_CGROUP_INET6_CONNECT,
192 BPF_CGROUP_INET4_POST_BIND,
193 BPF_CGROUP_INET6_POST_BIND,
194 BPF_CGROUP_UDP4_SENDMSG,
195 BPF_CGROUP_UDP6_SENDMSG,
199 BPF_CGROUP_UDP4_RECVMSG,
200 BPF_CGROUP_UDP6_RECVMSG,
201 BPF_CGROUP_GETSOCKOPT,
202 BPF_CGROUP_SETSOCKOPT,
204 __MAX_BPF_ATTACH_TYPE
207 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
209 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
211 * NONE(default): No further bpf programs allowed in the subtree.
213 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
214 * the program in this cgroup yields to sub-cgroup program.
216 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
217 * that cgroup program gets run in addition to the program in this cgroup.
219 * Only one program is allowed to be attached to a cgroup with
220 * NONE or BPF_F_ALLOW_OVERRIDE flag.
221 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
222 * release old program and attach the new one. Attach flags has to match.
224 * Multiple programs are allowed to be attached to a cgroup with
225 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
226 * (those that were attached first, run first)
227 * The programs of sub-cgroup are executed first, then programs of
228 * this cgroup and then programs of parent cgroup.
229 * When children program makes decision (like picking TCP CA or sock bind)
230 * parent program has a chance to override it.
232 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
233 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
235 * cgrp1 (MULTI progs A, B) ->
236 * cgrp2 (OVERRIDE prog C) ->
237 * cgrp3 (MULTI prog D) ->
238 * cgrp4 (OVERRIDE prog E) ->
239 * cgrp5 (NONE prog F)
240 * the event in cgrp5 triggers execution of F,D,A,B in that order.
241 * if prog F is detached, the execution is E,D,A,B
242 * if prog F and D are detached, the execution is E,A,B
243 * if prog F, E and D are detached, the execution is C,A,B
245 * All eligible programs are executed regardless of return code from
248 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
249 #define BPF_F_ALLOW_MULTI (1U << 1)
251 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
252 * verifier will perform strict alignment checking as if the kernel
253 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
254 * and NET_IP_ALIGN defined to 2.
256 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
258 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
259 * verifier will allow any alignment whatsoever. On platforms
260 * with strict alignment requirements for loads ands stores (such
261 * as sparc and mips) the verifier validates that all loads and
262 * stores provably follow this requirement. This flag turns that
263 * checking and enforcement off.
265 * It is mostly used for testing when we want to validate the
266 * context and memory access aspects of the verifier, but because
267 * of an unaligned access the alignment check would trigger before
268 * the one we are interested in.
270 #define BPF_F_ANY_ALIGNMENT (1U << 1)
272 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
273 * Verifier does sub-register def/use analysis and identifies instructions whose
274 * def only matters for low 32-bit, high 32-bit is never referenced later
275 * through implicit zero extension. Therefore verifier notifies JIT back-ends
276 * that it is safe to ignore clearing high 32-bit for these instructions. This
277 * saves some back-ends a lot of code-gen. However such optimization is not
278 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
279 * hence hasn't used verifier's analysis result. But, we really want to have a
280 * way to be able to verify the correctness of the described optimization on
281 * x86_64 on which testsuites are frequently exercised.
283 * So, this flag is introduced. Once it is set, verifier will randomize high
284 * 32-bit for those instructions who has been identified as safe to ignore them.
285 * Then, if verifier is not doing correct analysis, such randomization will
286 * regress tests to expose bugs.
288 #define BPF_F_TEST_RND_HI32 (1U << 2)
290 /* The verifier internal test flag. Behavior is undefined */
291 #define BPF_F_TEST_STATE_FREQ (1U << 3)
293 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
296 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
297 * insn[0].imm: map fd map fd
298 * insn[1].imm: 0 offset into value
301 * ldimm64 rewrite: address of map address of map[0]+offset
302 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
304 #define BPF_PSEUDO_MAP_FD 1
305 #define BPF_PSEUDO_MAP_VALUE 2
307 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
308 * offset to another bpf function
310 #define BPF_PSEUDO_CALL 1
312 /* flags for BPF_MAP_UPDATE_ELEM command */
313 #define BPF_ANY 0 /* create new element or update existing */
314 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
315 #define BPF_EXIST 2 /* update existing element */
316 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
318 /* flags for BPF_MAP_CREATE command */
319 #define BPF_F_NO_PREALLOC (1U << 0)
320 /* Instead of having one common LRU list in the
321 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
322 * which can scale and perform better.
323 * Note, the LRU nodes (including free nodes) cannot be moved
324 * across different LRU lists.
326 #define BPF_F_NO_COMMON_LRU (1U << 1)
327 /* Specify numa node during map creation */
328 #define BPF_F_NUMA_NODE (1U << 2)
330 #define BPF_OBJ_NAME_LEN 16U
332 /* Flags for accessing BPF object from syscall side. */
333 #define BPF_F_RDONLY (1U << 3)
334 #define BPF_F_WRONLY (1U << 4)
336 /* Flag for stack_map, store build_id+offset instead of pointer */
337 #define BPF_F_STACK_BUILD_ID (1U << 5)
339 /* Zero-initialize hash function seed. This should only be used for testing. */
340 #define BPF_F_ZERO_SEED (1U << 6)
342 /* Flags for accessing BPF object from program side. */
343 #define BPF_F_RDONLY_PROG (1U << 7)
344 #define BPF_F_WRONLY_PROG (1U << 8)
346 /* Clone map from listener for newly accepted socket */
347 #define BPF_F_CLONE (1U << 9)
349 /* flags for BPF_PROG_QUERY */
350 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
352 enum bpf_stack_build_id_status {
353 /* user space need an empty entry to identify end of a trace */
354 BPF_STACK_BUILD_ID_EMPTY = 0,
355 /* with valid build_id and offset */
356 BPF_STACK_BUILD_ID_VALID = 1,
357 /* couldn't get build_id, fallback to ip */
358 BPF_STACK_BUILD_ID_IP = 2,
361 #define BPF_BUILD_ID_SIZE 20
362 struct bpf_stack_build_id {
364 unsigned char build_id[BPF_BUILD_ID_SIZE];
372 struct { /* anonymous struct used by BPF_MAP_CREATE command */
373 __u32 map_type; /* one of enum bpf_map_type */
374 __u32 key_size; /* size of key in bytes */
375 __u32 value_size; /* size of value in bytes */
376 __u32 max_entries; /* max number of entries in a map */
377 __u32 map_flags; /* BPF_MAP_CREATE related
378 * flags defined above.
380 __u32 inner_map_fd; /* fd pointing to the inner map */
381 __u32 numa_node; /* numa node (effective only if
382 * BPF_F_NUMA_NODE is set).
384 char map_name[BPF_OBJ_NAME_LEN];
385 __u32 map_ifindex; /* ifindex of netdev to create on */
386 __u32 btf_fd; /* fd pointing to a BTF type data */
387 __u32 btf_key_type_id; /* BTF type_id of the key */
388 __u32 btf_value_type_id; /* BTF type_id of the value */
391 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
396 __aligned_u64 next_key;
401 struct { /* anonymous struct used by BPF_PROG_LOAD command */
402 __u32 prog_type; /* one of enum bpf_prog_type */
405 __aligned_u64 license;
406 __u32 log_level; /* verbosity level of verifier */
407 __u32 log_size; /* size of user buffer */
408 __aligned_u64 log_buf; /* user supplied buffer */
409 __u32 kern_version; /* not used */
411 char prog_name[BPF_OBJ_NAME_LEN];
412 __u32 prog_ifindex; /* ifindex of netdev to prep for */
413 /* For some prog types expected attach type must be known at
414 * load time to verify attach type specific parts of prog
415 * (context accesses, allowed helpers, etc).
417 __u32 expected_attach_type;
418 __u32 prog_btf_fd; /* fd pointing to BTF type data */
419 __u32 func_info_rec_size; /* userspace bpf_func_info size */
420 __aligned_u64 func_info; /* func info */
421 __u32 func_info_cnt; /* number of bpf_func_info records */
422 __u32 line_info_rec_size; /* userspace bpf_line_info size */
423 __aligned_u64 line_info; /* line info */
424 __u32 line_info_cnt; /* number of bpf_line_info records */
425 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
428 struct { /* anonymous struct used by BPF_OBJ_* commands */
429 __aligned_u64 pathname;
434 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
435 __u32 target_fd; /* container object to attach to */
436 __u32 attach_bpf_fd; /* eBPF program to attach */
441 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
444 __u32 data_size_in; /* input: len of data_in */
445 __u32 data_size_out; /* input/output: len of data_out
446 * returns ENOSPC if data_out
449 __aligned_u64 data_in;
450 __aligned_u64 data_out;
453 __u32 ctx_size_in; /* input: len of ctx_in */
454 __u32 ctx_size_out; /* input/output: len of ctx_out
455 * returns ENOSPC if ctx_out
458 __aligned_u64 ctx_in;
459 __aligned_u64 ctx_out;
462 struct { /* anonymous struct used by BPF_*_GET_*_ID */
473 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
479 struct { /* anonymous struct used by BPF_PROG_QUERY command */
480 __u32 target_fd; /* container object to query */
484 __aligned_u64 prog_ids;
493 struct { /* anonymous struct for BPF_BTF_LOAD */
495 __aligned_u64 btf_log_buf;
502 __u32 pid; /* input: pid */
503 __u32 fd; /* input: fd */
504 __u32 flags; /* input: flags */
505 __u32 buf_len; /* input/output: buf len */
506 __aligned_u64 buf; /* input/output:
507 * tp_name for tracepoint
509 * filename for uprobe
511 __u32 prog_id; /* output: prod_id */
512 __u32 fd_type; /* output: BPF_FD_TYPE_* */
513 __u64 probe_offset; /* output: probe_offset */
514 __u64 probe_addr; /* output: probe_addr */
516 } __attribute__((aligned(8)));
518 /* The description below is an attempt at providing documentation to eBPF
519 * developers about the multiple available eBPF helper functions. It can be
520 * parsed and used to produce a manual page. The workflow is the following,
521 * and requires the rst2man utility:
523 * $ ./scripts/bpf_helpers_doc.py \
524 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
525 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
526 * $ man /tmp/bpf-helpers.7
528 * Note that in order to produce this external documentation, some RST
529 * formatting is used in the descriptions to get "bold" and "italics" in
530 * manual pages. Also note that the few trailing white spaces are
531 * intentional, removing them would break paragraphs for rst2man.
533 * Start of BPF helper function descriptions:
535 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
537 * Perform a lookup in *map* for an entry associated to *key*.
539 * Map value associated to *key*, or **NULL** if no entry was
542 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
544 * Add or update the value of the entry associated to *key* in
545 * *map* with *value*. *flags* is one of:
548 * The entry for *key* must not exist in the map.
550 * The entry for *key* must already exist in the map.
552 * No condition on the existence of the entry for *key*.
554 * Flag value **BPF_NOEXIST** cannot be used for maps of types
555 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
556 * elements always exist), the helper would return an error.
558 * 0 on success, or a negative error in case of failure.
560 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
562 * Delete entry with *key* from *map*.
564 * 0 on success, or a negative error in case of failure.
566 * int bpf_probe_read(void *dst, u32 size, const void *src)
568 * For tracing programs, safely attempt to read *size* bytes from
569 * address *src* and store the data in *dst*.
571 * 0 on success, or a negative error in case of failure.
573 * u64 bpf_ktime_get_ns(void)
575 * Return the time elapsed since system boot, in nanoseconds.
579 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
581 * This helper is a "printk()-like" facility for debugging. It
582 * prints a message defined by format *fmt* (of size *fmt_size*)
583 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
584 * available. It can take up to three additional **u64**
585 * arguments (as an eBPF helpers, the total number of arguments is
588 * Each time the helper is called, it appends a line to the trace.
589 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
590 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
591 * The format of the trace is customizable, and the exact output
592 * one will get depends on the options set in
593 * *\/sys/kernel/debug/tracing/trace_options* (see also the
594 * *README* file under the same directory). However, it usually
595 * defaults to something like:
599 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
603 * * ``telnet`` is the name of the current task.
604 * * ``470`` is the PID of the current task.
605 * * ``001`` is the CPU number on which the task is
607 * * In ``.N..``, each character refers to a set of
608 * options (whether irqs are enabled, scheduling
609 * options, whether hard/softirqs are running, level of
610 * preempt_disabled respectively). **N** means that
611 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
613 * * ``419421.045894`` is a timestamp.
614 * * ``0x00000001`` is a fake value used by BPF for the
615 * instruction pointer register.
616 * * ``<formatted msg>`` is the message formatted with
619 * The conversion specifiers supported by *fmt* are similar, but
620 * more limited than for printk(). They are **%d**, **%i**,
621 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
622 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
623 * of field, padding with zeroes, etc.) is available, and the
624 * helper will return **-EINVAL** (but print nothing) if it
625 * encounters an unknown specifier.
627 * Also, note that **bpf_trace_printk**\ () is slow, and should
628 * only be used for debugging purposes. For this reason, a notice
629 * bloc (spanning several lines) is printed to kernel logs and
630 * states that the helper should not be used "for production use"
631 * the first time this helper is used (or more precisely, when
632 * **trace_printk**\ () buffers are allocated). For passing values
633 * to user space, perf events should be preferred.
635 * The number of bytes written to the buffer, or a negative error
636 * in case of failure.
638 * u32 bpf_get_prandom_u32(void)
640 * Get a pseudo-random number.
642 * From a security point of view, this helper uses its own
643 * pseudo-random internal state, and cannot be used to infer the
644 * seed of other random functions in the kernel. However, it is
645 * essential to note that the generator used by the helper is not
646 * cryptographically secure.
648 * A random 32-bit unsigned value.
650 * u32 bpf_get_smp_processor_id(void)
652 * Get the SMP (symmetric multiprocessing) processor id. Note that
653 * all programs run with preemption disabled, which means that the
654 * SMP processor id is stable during all the execution of the
657 * The SMP id of the processor running the program.
659 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
661 * Store *len* bytes from address *from* into the packet
662 * associated to *skb*, at *offset*. *flags* are a combination of
663 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
664 * checksum for the packet after storing the bytes) and
665 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
666 * **->swhash** and *skb*\ **->l4hash** to 0).
668 * A call to this helper is susceptible to change the underlying
669 * packet buffer. Therefore, at load time, all checks on pointers
670 * previously done by the verifier are invalidated and must be
671 * performed again, if the helper is used in combination with
672 * direct packet access.
674 * 0 on success, or a negative error in case of failure.
676 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
678 * Recompute the layer 3 (e.g. IP) checksum for the packet
679 * associated to *skb*. Computation is incremental, so the helper
680 * must know the former value of the header field that was
681 * modified (*from*), the new value of this field (*to*), and the
682 * number of bytes (2 or 4) for this field, stored in *size*.
683 * Alternatively, it is possible to store the difference between
684 * the previous and the new values of the header field in *to*, by
685 * setting *from* and *size* to 0. For both methods, *offset*
686 * indicates the location of the IP checksum within the packet.
688 * This helper works in combination with **bpf_csum_diff**\ (),
689 * which does not update the checksum in-place, but offers more
690 * flexibility and can handle sizes larger than 2 or 4 for the
691 * checksum to update.
693 * A call to this helper is susceptible to change the underlying
694 * packet buffer. Therefore, at load time, all checks on pointers
695 * previously done by the verifier are invalidated and must be
696 * performed again, if the helper is used in combination with
697 * direct packet access.
699 * 0 on success, or a negative error in case of failure.
701 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
703 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
704 * packet associated to *skb*. Computation is incremental, so the
705 * helper must know the former value of the header field that was
706 * modified (*from*), the new value of this field (*to*), and the
707 * number of bytes (2 or 4) for this field, stored on the lowest
708 * four bits of *flags*. Alternatively, it is possible to store
709 * the difference between the previous and the new values of the
710 * header field in *to*, by setting *from* and the four lowest
711 * bits of *flags* to 0. For both methods, *offset* indicates the
712 * location of the IP checksum within the packet. In addition to
713 * the size of the field, *flags* can be added (bitwise OR) actual
714 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
715 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
716 * for updates resulting in a null checksum the value is set to
717 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
718 * the checksum is to be computed against a pseudo-header.
720 * This helper works in combination with **bpf_csum_diff**\ (),
721 * which does not update the checksum in-place, but offers more
722 * flexibility and can handle sizes larger than 2 or 4 for the
723 * checksum to update.
725 * A call to this helper is susceptible to change the underlying
726 * packet buffer. Therefore, at load time, all checks on pointers
727 * previously done by the verifier are invalidated and must be
728 * performed again, if the helper is used in combination with
729 * direct packet access.
731 * 0 on success, or a negative error in case of failure.
733 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
735 * This special helper is used to trigger a "tail call", or in
736 * other words, to jump into another eBPF program. The same stack
737 * frame is used (but values on stack and in registers for the
738 * caller are not accessible to the callee). This mechanism allows
739 * for program chaining, either for raising the maximum number of
740 * available eBPF instructions, or to execute given programs in
741 * conditional blocks. For security reasons, there is an upper
742 * limit to the number of successive tail calls that can be
745 * Upon call of this helper, the program attempts to jump into a
746 * program referenced at index *index* in *prog_array_map*, a
747 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
748 * *ctx*, a pointer to the context.
750 * If the call succeeds, the kernel immediately runs the first
751 * instruction of the new program. This is not a function call,
752 * and it never returns to the previous program. If the call
753 * fails, then the helper has no effect, and the caller continues
754 * to run its subsequent instructions. A call can fail if the
755 * destination program for the jump does not exist (i.e. *index*
756 * is superior to the number of entries in *prog_array_map*), or
757 * if the maximum number of tail calls has been reached for this
758 * chain of programs. This limit is defined in the kernel by the
759 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
760 * which is currently set to 32.
762 * 0 on success, or a negative error in case of failure.
764 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
766 * Clone and redirect the packet associated to *skb* to another
767 * net device of index *ifindex*. Both ingress and egress
768 * interfaces can be used for redirection. The **BPF_F_INGRESS**
769 * value in *flags* is used to make the distinction (ingress path
770 * is selected if the flag is present, egress path otherwise).
771 * This is the only flag supported for now.
773 * In comparison with **bpf_redirect**\ () helper,
774 * **bpf_clone_redirect**\ () has the associated cost of
775 * duplicating the packet buffer, but this can be executed out of
776 * the eBPF program. Conversely, **bpf_redirect**\ () is more
777 * efficient, but it is handled through an action code where the
778 * redirection happens only after the eBPF program has returned.
780 * A call to this helper is susceptible to change the underlying
781 * packet buffer. Therefore, at load time, all checks on pointers
782 * previously done by the verifier are invalidated and must be
783 * performed again, if the helper is used in combination with
784 * direct packet access.
786 * 0 on success, or a negative error in case of failure.
788 * u64 bpf_get_current_pid_tgid(void)
790 * A 64-bit integer containing the current tgid and pid, and
792 * *current_task*\ **->tgid << 32 \|**
793 * *current_task*\ **->pid**.
795 * u64 bpf_get_current_uid_gid(void)
797 * A 64-bit integer containing the current GID and UID, and
798 * created as such: *current_gid* **<< 32 \|** *current_uid*.
800 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
802 * Copy the **comm** attribute of the current task into *buf* of
803 * *size_of_buf*. The **comm** attribute contains the name of
804 * the executable (excluding the path) for the current task. The
805 * *size_of_buf* must be strictly positive. On success, the
806 * helper makes sure that the *buf* is NUL-terminated. On failure,
807 * it is filled with zeroes.
809 * 0 on success, or a negative error in case of failure.
811 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
813 * Retrieve the classid for the current task, i.e. for the net_cls
814 * cgroup to which *skb* belongs.
816 * This helper can be used on TC egress path, but not on ingress.
818 * The net_cls cgroup provides an interface to tag network packets
819 * based on a user-provided identifier for all traffic coming from
820 * the tasks belonging to the related cgroup. See also the related
821 * kernel documentation, available from the Linux sources in file
822 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
824 * The Linux kernel has two versions for cgroups: there are
825 * cgroups v1 and cgroups v2. Both are available to users, who can
826 * use a mixture of them, but note that the net_cls cgroup is for
827 * cgroup v1 only. This makes it incompatible with BPF programs
828 * run on cgroups, which is a cgroup-v2-only feature (a socket can
829 * only hold data for one version of cgroups at a time).
831 * This helper is only available is the kernel was compiled with
832 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
833 * "**y**" or to "**m**".
835 * The classid, or 0 for the default unconfigured classid.
837 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
839 * Push a *vlan_tci* (VLAN tag control information) of protocol
840 * *vlan_proto* to the packet associated to *skb*, then update
841 * the checksum. Note that if *vlan_proto* is different from
842 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
843 * be **ETH_P_8021Q**.
845 * A call to this helper is susceptible to change the underlying
846 * packet buffer. Therefore, at load time, all checks on pointers
847 * previously done by the verifier are invalidated and must be
848 * performed again, if the helper is used in combination with
849 * direct packet access.
851 * 0 on success, or a negative error in case of failure.
853 * int bpf_skb_vlan_pop(struct sk_buff *skb)
855 * Pop a VLAN header from the packet associated to *skb*.
857 * A call to this helper is susceptible to change the underlying
858 * packet buffer. Therefore, at load time, all checks on pointers
859 * previously done by the verifier are invalidated and must be
860 * performed again, if the helper is used in combination with
861 * direct packet access.
863 * 0 on success, or a negative error in case of failure.
865 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
867 * Get tunnel metadata. This helper takes a pointer *key* to an
868 * empty **struct bpf_tunnel_key** of **size**, that will be
869 * filled with tunnel metadata for the packet associated to *skb*.
870 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
871 * indicates that the tunnel is based on IPv6 protocol instead of
874 * The **struct bpf_tunnel_key** is an object that generalizes the
875 * principal parameters used by various tunneling protocols into a
876 * single struct. This way, it can be used to easily make a
877 * decision based on the contents of the encapsulation header,
878 * "summarized" in this struct. In particular, it holds the IP
879 * address of the remote end (IPv4 or IPv6, depending on the case)
880 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
881 * this struct exposes the *key*\ **->tunnel_id**, which is
882 * generally mapped to a VNI (Virtual Network Identifier), making
883 * it programmable together with the **bpf_skb_set_tunnel_key**\
886 * Let's imagine that the following code is part of a program
887 * attached to the TC ingress interface, on one end of a GRE
888 * tunnel, and is supposed to filter out all messages coming from
889 * remote ends with IPv4 address other than 10.0.0.1:
894 * struct bpf_tunnel_key key = {};
896 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
898 * return TC_ACT_SHOT; // drop packet
900 * if (key.remote_ipv4 != 0x0a000001)
901 * return TC_ACT_SHOT; // drop packet
903 * return TC_ACT_OK; // accept packet
905 * This interface can also be used with all encapsulation devices
906 * that can operate in "collect metadata" mode: instead of having
907 * one network device per specific configuration, the "collect
908 * metadata" mode only requires a single device where the
909 * configuration can be extracted from this helper.
911 * This can be used together with various tunnels such as VXLan,
912 * Geneve, GRE or IP in IP (IPIP).
914 * 0 on success, or a negative error in case of failure.
916 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
918 * Populate tunnel metadata for packet associated to *skb.* The
919 * tunnel metadata is set to the contents of *key*, of *size*. The
920 * *flags* can be set to a combination of the following values:
922 * **BPF_F_TUNINFO_IPV6**
923 * Indicate that the tunnel is based on IPv6 protocol
925 * **BPF_F_ZERO_CSUM_TX**
926 * For IPv4 packets, add a flag to tunnel metadata
927 * indicating that checksum computation should be skipped
928 * and checksum set to zeroes.
929 * **BPF_F_DONT_FRAGMENT**
930 * Add a flag to tunnel metadata indicating that the
931 * packet should not be fragmented.
932 * **BPF_F_SEQ_NUMBER**
933 * Add a flag to tunnel metadata indicating that a
934 * sequence number should be added to tunnel header before
935 * sending the packet. This flag was added for GRE
936 * encapsulation, but might be used with other protocols
937 * as well in the future.
939 * Here is a typical usage on the transmit path:
943 * struct bpf_tunnel_key key;
945 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
946 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
948 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
949 * helper for additional information.
951 * 0 on success, or a negative error in case of failure.
953 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
955 * Read the value of a perf event counter. This helper relies on a
956 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
957 * the perf event counter is selected when *map* is updated with
958 * perf event file descriptors. The *map* is an array whose size
959 * is the number of available CPUs, and each cell contains a value
960 * relative to one CPU. The value to retrieve is indicated by
961 * *flags*, that contains the index of the CPU to look up, masked
962 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
963 * **BPF_F_CURRENT_CPU** to indicate that the value for the
964 * current CPU should be retrieved.
966 * Note that before Linux 4.13, only hardware perf event can be
969 * Also, be aware that the newer helper
970 * **bpf_perf_event_read_value**\ () is recommended over
971 * **bpf_perf_event_read**\ () in general. The latter has some ABI
972 * quirks where error and counter value are used as a return code
973 * (which is wrong to do since ranges may overlap). This issue is
974 * fixed with **bpf_perf_event_read_value**\ (), which at the same
975 * time provides more features over the **bpf_perf_event_read**\
976 * () interface. Please refer to the description of
977 * **bpf_perf_event_read_value**\ () for details.
979 * The value of the perf event counter read from the map, or a
980 * negative error code in case of failure.
982 * int bpf_redirect(u32 ifindex, u64 flags)
984 * Redirect the packet to another net device of index *ifindex*.
985 * This helper is somewhat similar to **bpf_clone_redirect**\
986 * (), except that the packet is not cloned, which provides
987 * increased performance.
989 * Except for XDP, both ingress and egress interfaces can be used
990 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
991 * to make the distinction (ingress path is selected if the flag
992 * is present, egress path otherwise). Currently, XDP only
993 * supports redirection to the egress interface, and accepts no
996 * The same effect can be attained with the more generic
997 * **bpf_redirect_map**\ (), which requires specific maps to be
998 * used but offers better performance.
1000 * For XDP, the helper returns **XDP_REDIRECT** on success or
1001 * **XDP_ABORTED** on error. For other program types, the values
1002 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1005 * u32 bpf_get_route_realm(struct sk_buff *skb)
1007 * Retrieve the realm or the route, that is to say the
1008 * **tclassid** field of the destination for the *skb*. The
1009 * indentifier retrieved is a user-provided tag, similar to the
1010 * one used with the net_cls cgroup (see description for
1011 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1012 * held by a route (a destination entry), not by a task.
1014 * Retrieving this identifier works with the clsact TC egress hook
1015 * (see also **tc-bpf(8)**), or alternatively on conventional
1016 * classful egress qdiscs, but not on TC ingress path. In case of
1017 * clsact TC egress hook, this has the advantage that, internally,
1018 * the destination entry has not been dropped yet in the transmit
1019 * path. Therefore, the destination entry does not need to be
1020 * artificially held via **netif_keep_dst**\ () for a classful
1021 * qdisc until the *skb* is freed.
1023 * This helper is available only if the kernel was compiled with
1024 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1026 * The realm of the route for the packet associated to *skb*, or 0
1027 * if none was found.
1029 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1031 * Write raw *data* blob into a special BPF perf event held by
1032 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1033 * event must have the following attributes: **PERF_SAMPLE_RAW**
1034 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1035 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1037 * The *flags* are used to indicate the index in *map* for which
1038 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1039 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1040 * to indicate that the index of the current CPU core should be
1043 * The value to write, of *size*, is passed through eBPF stack and
1044 * pointed by *data*.
1046 * The context of the program *ctx* needs also be passed to the
1049 * On user space, a program willing to read the values needs to
1050 * call **perf_event_open**\ () on the perf event (either for
1051 * one or for all CPUs) and to store the file descriptor into the
1052 * *map*. This must be done before the eBPF program can send data
1053 * into it. An example is available in file
1054 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1055 * tree (the eBPF program counterpart is in
1056 * *samples/bpf/trace_output_kern.c*).
1058 * **bpf_perf_event_output**\ () achieves better performance
1059 * than **bpf_trace_printk**\ () for sharing data with user
1060 * space, and is much better suitable for streaming data from eBPF
1063 * Note that this helper is not restricted to tracing use cases
1064 * and can be used with programs attached to TC or XDP as well,
1065 * where it allows for passing data to user space listeners. Data
1068 * * Only custom structs,
1069 * * Only the packet payload, or
1070 * * A combination of both.
1072 * 0 on success, or a negative error in case of failure.
1074 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1076 * This helper was provided as an easy way to load data from a
1077 * packet. It can be used to load *len* bytes from *offset* from
1078 * the packet associated to *skb*, into the buffer pointed by
1081 * Since Linux 4.7, usage of this helper has mostly been replaced
1082 * by "direct packet access", enabling packet data to be
1083 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1084 * pointing respectively to the first byte of packet data and to
1085 * the byte after the last byte of packet data. However, it
1086 * remains useful if one wishes to read large quantities of data
1087 * at once from a packet into the eBPF stack.
1089 * 0 on success, or a negative error in case of failure.
1091 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1093 * Walk a user or a kernel stack and return its id. To achieve
1094 * this, the helper needs *ctx*, which is a pointer to the context
1095 * on which the tracing program is executed, and a pointer to a
1096 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1098 * The last argument, *flags*, holds the number of stack frames to
1099 * skip (from 0 to 255), masked with
1100 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1101 * a combination of the following flags:
1103 * **BPF_F_USER_STACK**
1104 * Collect a user space stack instead of a kernel stack.
1105 * **BPF_F_FAST_STACK_CMP**
1106 * Compare stacks by hash only.
1107 * **BPF_F_REUSE_STACKID**
1108 * If two different stacks hash into the same *stackid*,
1109 * discard the old one.
1111 * The stack id retrieved is a 32 bit long integer handle which
1112 * can be further combined with other data (including other stack
1113 * ids) and used as a key into maps. This can be useful for
1114 * generating a variety of graphs (such as flame graphs or off-cpu
1117 * For walking a stack, this helper is an improvement over
1118 * **bpf_probe_read**\ (), which can be used with unrolled loops
1119 * but is not efficient and consumes a lot of eBPF instructions.
1120 * Instead, **bpf_get_stackid**\ () can collect up to
1121 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1122 * this limit can be controlled with the **sysctl** program, and
1123 * that it should be manually increased in order to profile long
1124 * user stacks (such as stacks for Java programs). To do so, use:
1128 * # sysctl kernel.perf_event_max_stack=<new value>
1130 * The positive or null stack id on success, or a negative error
1131 * in case of failure.
1133 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1135 * Compute a checksum difference, from the raw buffer pointed by
1136 * *from*, of length *from_size* (that must be a multiple of 4),
1137 * towards the raw buffer pointed by *to*, of size *to_size*
1138 * (same remark). An optional *seed* can be added to the value
1139 * (this can be cascaded, the seed may come from a previous call
1142 * This is flexible enough to be used in several ways:
1144 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1145 * checksum, it can be used when pushing new data.
1146 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1147 * checksum, it can be used when removing data from a packet.
1148 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1149 * can be used to compute a diff. Note that *from_size* and
1150 * *to_size* do not need to be equal.
1152 * This helper can be used in combination with
1153 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1154 * which one can feed in the difference computed with
1155 * **bpf_csum_diff**\ ().
1157 * The checksum result, or a negative error code in case of
1160 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1162 * Retrieve tunnel options metadata for the packet associated to
1163 * *skb*, and store the raw tunnel option data to the buffer *opt*
1166 * This helper can be used with encapsulation devices that can
1167 * operate in "collect metadata" mode (please refer to the related
1168 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1169 * more details). A particular example where this can be used is
1170 * in combination with the Geneve encapsulation protocol, where it
1171 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1172 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1173 * the eBPF program. This allows for full customization of these
1176 * The size of the option data retrieved.
1178 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1180 * Set tunnel options metadata for the packet associated to *skb*
1181 * to the option data contained in the raw buffer *opt* of *size*.
1183 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1184 * helper for additional information.
1186 * 0 on success, or a negative error in case of failure.
1188 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1190 * Change the protocol of the *skb* to *proto*. Currently
1191 * supported are transition from IPv4 to IPv6, and from IPv6 to
1192 * IPv4. The helper takes care of the groundwork for the
1193 * transition, including resizing the socket buffer. The eBPF
1194 * program is expected to fill the new headers, if any, via
1195 * **skb_store_bytes**\ () and to recompute the checksums with
1196 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1197 * (). The main case for this helper is to perform NAT64
1198 * operations out of an eBPF program.
1200 * Internally, the GSO type is marked as dodgy so that headers are
1201 * checked and segments are recalculated by the GSO/GRO engine.
1202 * The size for GSO target is adapted as well.
1204 * All values for *flags* are reserved for future usage, and must
1207 * A call to this helper is susceptible to change the underlying
1208 * packet buffer. Therefore, at load time, all checks on pointers
1209 * previously done by the verifier are invalidated and must be
1210 * performed again, if the helper is used in combination with
1211 * direct packet access.
1213 * 0 on success, or a negative error in case of failure.
1215 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1217 * Change the packet type for the packet associated to *skb*. This
1218 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1219 * the eBPF program does not have a write access to *skb*\
1220 * **->pkt_type** beside this helper. Using a helper here allows
1221 * for graceful handling of errors.
1223 * The major use case is to change incoming *skb*s to
1224 * **PACKET_HOST** in a programmatic way instead of having to
1225 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1228 * Note that *type* only allows certain values. At this time, they
1233 * **PACKET_BROADCAST**
1234 * Send packet to all.
1235 * **PACKET_MULTICAST**
1236 * Send packet to group.
1237 * **PACKET_OTHERHOST**
1238 * Send packet to someone else.
1240 * 0 on success, or a negative error in case of failure.
1242 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1244 * Check whether *skb* is a descendant of the cgroup2 held by
1245 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1247 * The return value depends on the result of the test, and can be:
1249 * * 0, if the *skb* failed the cgroup2 descendant test.
1250 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1251 * * A negative error code, if an error occurred.
1253 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1255 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1256 * not set, in particular if the hash was cleared due to mangling,
1257 * recompute this hash. Later accesses to the hash can be done
1258 * directly with *skb*\ **->hash**.
1260 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1261 * prototype with **bpf_skb_change_proto**\ (), or calling
1262 * **bpf_skb_store_bytes**\ () with the
1263 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1264 * the hash and to trigger a new computation for the next call to
1265 * **bpf_get_hash_recalc**\ ().
1269 * u64 bpf_get_current_task(void)
1271 * A pointer to the current task struct.
1273 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1275 * Attempt in a safe way to write *len* bytes from the buffer
1276 * *src* to *dst* in memory. It only works for threads that are in
1277 * user context, and *dst* must be a valid user space address.
1279 * This helper should not be used to implement any kind of
1280 * security mechanism because of TOC-TOU attacks, but rather to
1281 * debug, divert, and manipulate execution of semi-cooperative
1284 * Keep in mind that this feature is meant for experiments, and it
1285 * has a risk of crashing the system and running programs.
1286 * Therefore, when an eBPF program using this helper is attached,
1287 * a warning including PID and process name is printed to kernel
1290 * 0 on success, or a negative error in case of failure.
1292 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1294 * Check whether the probe is being run is the context of a given
1295 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1296 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1298 * The return value depends on the result of the test, and can be:
1300 * * 0, if the *skb* task belongs to the cgroup2.
1301 * * 1, if the *skb* task does not belong to the cgroup2.
1302 * * A negative error code, if an error occurred.
1304 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1306 * Resize (trim or grow) the packet associated to *skb* to the
1307 * new *len*. The *flags* are reserved for future usage, and must
1310 * The basic idea is that the helper performs the needed work to
1311 * change the size of the packet, then the eBPF program rewrites
1312 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1313 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1314 * and others. This helper is a slow path utility intended for
1315 * replies with control messages. And because it is targeted for
1316 * slow path, the helper itself can afford to be slow: it
1317 * implicitly linearizes, unclones and drops offloads from the
1320 * A call to this helper is susceptible to change the underlying
1321 * packet buffer. Therefore, at load time, all checks on pointers
1322 * previously done by the verifier are invalidated and must be
1323 * performed again, if the helper is used in combination with
1324 * direct packet access.
1326 * 0 on success, or a negative error in case of failure.
1328 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1330 * Pull in non-linear data in case the *skb* is non-linear and not
1331 * all of *len* are part of the linear section. Make *len* bytes
1332 * from *skb* readable and writable. If a zero value is passed for
1333 * *len*, then the whole length of the *skb* is pulled.
1335 * This helper is only needed for reading and writing with direct
1338 * For direct packet access, testing that offsets to access
1339 * are within packet boundaries (test on *skb*\ **->data_end**) is
1340 * susceptible to fail if offsets are invalid, or if the requested
1341 * data is in non-linear parts of the *skb*. On failure the
1342 * program can just bail out, or in the case of a non-linear
1343 * buffer, use a helper to make the data available. The
1344 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1345 * the data. Another one consists in using **bpf_skb_pull_data**
1346 * to pull in once the non-linear parts, then retesting and
1347 * eventually access the data.
1349 * At the same time, this also makes sure the *skb* is uncloned,
1350 * which is a necessary condition for direct write. As this needs
1351 * to be an invariant for the write part only, the verifier
1352 * detects writes and adds a prologue that is calling
1353 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1354 * the very beginning in case it is indeed cloned.
1356 * A call to this helper is susceptible to change the underlying
1357 * packet buffer. Therefore, at load time, all checks on pointers
1358 * previously done by the verifier are invalidated and must be
1359 * performed again, if the helper is used in combination with
1360 * direct packet access.
1362 * 0 on success, or a negative error in case of failure.
1364 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1366 * Add the checksum *csum* into *skb*\ **->csum** in case the
1367 * driver has supplied a checksum for the entire packet into that
1368 * field. Return an error otherwise. This helper is intended to be
1369 * used in combination with **bpf_csum_diff**\ (), in particular
1370 * when the checksum needs to be updated after data has been
1371 * written into the packet through direct packet access.
1373 * The checksum on success, or a negative error code in case of
1376 * void bpf_set_hash_invalid(struct sk_buff *skb)
1378 * Invalidate the current *skb*\ **->hash**. It can be used after
1379 * mangling on headers through direct packet access, in order to
1380 * indicate that the hash is outdated and to trigger a
1381 * recalculation the next time the kernel tries to access this
1382 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1384 * int bpf_get_numa_node_id(void)
1386 * Return the id of the current NUMA node. The primary use case
1387 * for this helper is the selection of sockets for the local NUMA
1388 * node, when the program is attached to sockets using the
1389 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1390 * but the helper is also available to other eBPF program types,
1391 * similarly to **bpf_get_smp_processor_id**\ ().
1393 * The id of current NUMA node.
1395 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1397 * Grows headroom of packet associated to *skb* and adjusts the
1398 * offset of the MAC header accordingly, adding *len* bytes of
1399 * space. It automatically extends and reallocates memory as
1402 * This helper can be used on a layer 3 *skb* to push a MAC header
1403 * for redirection into a layer 2 device.
1405 * All values for *flags* are reserved for future usage, and must
1408 * A call to this helper is susceptible to change the underlying
1409 * packet buffer. Therefore, at load time, all checks on pointers
1410 * previously done by the verifier are invalidated and must be
1411 * performed again, if the helper is used in combination with
1412 * direct packet access.
1414 * 0 on success, or a negative error in case of failure.
1416 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1418 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1419 * it is possible to use a negative value for *delta*. This helper
1420 * can be used to prepare the packet for pushing or popping
1423 * A call to this helper is susceptible to change the underlying
1424 * packet buffer. Therefore, at load time, all checks on pointers
1425 * previously done by the verifier are invalidated and must be
1426 * performed again, if the helper is used in combination with
1427 * direct packet access.
1429 * 0 on success, or a negative error in case of failure.
1431 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1433 * Copy a NUL terminated string from an unsafe address
1434 * *unsafe_ptr* to *dst*. The *size* should include the
1435 * terminating NUL byte. In case the string length is smaller than
1436 * *size*, the target is not padded with further NUL bytes. If the
1437 * string length is larger than *size*, just *size*-1 bytes are
1438 * copied and the last byte is set to NUL.
1440 * On success, the length of the copied string is returned. This
1441 * makes this helper useful in tracing programs for reading
1442 * strings, and more importantly to get its length at runtime. See
1443 * the following snippet:
1447 * SEC("kprobe/sys_open")
1448 * void bpf_sys_open(struct pt_regs *ctx)
1450 * char buf[PATHLEN]; // PATHLEN is defined to 256
1451 * int res = bpf_probe_read_str(buf, sizeof(buf),
1454 * // Consume buf, for example push it to
1455 * // userspace via bpf_perf_event_output(); we
1456 * // can use res (the string length) as event
1457 * // size, after checking its boundaries.
1460 * In comparison, using **bpf_probe_read()** helper here instead
1461 * to read the string would require to estimate the length at
1462 * compile time, and would often result in copying more memory
1465 * Another useful use case is when parsing individual process
1466 * arguments or individual environment variables navigating
1467 * *current*\ **->mm->arg_start** and *current*\
1468 * **->mm->env_start**: using this helper and the return value,
1469 * one can quickly iterate at the right offset of the memory area.
1471 * On success, the strictly positive length of the string,
1472 * including the trailing NUL character. On error, a negative
1475 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1477 * If the **struct sk_buff** pointed by *skb* has a known socket,
1478 * retrieve the cookie (generated by the kernel) of this socket.
1479 * If no cookie has been set yet, generate a new cookie. Once
1480 * generated, the socket cookie remains stable for the life of the
1481 * socket. This helper can be useful for monitoring per socket
1482 * networking traffic statistics as it provides a global socket
1483 * identifier that can be assumed unique.
1485 * A 8-byte long non-decreasing number on success, or 0 if the
1486 * socket field is missing inside *skb*.
1488 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1490 * Equivalent to bpf_get_socket_cookie() helper that accepts
1491 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1493 * A 8-byte long non-decreasing number.
1495 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1497 * Equivalent to bpf_get_socket_cookie() helper that accepts
1498 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1500 * A 8-byte long non-decreasing number.
1502 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1504 * The owner UID of the socket associated to *skb*. If the socket
1505 * is **NULL**, or if it is not a full socket (i.e. if it is a
1506 * time-wait or a request socket instead), **overflowuid** value
1507 * is returned (note that **overflowuid** might also be the actual
1508 * UID value for the socket).
1510 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1512 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1517 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1519 * Emulate a call to **setsockopt()** on the socket associated to
1520 * *bpf_socket*, which must be a full socket. The *level* at
1521 * which the option resides and the name *optname* of the option
1522 * must be specified, see **setsockopt(2)** for more information.
1523 * The option value of length *optlen* is pointed by *optval*.
1525 * This helper actually implements a subset of **setsockopt()**.
1526 * It supports the following *level*\ s:
1528 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1529 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1530 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1531 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1532 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1533 * **TCP_BPF_SNDCWND_CLAMP**.
1534 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1535 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1537 * 0 on success, or a negative error in case of failure.
1539 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1541 * Grow or shrink the room for data in the packet associated to
1542 * *skb* by *len_diff*, and according to the selected *mode*.
1544 * There are two supported modes at this time:
1546 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1547 * (room space is added or removed below the layer 2 header).
1549 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1550 * (room space is added or removed below the layer 3 header).
1552 * The following flags are supported at this time:
1554 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1555 * Adjusting mss in this way is not allowed for datagrams.
1557 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1558 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1559 * Any new space is reserved to hold a tunnel header.
1560 * Configure skb offsets and other fields accordingly.
1562 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1563 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1564 * Use with ENCAP_L3 flags to further specify the tunnel type.
1566 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1567 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1568 * type; *len* is the length of the inner MAC header.
1570 * A call to this helper is susceptible to change the underlying
1571 * packet buffer. Therefore, at load time, all checks on pointers
1572 * previously done by the verifier are invalidated and must be
1573 * performed again, if the helper is used in combination with
1574 * direct packet access.
1576 * 0 on success, or a negative error in case of failure.
1578 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1580 * Redirect the packet to the endpoint referenced by *map* at
1581 * index *key*. Depending on its type, this *map* can contain
1582 * references to net devices (for forwarding packets through other
1583 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1584 * but this is only implemented for native XDP (with driver
1585 * support) as of this writing).
1587 * The lower two bits of *flags* are used as the return code if
1588 * the map lookup fails. This is so that the return value can be
1589 * one of the XDP program return codes up to XDP_TX, as chosen by
1590 * the caller. Any higher bits in the *flags* argument must be
1593 * When used to redirect packets to net devices, this helper
1594 * provides a high performance increase over **bpf_redirect**\ ().
1595 * This is due to various implementation details of the underlying
1596 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1597 * () tries to send packet as a "bulk" to the device.
1599 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1601 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1603 * Redirect the packet to the socket referenced by *map* (of type
1604 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1605 * egress interfaces can be used for redirection. The
1606 * **BPF_F_INGRESS** value in *flags* is used to make the
1607 * distinction (ingress path is selected if the flag is present,
1608 * egress path otherwise). This is the only flag supported for now.
1610 * **SK_PASS** on success, or **SK_DROP** on error.
1612 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1614 * Add an entry to, or update a *map* referencing sockets. The
1615 * *skops* is used as a new value for the entry associated to
1616 * *key*. *flags* is one of:
1619 * The entry for *key* must not exist in the map.
1621 * The entry for *key* must already exist in the map.
1623 * No condition on the existence of the entry for *key*.
1625 * If the *map* has eBPF programs (parser and verdict), those will
1626 * be inherited by the socket being added. If the socket is
1627 * already attached to eBPF programs, this results in an error.
1629 * 0 on success, or a negative error in case of failure.
1631 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1633 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1634 * *delta* (which can be positive or negative). Note that this
1635 * operation modifies the address stored in *xdp_md*\ **->data**,
1636 * so the latter must be loaded only after the helper has been
1639 * The use of *xdp_md*\ **->data_meta** is optional and programs
1640 * are not required to use it. The rationale is that when the
1641 * packet is processed with XDP (e.g. as DoS filter), it is
1642 * possible to push further meta data along with it before passing
1643 * to the stack, and to give the guarantee that an ingress eBPF
1644 * program attached as a TC classifier on the same device can pick
1645 * this up for further post-processing. Since TC works with socket
1646 * buffers, it remains possible to set from XDP the **mark** or
1647 * **priority** pointers, or other pointers for the socket buffer.
1648 * Having this scratch space generic and programmable allows for
1649 * more flexibility as the user is free to store whatever meta
1652 * A call to this helper is susceptible to change the underlying
1653 * packet buffer. Therefore, at load time, all checks on pointers
1654 * previously done by the verifier are invalidated and must be
1655 * performed again, if the helper is used in combination with
1656 * direct packet access.
1658 * 0 on success, or a negative error in case of failure.
1660 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1662 * Read the value of a perf event counter, and store it into *buf*
1663 * of size *buf_size*. This helper relies on a *map* of type
1664 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1665 * counter is selected when *map* is updated with perf event file
1666 * descriptors. The *map* is an array whose size is the number of
1667 * available CPUs, and each cell contains a value relative to one
1668 * CPU. The value to retrieve is indicated by *flags*, that
1669 * contains the index of the CPU to look up, masked with
1670 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1671 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1672 * current CPU should be retrieved.
1674 * This helper behaves in a way close to
1675 * **bpf_perf_event_read**\ () helper, save that instead of
1676 * just returning the value observed, it fills the *buf*
1677 * structure. This allows for additional data to be retrieved: in
1678 * particular, the enabled and running times (in *buf*\
1679 * **->enabled** and *buf*\ **->running**, respectively) are
1680 * copied. In general, **bpf_perf_event_read_value**\ () is
1681 * recommended over **bpf_perf_event_read**\ (), which has some
1682 * ABI issues and provides fewer functionalities.
1684 * These values are interesting, because hardware PMU (Performance
1685 * Monitoring Unit) counters are limited resources. When there are
1686 * more PMU based perf events opened than available counters,
1687 * kernel will multiplex these events so each event gets certain
1688 * percentage (but not all) of the PMU time. In case that
1689 * multiplexing happens, the number of samples or counter value
1690 * will not reflect the case compared to when no multiplexing
1691 * occurs. This makes comparison between different runs difficult.
1692 * Typically, the counter value should be normalized before
1693 * comparing to other experiments. The usual normalization is done
1698 * normalized_counter = counter * t_enabled / t_running
1700 * Where t_enabled is the time enabled for event and t_running is
1701 * the time running for event since last normalization. The
1702 * enabled and running times are accumulated since the perf event
1703 * open. To achieve scaling factor between two invocations of an
1704 * eBPF program, users can can use CPU id as the key (which is
1705 * typical for perf array usage model) to remember the previous
1706 * value and do the calculation inside the eBPF program.
1708 * 0 on success, or a negative error in case of failure.
1710 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1712 * For en eBPF program attached to a perf event, retrieve the
1713 * value of the event counter associated to *ctx* and store it in
1714 * the structure pointed by *buf* and of size *buf_size*. Enabled
1715 * and running times are also stored in the structure (see
1716 * description of helper **bpf_perf_event_read_value**\ () for
1719 * 0 on success, or a negative error in case of failure.
1721 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
1723 * Emulate a call to **getsockopt()** on the socket associated to
1724 * *bpf_socket*, which must be a full socket. The *level* at
1725 * which the option resides and the name *optname* of the option
1726 * must be specified, see **getsockopt(2)** for more information.
1727 * The retrieved value is stored in the structure pointed by
1728 * *opval* and of length *optlen*.
1730 * This helper actually implements a subset of **getsockopt()**.
1731 * It supports the following *level*\ s:
1733 * * **IPPROTO_TCP**, which supports *optname*
1734 * **TCP_CONGESTION**.
1735 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1736 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1738 * 0 on success, or a negative error in case of failure.
1740 * int bpf_override_return(struct pt_regs *regs, u64 rc)
1742 * Used for error injection, this helper uses kprobes to override
1743 * the return value of the probed function, and to set it to *rc*.
1744 * The first argument is the context *regs* on which the kprobe
1747 * This helper works by setting setting the PC (program counter)
1748 * to an override function which is run in place of the original
1749 * probed function. This means the probed function is not run at
1750 * all. The replacement function just returns with the required
1753 * This helper has security implications, and thus is subject to
1754 * restrictions. It is only available if the kernel was compiled
1755 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1756 * option, and in this case it only works on functions tagged with
1757 * **ALLOW_ERROR_INJECTION** in the kernel code.
1759 * Also, the helper is only available for the architectures having
1760 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1761 * x86 architecture is the only one to support this feature.
1765 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1767 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1768 * for the full TCP socket associated to *bpf_sock_ops* to
1771 * The primary use of this field is to determine if there should
1772 * be calls to eBPF programs of type
1773 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1774 * code. A program of the same type can change its value, per
1775 * connection and as necessary, when the connection is
1776 * established. This field is directly accessible for reading, but
1777 * this helper must be used for updates in order to return an
1778 * error if an eBPF program tries to set a callback that is not
1779 * supported in the current kernel.
1781 * *argval* is a flag array which can combine these flags:
1783 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1784 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1785 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1786 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1788 * Therefore, this function can be used to clear a callback flag by
1789 * setting the appropriate bit to zero. e.g. to disable the RTO
1792 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1793 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1795 * Here are some examples of where one could call such eBPF
1799 * * When a packet is retransmitted.
1800 * * When the connection terminates.
1801 * * When a packet is sent.
1802 * * When a packet is received.
1804 * Code **-EINVAL** if the socket is not a full TCP socket;
1805 * otherwise, a positive number containing the bits that could not
1806 * be set is returned (which comes down to 0 if all bits were set
1809 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1811 * This helper is used in programs implementing policies at the
1812 * socket level. If the message *msg* is allowed to pass (i.e. if
1813 * the verdict eBPF program returns **SK_PASS**), redirect it to
1814 * the socket referenced by *map* (of type
1815 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1816 * egress interfaces can be used for redirection. The
1817 * **BPF_F_INGRESS** value in *flags* is used to make the
1818 * distinction (ingress path is selected if the flag is present,
1819 * egress path otherwise). This is the only flag supported for now.
1821 * **SK_PASS** on success, or **SK_DROP** on error.
1823 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1825 * For socket policies, apply the verdict of the eBPF program to
1826 * the next *bytes* (number of bytes) of message *msg*.
1828 * For example, this helper can be used in the following cases:
1830 * * A single **sendmsg**\ () or **sendfile**\ () system call
1831 * contains multiple logical messages that the eBPF program is
1832 * supposed to read and for which it should apply a verdict.
1833 * * An eBPF program only cares to read the first *bytes* of a
1834 * *msg*. If the message has a large payload, then setting up
1835 * and calling the eBPF program repeatedly for all bytes, even
1836 * though the verdict is already known, would create unnecessary
1839 * When called from within an eBPF program, the helper sets a
1840 * counter internal to the BPF infrastructure, that is used to
1841 * apply the last verdict to the next *bytes*. If *bytes* is
1842 * smaller than the current data being processed from a
1843 * **sendmsg**\ () or **sendfile**\ () system call, the first
1844 * *bytes* will be sent and the eBPF program will be re-run with
1845 * the pointer for start of data pointing to byte number *bytes*
1846 * **+ 1**. If *bytes* is larger than the current data being
1847 * processed, then the eBPF verdict will be applied to multiple
1848 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1851 * Note that if a socket closes with the internal counter holding
1852 * a non-zero value, this is not a problem because data is not
1853 * being buffered for *bytes* and is sent as it is received.
1857 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1859 * For socket policies, prevent the execution of the verdict eBPF
1860 * program for message *msg* until *bytes* (byte number) have been
1863 * This can be used when one needs a specific number of bytes
1864 * before a verdict can be assigned, even if the data spans
1865 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1866 * case would be a user calling **sendmsg**\ () repeatedly with
1867 * 1-byte long message segments. Obviously, this is bad for
1868 * performance, but it is still valid. If the eBPF program needs
1869 * *bytes* bytes to validate a header, this helper can be used to
1870 * prevent the eBPF program to be called again until *bytes* have
1875 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1877 * For socket policies, pull in non-linear data from user space
1878 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1879 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1882 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1883 * *msg* it can only parse data that the (**data**, **data_end**)
1884 * pointers have already consumed. For **sendmsg**\ () hooks this
1885 * is likely the first scatterlist element. But for calls relying
1886 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1887 * be the range (**0**, **0**) because the data is shared with
1888 * user space and by default the objective is to avoid allowing
1889 * user space to modify data while (or after) eBPF verdict is
1890 * being decided. This helper can be used to pull in data and to
1891 * set the start and end pointer to given values. Data will be
1892 * copied if necessary (i.e. if data was not linear and if start
1893 * and end pointers do not point to the same chunk).
1895 * A call to this helper is susceptible to change the underlying
1896 * packet buffer. Therefore, at load time, all checks on pointers
1897 * previously done by the verifier are invalidated and must be
1898 * performed again, if the helper is used in combination with
1899 * direct packet access.
1901 * All values for *flags* are reserved for future usage, and must
1904 * 0 on success, or a negative error in case of failure.
1906 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1908 * Bind the socket associated to *ctx* to the address pointed by
1909 * *addr*, of length *addr_len*. This allows for making outgoing
1910 * connection from the desired IP address, which can be useful for
1911 * example when all processes inside a cgroup should use one
1912 * single IP address on a host that has multiple IP configured.
1914 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1915 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1916 * **AF_INET6**). Looking for a free port to bind to can be
1917 * expensive, therefore binding to port is not permitted by the
1918 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1919 * must be set to zero.
1921 * 0 on success, or a negative error in case of failure.
1923 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1925 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1926 * only possible to shrink the packet as of this writing,
1927 * therefore *delta* must be a negative integer.
1929 * A call to this helper is susceptible to change the underlying
1930 * packet buffer. Therefore, at load time, all checks on pointers
1931 * previously done by the verifier are invalidated and must be
1932 * performed again, if the helper is used in combination with
1933 * direct packet access.
1935 * 0 on success, or a negative error in case of failure.
1937 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1939 * Retrieve the XFRM state (IP transform framework, see also
1940 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1942 * The retrieved value is stored in the **struct bpf_xfrm_state**
1943 * pointed by *xfrm_state* and of length *size*.
1945 * All values for *flags* are reserved for future usage, and must
1948 * This helper is available only if the kernel was compiled with
1949 * **CONFIG_XFRM** configuration option.
1951 * 0 on success, or a negative error in case of failure.
1953 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
1955 * Return a user or a kernel stack in bpf program provided buffer.
1956 * To achieve this, the helper needs *ctx*, which is a pointer
1957 * to the context on which the tracing program is executed.
1958 * To store the stacktrace, the bpf program provides *buf* with
1959 * a nonnegative *size*.
1961 * The last argument, *flags*, holds the number of stack frames to
1962 * skip (from 0 to 255), masked with
1963 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1964 * the following flags:
1966 * **BPF_F_USER_STACK**
1967 * Collect a user space stack instead of a kernel stack.
1968 * **BPF_F_USER_BUILD_ID**
1969 * Collect buildid+offset instead of ips for user stack,
1970 * only valid if **BPF_F_USER_STACK** is also specified.
1972 * **bpf_get_stack**\ () can collect up to
1973 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1974 * to sufficient large buffer size. Note that
1975 * this limit can be controlled with the **sysctl** program, and
1976 * that it should be manually increased in order to profile long
1977 * user stacks (such as stacks for Java programs). To do so, use:
1981 * # sysctl kernel.perf_event_max_stack=<new value>
1983 * A non-negative value equal to or less than *size* on success,
1984 * or a negative error in case of failure.
1986 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
1988 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1989 * it provides an easy way to load *len* bytes from *offset*
1990 * from the packet associated to *skb*, into the buffer pointed
1991 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1992 * a fifth argument *start_header* exists in order to select a
1993 * base offset to start from. *start_header* can be one of:
1995 * **BPF_HDR_START_MAC**
1996 * Base offset to load data from is *skb*'s mac header.
1997 * **BPF_HDR_START_NET**
1998 * Base offset to load data from is *skb*'s network header.
2000 * In general, "direct packet access" is the preferred method to
2001 * access packet data, however, this helper is in particular useful
2002 * in socket filters where *skb*\ **->data** does not always point
2003 * to the start of the mac header and where "direct packet access"
2006 * 0 on success, or a negative error in case of failure.
2008 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2010 * Do FIB lookup in kernel tables using parameters in *params*.
2011 * If lookup is successful and result shows packet is to be
2012 * forwarded, the neighbor tables are searched for the nexthop.
2013 * If successful (ie., FIB lookup shows forwarding and nexthop
2014 * is resolved), the nexthop address is returned in ipv4_dst
2015 * or ipv6_dst based on family, smac is set to mac address of
2016 * egress device, dmac is set to nexthop mac address, rt_metric
2017 * is set to metric from route (IPv4/IPv6 only), and ifindex
2018 * is set to the device index of the nexthop from the FIB lookup.
2020 * *plen* argument is the size of the passed in struct.
2021 * *flags* argument can be a combination of one or more of the
2024 * **BPF_FIB_LOOKUP_DIRECT**
2025 * Do a direct table lookup vs full lookup using FIB
2027 * **BPF_FIB_LOOKUP_OUTPUT**
2028 * Perform lookup from an egress perspective (default is
2031 * *ctx* is either **struct xdp_md** for XDP programs or
2032 * **struct sk_buff** tc cls_act programs.
2034 * * < 0 if any input argument is invalid
2035 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2036 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2037 * packet is not forwarded or needs assist from full stack
2039 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2041 * Add an entry to, or update a sockhash *map* referencing sockets.
2042 * The *skops* is used as a new value for the entry associated to
2043 * *key*. *flags* is one of:
2046 * The entry for *key* must not exist in the map.
2048 * The entry for *key* must already exist in the map.
2050 * No condition on the existence of the entry for *key*.
2052 * If the *map* has eBPF programs (parser and verdict), those will
2053 * be inherited by the socket being added. If the socket is
2054 * already attached to eBPF programs, this results in an error.
2056 * 0 on success, or a negative error in case of failure.
2058 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2060 * This helper is used in programs implementing policies at the
2061 * socket level. If the message *msg* is allowed to pass (i.e. if
2062 * the verdict eBPF program returns **SK_PASS**), redirect it to
2063 * the socket referenced by *map* (of type
2064 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2065 * egress interfaces can be used for redirection. The
2066 * **BPF_F_INGRESS** value in *flags* is used to make the
2067 * distinction (ingress path is selected if the flag is present,
2068 * egress path otherwise). This is the only flag supported for now.
2070 * **SK_PASS** on success, or **SK_DROP** on error.
2072 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2074 * This helper is used in programs implementing policies at the
2075 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2076 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2077 * to the socket referenced by *map* (of type
2078 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2079 * egress interfaces can be used for redirection. The
2080 * **BPF_F_INGRESS** value in *flags* is used to make the
2081 * distinction (ingress path is selected if the flag is present,
2082 * egress otherwise). This is the only flag supported for now.
2084 * **SK_PASS** on success, or **SK_DROP** on error.
2086 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2088 * Encapsulate the packet associated to *skb* within a Layer 3
2089 * protocol header. This header is provided in the buffer at
2090 * address *hdr*, with *len* its size in bytes. *type* indicates
2091 * the protocol of the header and can be one of:
2093 * **BPF_LWT_ENCAP_SEG6**
2094 * IPv6 encapsulation with Segment Routing Header
2095 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2096 * the IPv6 header is computed by the kernel.
2097 * **BPF_LWT_ENCAP_SEG6_INLINE**
2098 * Only works if *skb* contains an IPv6 packet. Insert a
2099 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2101 * **BPF_LWT_ENCAP_IP**
2102 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2103 * must be IPv4 or IPv6, followed by zero or more
2104 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2105 * total bytes in all prepended headers. Please note that
2106 * if **skb_is_gso**\ (*skb*) is true, no more than two
2107 * headers can be prepended, and the inner header, if
2108 * present, should be either GRE or UDP/GUE.
2110 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2111 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2112 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2113 * **BPF_PROG_TYPE_LWT_XMIT**.
2115 * A call to this helper is susceptible to change the underlying
2116 * packet buffer. Therefore, at load time, all checks on pointers
2117 * previously done by the verifier are invalidated and must be
2118 * performed again, if the helper is used in combination with
2119 * direct packet access.
2121 * 0 on success, or a negative error in case of failure.
2123 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2125 * Store *len* bytes from address *from* into the packet
2126 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2127 * inside the outermost IPv6 Segment Routing Header can be
2128 * modified through this helper.
2130 * A call to this helper is susceptible to change the underlying
2131 * packet buffer. Therefore, at load time, all checks on pointers
2132 * previously done by the verifier are invalidated and must be
2133 * performed again, if the helper is used in combination with
2134 * direct packet access.
2136 * 0 on success, or a negative error in case of failure.
2138 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2140 * Adjust the size allocated to TLVs in the outermost IPv6
2141 * Segment Routing Header contained in the packet associated to
2142 * *skb*, at position *offset* by *delta* bytes. Only offsets
2143 * after the segments are accepted. *delta* can be as well
2144 * positive (growing) as negative (shrinking).
2146 * A call to this helper is susceptible to change the underlying
2147 * packet buffer. Therefore, at load time, all checks on pointers
2148 * previously done by the verifier are invalidated and must be
2149 * performed again, if the helper is used in combination with
2150 * direct packet access.
2152 * 0 on success, or a negative error in case of failure.
2154 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2156 * Apply an IPv6 Segment Routing action of type *action* to the
2157 * packet associated to *skb*. Each action takes a parameter
2158 * contained at address *param*, and of length *param_len* bytes.
2159 * *action* can be one of:
2161 * **SEG6_LOCAL_ACTION_END_X**
2162 * End.X action: Endpoint with Layer-3 cross-connect.
2163 * Type of *param*: **struct in6_addr**.
2164 * **SEG6_LOCAL_ACTION_END_T**
2165 * End.T action: Endpoint with specific IPv6 table lookup.
2166 * Type of *param*: **int**.
2167 * **SEG6_LOCAL_ACTION_END_B6**
2168 * End.B6 action: Endpoint bound to an SRv6 policy.
2169 * Type of *param*: **struct ipv6_sr_hdr**.
2170 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2171 * End.B6.Encap action: Endpoint bound to an SRv6
2172 * encapsulation policy.
2173 * Type of *param*: **struct ipv6_sr_hdr**.
2175 * A call to this helper is susceptible to change the underlying
2176 * packet buffer. Therefore, at load time, all checks on pointers
2177 * previously done by the verifier are invalidated and must be
2178 * performed again, if the helper is used in combination with
2179 * direct packet access.
2181 * 0 on success, or a negative error in case of failure.
2183 * int bpf_rc_repeat(void *ctx)
2185 * This helper is used in programs implementing IR decoding, to
2186 * report a successfully decoded repeat key message. This delays
2187 * the generation of a key up event for previously generated
2190 * Some IR protocols like NEC have a special IR message for
2191 * repeating last button, for when a button is held down.
2193 * The *ctx* should point to the lirc sample as passed into
2196 * This helper is only available is the kernel was compiled with
2197 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2202 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2204 * This helper is used in programs implementing IR decoding, to
2205 * report a successfully decoded key press with *scancode*,
2206 * *toggle* value in the given *protocol*. The scancode will be
2207 * translated to a keycode using the rc keymap, and reported as
2208 * an input key down event. After a period a key up event is
2209 * generated. This period can be extended by calling either
2210 * **bpf_rc_keydown**\ () again with the same values, or calling
2211 * **bpf_rc_repeat**\ ().
2213 * Some protocols include a toggle bit, in case the button was
2214 * released and pressed again between consecutive scancodes.
2216 * The *ctx* should point to the lirc sample as passed into
2219 * The *protocol* is the decoded protocol number (see
2220 * **enum rc_proto** for some predefined values).
2222 * This helper is only available is the kernel was compiled with
2223 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2228 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2230 * Return the cgroup v2 id of the socket associated with the *skb*.
2231 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2232 * helper for cgroup v1 by providing a tag resp. identifier that
2233 * can be matched on or used for map lookups e.g. to implement
2234 * policy. The cgroup v2 id of a given path in the hierarchy is
2235 * exposed in user space through the f_handle API in order to get
2236 * to the same 64-bit id.
2238 * This helper can be used on TC egress path, but not on ingress,
2239 * and is available only if the kernel was compiled with the
2240 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2242 * The id is returned or 0 in case the id could not be retrieved.
2244 * u64 bpf_get_current_cgroup_id(void)
2246 * A 64-bit integer containing the current cgroup id based
2247 * on the cgroup within which the current task is running.
2249 * void *bpf_get_local_storage(void *map, u64 flags)
2251 * Get the pointer to the local storage area.
2252 * The type and the size of the local storage is defined
2253 * by the *map* argument.
2254 * The *flags* meaning is specific for each map type,
2255 * and has to be 0 for cgroup local storage.
2257 * Depending on the BPF program type, a local storage area
2258 * can be shared between multiple instances of the BPF program,
2259 * running simultaneously.
2261 * A user should care about the synchronization by himself.
2262 * For example, by using the **BPF_STX_XADD** instruction to alter
2265 * A pointer to the local storage area.
2267 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2269 * Select a **SO_REUSEPORT** socket from a
2270 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2271 * It checks the selected socket is matching the incoming
2272 * request in the socket buffer.
2274 * 0 on success, or a negative error in case of failure.
2276 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2278 * Return id of cgroup v2 that is ancestor of cgroup associated
2279 * with the *skb* at the *ancestor_level*. The root cgroup is at
2280 * *ancestor_level* zero and each step down the hierarchy
2281 * increments the level. If *ancestor_level* == level of cgroup
2282 * associated with *skb*, then return value will be same as that
2283 * of **bpf_skb_cgroup_id**\ ().
2285 * The helper is useful to implement policies based on cgroups
2286 * that are upper in hierarchy than immediate cgroup associated
2289 * The format of returned id and helper limitations are same as in
2290 * **bpf_skb_cgroup_id**\ ().
2292 * The id is returned or 0 in case the id could not be retrieved.
2294 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2296 * Look for TCP socket matching *tuple*, optionally in a child
2297 * network namespace *netns*. The return value must be checked,
2298 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2300 * The *ctx* should point to the context of the program, such as
2301 * the skb or socket (depending on the hook in use). This is used
2302 * to determine the base network namespace for the lookup.
2304 * *tuple_size* must be one of:
2306 * **sizeof**\ (*tuple*\ **->ipv4**)
2307 * Look for an IPv4 socket.
2308 * **sizeof**\ (*tuple*\ **->ipv6**)
2309 * Look for an IPv6 socket.
2311 * If the *netns* is a negative signed 32-bit integer, then the
2312 * socket lookup table in the netns associated with the *ctx* will
2313 * will be used. For the TC hooks, this is the netns of the device
2314 * in the skb. For socket hooks, this is the netns of the socket.
2315 * If *netns* is any other signed 32-bit value greater than or
2316 * equal to zero then it specifies the ID of the netns relative to
2317 * the netns associated with the *ctx*. *netns* values beyond the
2318 * range of 32-bit integers are reserved for future use.
2320 * All values for *flags* are reserved for future usage, and must
2323 * This helper is available only if the kernel was compiled with
2324 * **CONFIG_NET** configuration option.
2326 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2327 * For sockets with reuseport option, the **struct bpf_sock**
2328 * result is from *reuse*\ **->socks**\ [] using the hash of the
2331 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2333 * Look for UDP socket matching *tuple*, optionally in a child
2334 * network namespace *netns*. The return value must be checked,
2335 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2337 * The *ctx* should point to the context of the program, such as
2338 * the skb or socket (depending on the hook in use). This is used
2339 * to determine the base network namespace for the lookup.
2341 * *tuple_size* must be one of:
2343 * **sizeof**\ (*tuple*\ **->ipv4**)
2344 * Look for an IPv4 socket.
2345 * **sizeof**\ (*tuple*\ **->ipv6**)
2346 * Look for an IPv6 socket.
2348 * If the *netns* is a negative signed 32-bit integer, then the
2349 * socket lookup table in the netns associated with the *ctx* will
2350 * will be used. For the TC hooks, this is the netns of the device
2351 * in the skb. For socket hooks, this is the netns of the socket.
2352 * If *netns* is any other signed 32-bit value greater than or
2353 * equal to zero then it specifies the ID of the netns relative to
2354 * the netns associated with the *ctx*. *netns* values beyond the
2355 * range of 32-bit integers are reserved for future use.
2357 * All values for *flags* are reserved for future usage, and must
2360 * This helper is available only if the kernel was compiled with
2361 * **CONFIG_NET** configuration option.
2363 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2364 * For sockets with reuseport option, the **struct bpf_sock**
2365 * result is from *reuse*\ **->socks**\ [] using the hash of the
2368 * int bpf_sk_release(struct bpf_sock *sock)
2370 * Release the reference held by *sock*. *sock* must be a
2371 * non-**NULL** pointer that was returned from
2372 * **bpf_sk_lookup_xxx**\ ().
2374 * 0 on success, or a negative error in case of failure.
2376 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2378 * Push an element *value* in *map*. *flags* is one of:
2381 * If the queue/stack is full, the oldest element is
2382 * removed to make room for this.
2384 * 0 on success, or a negative error in case of failure.
2386 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2388 * Pop an element from *map*.
2390 * 0 on success, or a negative error in case of failure.
2392 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2394 * Get an element from *map* without removing it.
2396 * 0 on success, or a negative error in case of failure.
2398 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2400 * For socket policies, insert *len* bytes into *msg* at offset
2403 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2404 * *msg* it may want to insert metadata or options into the *msg*.
2405 * This can later be read and used by any of the lower layer BPF
2408 * This helper may fail if under memory pressure (a malloc
2409 * fails) in these cases BPF programs will get an appropriate
2410 * error and BPF programs will need to handle them.
2412 * 0 on success, or a negative error in case of failure.
2414 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2416 * Will remove *len* bytes from a *msg* starting at byte *start*.
2417 * This may result in **ENOMEM** errors under certain situations if
2418 * an allocation and copy are required due to a full ring buffer.
2419 * However, the helper will try to avoid doing the allocation
2420 * if possible. Other errors can occur if input parameters are
2421 * invalid either due to *start* byte not being valid part of *msg*
2422 * payload and/or *pop* value being to large.
2424 * 0 on success, or a negative error in case of failure.
2426 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2428 * This helper is used in programs implementing IR decoding, to
2429 * report a successfully decoded pointer movement.
2431 * The *ctx* should point to the lirc sample as passed into
2434 * This helper is only available is the kernel was compiled with
2435 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2440 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2442 * Acquire a spinlock represented by the pointer *lock*, which is
2443 * stored as part of a value of a map. Taking the lock allows to
2444 * safely update the rest of the fields in that value. The
2445 * spinlock can (and must) later be released with a call to
2446 * **bpf_spin_unlock**\ (\ *lock*\ ).
2448 * Spinlocks in BPF programs come with a number of restrictions
2451 * * **bpf_spin_lock** objects are only allowed inside maps of
2452 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2453 * list could be extended in the future).
2454 * * BTF description of the map is mandatory.
2455 * * The BPF program can take ONE lock at a time, since taking two
2456 * or more could cause dead locks.
2457 * * Only one **struct bpf_spin_lock** is allowed per map element.
2458 * * When the lock is taken, calls (either BPF to BPF or helpers)
2460 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2461 * allowed inside a spinlock-ed region.
2462 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2463 * the lock, on all execution paths, before it returns.
2464 * * The BPF program can access **struct bpf_spin_lock** only via
2465 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2466 * helpers. Loading or storing data into the **struct
2467 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2468 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2469 * of the map value must be a struct and have **struct
2470 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2471 * Nested lock inside another struct is not allowed.
2472 * * The **struct bpf_spin_lock** *lock* field in a map value must
2473 * be aligned on a multiple of 4 bytes in that value.
2474 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2475 * the **bpf_spin_lock** field to user space.
2476 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2477 * a BPF program, do not update the **bpf_spin_lock** field.
2478 * * **bpf_spin_lock** cannot be on the stack or inside a
2479 * networking packet (it can only be inside of a map values).
2480 * * **bpf_spin_lock** is available to root only.
2481 * * Tracing programs and socket filter programs cannot use
2482 * **bpf_spin_lock**\ () due to insufficient preemption checks
2483 * (but this may change in the future).
2484 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2488 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2490 * Release the *lock* previously locked by a call to
2491 * **bpf_spin_lock**\ (\ *lock*\ ).
2495 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2497 * This helper gets a **struct bpf_sock** pointer such
2498 * that all the fields in this **bpf_sock** can be accessed.
2500 * A **struct bpf_sock** pointer on success, or **NULL** in
2503 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2505 * This helper gets a **struct bpf_tcp_sock** pointer from a
2506 * **struct bpf_sock** pointer.
2508 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2511 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
2513 * Set ECN (Explicit Congestion Notification) field of IP header
2514 * to **CE** (Congestion Encountered) if current value is **ECT**
2515 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2518 * 1 if the **CE** flag is set (either by the current helper call
2519 * or because it was already present), 0 if it is not set.
2521 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2523 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2524 * **bpf_sk_release**\ () is unnecessary and not allowed.
2526 * A **struct bpf_sock** pointer on success, or **NULL** in
2529 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2531 * Look for TCP socket matching *tuple*, optionally in a child
2532 * network namespace *netns*. The return value must be checked,
2533 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2535 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2536 * that it also returns timewait or request sockets. Use
2537 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2540 * This helper is available only if the kernel was compiled with
2541 * **CONFIG_NET** configuration option.
2543 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2544 * For sockets with reuseport option, the **struct bpf_sock**
2545 * result is from *reuse*\ **->socks**\ [] using the hash of the
2548 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2550 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2551 * the listening socket in *sk*.
2553 * *iph* points to the start of the IPv4 or IPv6 header, while
2554 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2555 * **sizeof**\ (**struct ip6hdr**).
2557 * *th* points to the start of the TCP header, while *th_len*
2558 * contains **sizeof**\ (**struct tcphdr**).
2561 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2564 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2566 * Get name of sysctl in /proc/sys/ and copy it into provided by
2567 * program buffer *buf* of size *buf_len*.
2569 * The buffer is always NUL terminated, unless it's zero-sized.
2571 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2572 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2573 * only (e.g. "tcp_mem").
2575 * Number of character copied (not including the trailing NUL).
2577 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2578 * truncated name in this case).
2580 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2582 * Get current value of sysctl as it is presented in /proc/sys
2583 * (incl. newline, etc), and copy it as a string into provided
2584 * by program buffer *buf* of size *buf_len*.
2586 * The whole value is copied, no matter what file position user
2587 * space issued e.g. sys_read at.
2589 * The buffer is always NUL terminated, unless it's zero-sized.
2591 * Number of character copied (not including the trailing NUL).
2593 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2594 * truncated name in this case).
2596 * **-EINVAL** if current value was unavailable, e.g. because
2597 * sysctl is uninitialized and read returns -EIO for it.
2599 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2601 * Get new value being written by user space to sysctl (before
2602 * the actual write happens) and copy it as a string into
2603 * provided by program buffer *buf* of size *buf_len*.
2605 * User space may write new value at file position > 0.
2607 * The buffer is always NUL terminated, unless it's zero-sized.
2609 * Number of character copied (not including the trailing NUL).
2611 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2612 * truncated name in this case).
2614 * **-EINVAL** if sysctl is being read.
2616 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2618 * Override new value being written by user space to sysctl with
2619 * value provided by program in buffer *buf* of size *buf_len*.
2621 * *buf* should contain a string in same form as provided by user
2622 * space on sysctl write.
2624 * User space may write new value at file position > 0. To override
2625 * the whole sysctl value file position should be set to zero.
2629 * **-E2BIG** if the *buf_len* is too big.
2631 * **-EINVAL** if sysctl is being read.
2633 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2635 * Convert the initial part of the string from buffer *buf* of
2636 * size *buf_len* to a long integer according to the given base
2637 * and save the result in *res*.
2639 * The string may begin with an arbitrary amount of white space
2640 * (as determined by **isspace**\ (3)) followed by a single
2641 * optional '**-**' sign.
2643 * Five least significant bits of *flags* encode base, other bits
2644 * are currently unused.
2646 * Base must be either 8, 10, 16 or 0 to detect it automatically
2647 * similar to user space **strtol**\ (3).
2649 * Number of characters consumed on success. Must be positive but
2650 * no more than *buf_len*.
2652 * **-EINVAL** if no valid digits were found or unsupported base
2655 * **-ERANGE** if resulting value was out of range.
2657 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2659 * Convert the initial part of the string from buffer *buf* of
2660 * size *buf_len* to an unsigned long integer according to the
2661 * given base and save the result in *res*.
2663 * The string may begin with an arbitrary amount of white space
2664 * (as determined by **isspace**\ (3)).
2666 * Five least significant bits of *flags* encode base, other bits
2667 * are currently unused.
2669 * Base must be either 8, 10, 16 or 0 to detect it automatically
2670 * similar to user space **strtoul**\ (3).
2672 * Number of characters consumed on success. Must be positive but
2673 * no more than *buf_len*.
2675 * **-EINVAL** if no valid digits were found or unsupported base
2678 * **-ERANGE** if resulting value was out of range.
2680 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2682 * Get a bpf-local-storage from a *sk*.
2684 * Logically, it could be thought of getting the value from
2685 * a *map* with *sk* as the **key**. From this
2686 * perspective, the usage is not much different from
2687 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2688 * helper enforces the key must be a full socket and the map must
2689 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2691 * Underneath, the value is stored locally at *sk* instead of
2692 * the *map*. The *map* is used as the bpf-local-storage
2693 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2694 * searched against all bpf-local-storages residing at *sk*.
2696 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2697 * used such that a new bpf-local-storage will be
2698 * created if one does not exist. *value* can be used
2699 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2700 * the initial value of a bpf-local-storage. If *value* is
2701 * **NULL**, the new bpf-local-storage will be zero initialized.
2703 * A bpf-local-storage pointer is returned on success.
2705 * **NULL** if not found or there was an error in adding
2706 * a new bpf-local-storage.
2708 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2710 * Delete a bpf-local-storage from a *sk*.
2714 * **-ENOENT** if the bpf-local-storage cannot be found.
2716 * int bpf_send_signal(u32 sig)
2718 * Send signal *sig* to the current task.
2720 * 0 on success or successfully queued.
2722 * **-EBUSY** if work queue under nmi is full.
2724 * **-EINVAL** if *sig* is invalid.
2726 * **-EPERM** if no permission to send the *sig*.
2728 * **-EAGAIN** if bpf program can try again.
2730 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2732 * Try to issue a SYN cookie for the packet with corresponding
2733 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2735 * *iph* points to the start of the IPv4 or IPv6 header, while
2736 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2737 * **sizeof**\ (**struct ip6hdr**).
2739 * *th* points to the start of the TCP header, while *th_len*
2740 * contains the length of the TCP header.
2743 * On success, lower 32 bits hold the generated SYN cookie in
2744 * followed by 16 bits which hold the MSS value for that cookie,
2745 * and the top 16 bits are unused.
2747 * On failure, the returned value is one of the following:
2749 * **-EINVAL** SYN cookie cannot be issued due to error
2751 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2753 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2755 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2757 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2759 * Write raw *data* blob into a special BPF perf event held by
2760 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2761 * event must have the following attributes: **PERF_SAMPLE_RAW**
2762 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2763 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2765 * The *flags* are used to indicate the index in *map* for which
2766 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2767 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2768 * to indicate that the index of the current CPU core should be
2771 * The value to write, of *size*, is passed through eBPF stack and
2772 * pointed by *data*.
2774 * *ctx* is a pointer to in-kernel struct sk_buff.
2776 * This helper is similar to **bpf_perf_event_output**\ () but
2777 * restricted to raw_tracepoint bpf programs.
2779 * 0 on success, or a negative error in case of failure.
2781 #define __BPF_FUNC_MAPPER(FN) \
2783 FN(map_lookup_elem), \
2784 FN(map_update_elem), \
2785 FN(map_delete_elem), \
2789 FN(get_prandom_u32), \
2790 FN(get_smp_processor_id), \
2791 FN(skb_store_bytes), \
2792 FN(l3_csum_replace), \
2793 FN(l4_csum_replace), \
2795 FN(clone_redirect), \
2796 FN(get_current_pid_tgid), \
2797 FN(get_current_uid_gid), \
2798 FN(get_current_comm), \
2799 FN(get_cgroup_classid), \
2800 FN(skb_vlan_push), \
2802 FN(skb_get_tunnel_key), \
2803 FN(skb_set_tunnel_key), \
2804 FN(perf_event_read), \
2806 FN(get_route_realm), \
2807 FN(perf_event_output), \
2808 FN(skb_load_bytes), \
2811 FN(skb_get_tunnel_opt), \
2812 FN(skb_set_tunnel_opt), \
2813 FN(skb_change_proto), \
2814 FN(skb_change_type), \
2815 FN(skb_under_cgroup), \
2816 FN(get_hash_recalc), \
2817 FN(get_current_task), \
2818 FN(probe_write_user), \
2819 FN(current_task_under_cgroup), \
2820 FN(skb_change_tail), \
2821 FN(skb_pull_data), \
2823 FN(set_hash_invalid), \
2824 FN(get_numa_node_id), \
2825 FN(skb_change_head), \
2826 FN(xdp_adjust_head), \
2827 FN(probe_read_str), \
2828 FN(get_socket_cookie), \
2829 FN(get_socket_uid), \
2832 FN(skb_adjust_room), \
2834 FN(sk_redirect_map), \
2835 FN(sock_map_update), \
2836 FN(xdp_adjust_meta), \
2837 FN(perf_event_read_value), \
2838 FN(perf_prog_read_value), \
2840 FN(override_return), \
2841 FN(sock_ops_cb_flags_set), \
2842 FN(msg_redirect_map), \
2843 FN(msg_apply_bytes), \
2844 FN(msg_cork_bytes), \
2845 FN(msg_pull_data), \
2847 FN(xdp_adjust_tail), \
2848 FN(skb_get_xfrm_state), \
2850 FN(skb_load_bytes_relative), \
2852 FN(sock_hash_update), \
2853 FN(msg_redirect_hash), \
2854 FN(sk_redirect_hash), \
2855 FN(lwt_push_encap), \
2856 FN(lwt_seg6_store_bytes), \
2857 FN(lwt_seg6_adjust_srh), \
2858 FN(lwt_seg6_action), \
2861 FN(skb_cgroup_id), \
2862 FN(get_current_cgroup_id), \
2863 FN(get_local_storage), \
2864 FN(sk_select_reuseport), \
2865 FN(skb_ancestor_cgroup_id), \
2866 FN(sk_lookup_tcp), \
2867 FN(sk_lookup_udp), \
2869 FN(map_push_elem), \
2871 FN(map_peek_elem), \
2872 FN(msg_push_data), \
2874 FN(rc_pointer_rel), \
2879 FN(skb_ecn_set_ce), \
2880 FN(get_listener_sock), \
2881 FN(skc_lookup_tcp), \
2882 FN(tcp_check_syncookie), \
2883 FN(sysctl_get_name), \
2884 FN(sysctl_get_current_value), \
2885 FN(sysctl_get_new_value), \
2886 FN(sysctl_set_new_value), \
2889 FN(sk_storage_get), \
2890 FN(sk_storage_delete), \
2892 FN(tcp_gen_syncookie), \
2895 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2896 * function eBPF program intends to call
2898 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2900 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2903 #undef __BPF_ENUM_FN
2905 /* All flags used by eBPF helper functions, placed here. */
2907 /* BPF_FUNC_skb_store_bytes flags. */
2908 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2909 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2911 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2912 * First 4 bits are for passing the header field size.
2914 #define BPF_F_HDR_FIELD_MASK 0xfULL
2916 /* BPF_FUNC_l4_csum_replace flags. */
2917 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2918 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2919 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2921 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2922 #define BPF_F_INGRESS (1ULL << 0)
2924 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2925 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2927 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2928 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2929 #define BPF_F_USER_STACK (1ULL << 8)
2930 /* flags used by BPF_FUNC_get_stackid only. */
2931 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2932 #define BPF_F_REUSE_STACKID (1ULL << 10)
2933 /* flags used by BPF_FUNC_get_stack only. */
2934 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2936 /* BPF_FUNC_skb_set_tunnel_key flags. */
2937 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2938 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2939 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2941 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2942 * BPF_FUNC_perf_event_read_value flags.
2944 #define BPF_F_INDEX_MASK 0xffffffffULL
2945 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2946 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2947 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2949 /* Current network namespace */
2950 #define BPF_F_CURRENT_NETNS (-1L)
2952 /* BPF_FUNC_skb_adjust_room flags. */
2953 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
2955 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
2956 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
2958 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
2959 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
2960 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
2961 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
2962 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
2963 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
2964 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
2966 /* BPF_FUNC_sysctl_get_name flags. */
2967 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
2969 /* BPF_FUNC_sk_storage_get flags */
2970 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0)
2972 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2973 enum bpf_adj_room_mode {
2978 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2979 enum bpf_hdr_start_off {
2984 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2985 enum bpf_lwt_encap_mode {
2987 BPF_LWT_ENCAP_SEG6_INLINE,
2991 #define __bpf_md_ptr(type, name) \
2995 } __attribute__((aligned(8)))
2997 /* user accessible mirror of in-kernel sk_buff.
2998 * new fields can only be added to the end of this structure
3004 __u32 queue_mapping;
3010 __u32 ingress_ifindex;
3020 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
3022 __u32 remote_ip4; /* Stored in network byte order */
3023 __u32 local_ip4; /* Stored in network byte order */
3024 __u32 remote_ip6[4]; /* Stored in network byte order */
3025 __u32 local_ip6[4]; /* Stored in network byte order */
3026 __u32 remote_port; /* Stored in network byte order */
3027 __u32 local_port; /* stored in host byte order */
3031 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
3035 __bpf_md_ptr(struct bpf_sock *, sk);
3038 struct bpf_tunnel_key {
3042 __u32 remote_ipv6[4];
3046 __u16 tunnel_ext; /* Padding, future use. */
3050 /* user accessible mirror of in-kernel xfrm_state.
3051 * new fields can only be added to the end of this structure
3053 struct bpf_xfrm_state {
3055 __u32 spi; /* Stored in network byte order */
3057 __u16 ext; /* Padding, future use. */
3059 __u32 remote_ipv4; /* Stored in network byte order */
3060 __u32 remote_ipv6[4]; /* Stored in network byte order */
3064 /* Generic BPF return codes which all BPF program types may support.
3065 * The values are binary compatible with their TC_ACT_* counter-part to
3066 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3069 * XDP is handled seprately, see XDP_*.
3077 /* >127 are reserved for prog type specific return codes.
3079 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3080 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3081 * changed and should be routed based on its new L3 header.
3082 * (This is an L3 redirect, as opposed to L2 redirect
3083 * represented by BPF_REDIRECT above).
3085 BPF_LWT_REROUTE = 128,
3095 /* IP address also allows 1 and 2 bytes access */
3098 __u32 src_port; /* host byte order */
3099 __u32 dst_port; /* network byte order */
3105 struct bpf_tcp_sock {
3106 __u32 snd_cwnd; /* Sending congestion window */
3107 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3109 __u32 snd_ssthresh; /* Slow start size threshold */
3110 __u32 rcv_nxt; /* What we want to receive next */
3111 __u32 snd_nxt; /* Next sequence we send */
3112 __u32 snd_una; /* First byte we want an ack for */
3113 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3114 __u32 ecn_flags; /* ECN status bits. */
3115 __u32 rate_delivered; /* saved rate sample: packets delivered */
3116 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3117 __u32 packets_out; /* Packets which are "in flight" */
3118 __u32 retrans_out; /* Retransmitted packets out */
3119 __u32 total_retrans; /* Total retransmits for entire connection */
3120 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3121 * total number of segments in.
3123 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3124 * total number of data segments in.
3126 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3127 * The total number of segments sent.
3129 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3130 * total number of data segments sent.
3132 __u32 lost_out; /* Lost packets */
3133 __u32 sacked_out; /* SACK'd packets */
3134 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3135 * sum(delta(rcv_nxt)), or how many bytes
3138 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3139 * sum(delta(snd_una)), or how many bytes
3142 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3143 * total number of DSACK blocks received
3145 __u32 delivered; /* Total data packets delivered incl. rexmits */
3146 __u32 delivered_ce; /* Like the above but only ECE marked packets */
3147 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
3150 struct bpf_sock_tuple {
3167 struct bpf_xdp_sock {
3171 #define XDP_PACKET_HEADROOM 256
3173 /* User return codes for XDP prog type.
3174 * A valid XDP program must return one of these defined values. All other
3175 * return codes are reserved for future use. Unknown return codes will
3176 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3186 /* user accessible metadata for XDP packet hook
3187 * new fields must be added to the end of this structure
3193 /* Below access go through struct xdp_rxq_info */
3194 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3195 __u32 rx_queue_index; /* rxq->queue_index */
3203 /* user accessible metadata for SK_MSG packet hook, new fields must
3204 * be added to the end of this structure
3207 __bpf_md_ptr(void *, data);
3208 __bpf_md_ptr(void *, data_end);
3211 __u32 remote_ip4; /* Stored in network byte order */
3212 __u32 local_ip4; /* Stored in network byte order */
3213 __u32 remote_ip6[4]; /* Stored in network byte order */
3214 __u32 local_ip6[4]; /* Stored in network byte order */
3215 __u32 remote_port; /* Stored in network byte order */
3216 __u32 local_port; /* stored in host byte order */
3217 __u32 size; /* Total size of sk_msg */
3220 struct sk_reuseport_md {
3222 * Start of directly accessible data. It begins from
3223 * the tcp/udp header.
3225 __bpf_md_ptr(void *, data);
3226 /* End of directly accessible data */
3227 __bpf_md_ptr(void *, data_end);
3229 * Total length of packet (starting from the tcp/udp header).
3230 * Note that the directly accessible bytes (data_end - data)
3231 * could be less than this "len". Those bytes could be
3232 * indirectly read by a helper "bpf_skb_load_bytes()".
3236 * Eth protocol in the mac header (network byte order). e.g.
3237 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3240 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3241 __u32 bind_inany; /* Is sock bound to an INANY address? */
3242 __u32 hash; /* A hash of the packet 4 tuples */
3245 #define BPF_TAG_SIZE 8
3247 struct bpf_prog_info {
3250 __u8 tag[BPF_TAG_SIZE];
3251 __u32 jited_prog_len;
3252 __u32 xlated_prog_len;
3253 __aligned_u64 jited_prog_insns;
3254 __aligned_u64 xlated_prog_insns;
3255 __u64 load_time; /* ns since boottime */
3256 __u32 created_by_uid;
3258 __aligned_u64 map_ids;
3259 char name[BPF_OBJ_NAME_LEN];
3261 __u32 gpl_compatible:1;
3262 __u32 :31; /* alignment pad */
3265 __u32 nr_jited_ksyms;
3266 __u32 nr_jited_func_lens;
3267 __aligned_u64 jited_ksyms;
3268 __aligned_u64 jited_func_lens;
3270 __u32 func_info_rec_size;
3271 __aligned_u64 func_info;
3274 __aligned_u64 line_info;
3275 __aligned_u64 jited_line_info;
3276 __u32 nr_jited_line_info;
3277 __u32 line_info_rec_size;
3278 __u32 jited_line_info_rec_size;
3280 __aligned_u64 prog_tags;
3283 } __attribute__((aligned(8)));
3285 struct bpf_map_info {
3292 char name[BPF_OBJ_NAME_LEN];
3298 __u32 btf_key_type_id;
3299 __u32 btf_value_type_id;
3300 } __attribute__((aligned(8)));
3302 struct bpf_btf_info {
3306 } __attribute__((aligned(8)));
3308 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3309 * by user and intended to be used by socket (e.g. to bind to, depends on
3310 * attach attach type).
3312 struct bpf_sock_addr {
3313 __u32 user_family; /* Allows 4-byte read, but no write. */
3314 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3315 * Stored in network byte order.
3317 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3318 * Stored in network byte order.
3320 __u32 user_port; /* Allows 4-byte read and write.
3321 * Stored in network byte order
3323 __u32 family; /* Allows 4-byte read, but no write */
3324 __u32 type; /* Allows 4-byte read, but no write */
3325 __u32 protocol; /* Allows 4-byte read, but no write */
3326 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3327 * Stored in network byte order.
3329 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3330 * Stored in network byte order.
3332 __bpf_md_ptr(struct bpf_sock *, sk);
3335 /* User bpf_sock_ops struct to access socket values and specify request ops
3336 * and their replies.
3337 * Some of this fields are in network (bigendian) byte order and may need
3338 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3339 * New fields can only be added at the end of this structure
3341 struct bpf_sock_ops {
3344 __u32 args[4]; /* Optionally passed to bpf program */
3345 __u32 reply; /* Returned by bpf program */
3346 __u32 replylong[4]; /* Optionally returned by bpf prog */
3349 __u32 remote_ip4; /* Stored in network byte order */
3350 __u32 local_ip4; /* Stored in network byte order */
3351 __u32 remote_ip6[4]; /* Stored in network byte order */
3352 __u32 local_ip6[4]; /* Stored in network byte order */
3353 __u32 remote_port; /* Stored in network byte order */
3354 __u32 local_port; /* stored in host byte order */
3355 __u32 is_fullsock; /* Some TCP fields are only valid if
3356 * there is a full socket. If not, the
3357 * fields read as zero.
3360 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
3361 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3370 __u32 rate_delivered;
3371 __u32 rate_interval_us;
3374 __u32 total_retrans;
3378 __u32 data_segs_out;
3382 __u64 bytes_received;
3384 __bpf_md_ptr(struct bpf_sock *, sk);
3387 /* Definitions for bpf_sock_ops_cb_flags */
3388 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
3389 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
3390 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3391 #define BPF_SOCK_OPS_RTT_CB_FLAG (1<<3)
3392 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0xF /* Mask of all currently
3393 * supported cb flags
3396 /* List of known BPF sock_ops operators.
3397 * New entries can only be added at the end
3401 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3402 * -1 if default value should be used
3404 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3405 * window (in packets) or -1 if default
3406 * value should be used
3408 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3409 * active connection is initialized
3411 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3412 * active connection is
3415 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3416 * passive connection is
3419 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3422 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3423 * based on the path and may be
3424 * dependent on the congestion control
3425 * algorithm. In general it indicates
3426 * a congestion threshold. RTTs above
3427 * this indicate congestion
3429 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3430 * Arg1: value of icsk_retransmits
3431 * Arg2: value of icsk_rto
3432 * Arg3: whether RTO has expired
3434 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3435 * Arg1: sequence number of 1st byte
3437 * Arg3: return value of
3438 * tcp_transmit_skb (0 => success)
3440 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3444 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3445 * socket transition to LISTEN state.
3447 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3451 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3452 * changes between the TCP and BPF versions. Ideally this should never happen.
3453 * If it does, we need to add code to convert them before calling
3454 * the BPF sock_ops function.
3457 BPF_TCP_ESTABLISHED = 1,
3467 BPF_TCP_CLOSING, /* Now a valid state */
3468 BPF_TCP_NEW_SYN_RECV,
3470 BPF_TCP_MAX_STATES /* Leave at the end! */
3473 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3474 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3476 struct bpf_perf_event_value {
3482 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3483 #define BPF_DEVCG_ACC_READ (1ULL << 1)
3484 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3486 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3487 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3489 struct bpf_cgroup_dev_ctx {
3490 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3496 struct bpf_raw_tracepoint_args {
3500 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3501 * OUTPUT: Do lookup from egress perspective; default is ingress
3503 #define BPF_FIB_LOOKUP_DIRECT (1U << 0)
3504 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
3507 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3508 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3509 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3510 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3511 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3512 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3513 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3514 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3515 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3518 struct bpf_fib_lookup {
3519 /* input: network family for lookup (AF_INET, AF_INET6)
3520 * output: network family of egress nexthop
3524 /* set if lookup is to consider L4 data - e.g., FIB rules */
3529 /* total length of packet from network header - used for MTU check */
3532 /* input: L3 device index for lookup
3533 * output: device index from FIB lookup
3538 /* inputs to lookup */
3539 __u8 tos; /* AF_INET */
3540 __be32 flowinfo; /* AF_INET6, flow_label + priority */
3542 /* output: metric of fib result (IPv4/IPv6 only) */
3548 __u32 ipv6_src[4]; /* in6_addr; network order */
3551 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3552 * network header. output: bpf_fib_lookup sets to gateway address
3553 * if FIB lookup returns gateway route
3557 __u32 ipv6_dst[4]; /* in6_addr; network order */
3561 __be16 h_vlan_proto;
3563 __u8 smac[6]; /* ETH_ALEN */
3564 __u8 dmac[6]; /* ETH_ALEN */
3567 enum bpf_task_fd_type {
3568 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3569 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3570 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3571 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3572 BPF_FD_TYPE_UPROBE, /* filename + offset */
3573 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3576 #define BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG (1U << 0)
3577 #define BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL (1U << 1)
3578 #define BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP (1U << 2)
3580 struct bpf_flow_keys {
3583 __u16 addr_proto; /* ETH_P_* of valid addrs */
3597 __u32 ipv6_src[4]; /* in6_addr; network order */
3598 __u32 ipv6_dst[4]; /* in6_addr; network order */
3605 struct bpf_func_info {
3610 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3611 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3613 struct bpf_line_info {
3615 __u32 file_name_off;
3620 struct bpf_spin_lock {
3625 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3626 * Allows 1,2,4-byte read, but no write.
3628 __u32 file_pos; /* Sysctl file position to read from, write to.
3629 * Allows 1,2,4-byte read an 4-byte write.
3633 struct bpf_sockopt {
3634 __bpf_md_ptr(struct bpf_sock *, sk);
3635 __bpf_md_ptr(void *, optval);
3636 __bpf_md_ptr(void *, optval_end);
3644 #endif /* _UAPI__LINUX_BPF_H__ */