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,
178 enum bpf_attach_type {
179 BPF_CGROUP_INET_INGRESS,
180 BPF_CGROUP_INET_EGRESS,
181 BPF_CGROUP_INET_SOCK_CREATE,
183 BPF_SK_SKB_STREAM_PARSER,
184 BPF_SK_SKB_STREAM_VERDICT,
187 BPF_CGROUP_INET4_BIND,
188 BPF_CGROUP_INET6_BIND,
189 BPF_CGROUP_INET4_CONNECT,
190 BPF_CGROUP_INET6_CONNECT,
191 BPF_CGROUP_INET4_POST_BIND,
192 BPF_CGROUP_INET6_POST_BIND,
193 BPF_CGROUP_UDP4_SENDMSG,
194 BPF_CGROUP_UDP6_SENDMSG,
198 BPF_CGROUP_UDP4_RECVMSG,
199 BPF_CGROUP_UDP6_RECVMSG,
200 BPF_CGROUP_GETSOCKOPT,
201 BPF_CGROUP_SETSOCKOPT,
202 __MAX_BPF_ATTACH_TYPE
205 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
207 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
209 * NONE(default): No further bpf programs allowed in the subtree.
211 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
212 * the program in this cgroup yields to sub-cgroup program.
214 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
215 * that cgroup program gets run in addition to the program in this cgroup.
217 * Only one program is allowed to be attached to a cgroup with
218 * NONE or BPF_F_ALLOW_OVERRIDE flag.
219 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
220 * release old program and attach the new one. Attach flags has to match.
222 * Multiple programs are allowed to be attached to a cgroup with
223 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
224 * (those that were attached first, run first)
225 * The programs of sub-cgroup are executed first, then programs of
226 * this cgroup and then programs of parent cgroup.
227 * When children program makes decision (like picking TCP CA or sock bind)
228 * parent program has a chance to override it.
230 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
231 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
233 * cgrp1 (MULTI progs A, B) ->
234 * cgrp2 (OVERRIDE prog C) ->
235 * cgrp3 (MULTI prog D) ->
236 * cgrp4 (OVERRIDE prog E) ->
237 * cgrp5 (NONE prog F)
238 * the event in cgrp5 triggers execution of F,D,A,B in that order.
239 * if prog F is detached, the execution is E,D,A,B
240 * if prog F and D are detached, the execution is E,A,B
241 * if prog F, E and D are detached, the execution is C,A,B
243 * All eligible programs are executed regardless of return code from
246 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
247 #define BPF_F_ALLOW_MULTI (1U << 1)
249 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
250 * verifier will perform strict alignment checking as if the kernel
251 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
252 * and NET_IP_ALIGN defined to 2.
254 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
256 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
257 * verifier will allow any alignment whatsoever. On platforms
258 * with strict alignment requirements for loads ands stores (such
259 * as sparc and mips) the verifier validates that all loads and
260 * stores provably follow this requirement. This flag turns that
261 * checking and enforcement off.
263 * It is mostly used for testing when we want to validate the
264 * context and memory access aspects of the verifier, but because
265 * of an unaligned access the alignment check would trigger before
266 * the one we are interested in.
268 #define BPF_F_ANY_ALIGNMENT (1U << 1)
270 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
271 * Verifier does sub-register def/use analysis and identifies instructions whose
272 * def only matters for low 32-bit, high 32-bit is never referenced later
273 * through implicit zero extension. Therefore verifier notifies JIT back-ends
274 * that it is safe to ignore clearing high 32-bit for these instructions. This
275 * saves some back-ends a lot of code-gen. However such optimization is not
276 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
277 * hence hasn't used verifier's analysis result. But, we really want to have a
278 * way to be able to verify the correctness of the described optimization on
279 * x86_64 on which testsuites are frequently exercised.
281 * So, this flag is introduced. Once it is set, verifier will randomize high
282 * 32-bit for those instructions who has been identified as safe to ignore them.
283 * Then, if verifier is not doing correct analysis, such randomization will
284 * regress tests to expose bugs.
286 #define BPF_F_TEST_RND_HI32 (1U << 2)
288 /* The verifier internal test flag. Behavior is undefined */
289 #define BPF_F_TEST_STATE_FREQ (1U << 3)
291 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
294 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
295 * insn[0].imm: map fd map fd
296 * insn[1].imm: 0 offset into value
299 * ldimm64 rewrite: address of map address of map[0]+offset
300 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
302 #define BPF_PSEUDO_MAP_FD 1
303 #define BPF_PSEUDO_MAP_VALUE 2
305 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
306 * offset to another bpf function
308 #define BPF_PSEUDO_CALL 1
310 /* flags for BPF_MAP_UPDATE_ELEM command */
311 #define BPF_ANY 0 /* create new element or update existing */
312 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
313 #define BPF_EXIST 2 /* update existing element */
314 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
316 /* flags for BPF_MAP_CREATE command */
317 #define BPF_F_NO_PREALLOC (1U << 0)
318 /* Instead of having one common LRU list in the
319 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
320 * which can scale and perform better.
321 * Note, the LRU nodes (including free nodes) cannot be moved
322 * across different LRU lists.
324 #define BPF_F_NO_COMMON_LRU (1U << 1)
325 /* Specify numa node during map creation */
326 #define BPF_F_NUMA_NODE (1U << 2)
328 #define BPF_OBJ_NAME_LEN 16U
330 /* Flags for accessing BPF object from syscall side. */
331 #define BPF_F_RDONLY (1U << 3)
332 #define BPF_F_WRONLY (1U << 4)
334 /* Flag for stack_map, store build_id+offset instead of pointer */
335 #define BPF_F_STACK_BUILD_ID (1U << 5)
337 /* Zero-initialize hash function seed. This should only be used for testing. */
338 #define BPF_F_ZERO_SEED (1U << 6)
340 /* Flags for accessing BPF object from program side. */
341 #define BPF_F_RDONLY_PROG (1U << 7)
342 #define BPF_F_WRONLY_PROG (1U << 8)
344 /* Clone map from listener for newly accepted socket */
345 #define BPF_F_CLONE (1U << 9)
347 /* flags for BPF_PROG_QUERY */
348 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
350 enum bpf_stack_build_id_status {
351 /* user space need an empty entry to identify end of a trace */
352 BPF_STACK_BUILD_ID_EMPTY = 0,
353 /* with valid build_id and offset */
354 BPF_STACK_BUILD_ID_VALID = 1,
355 /* couldn't get build_id, fallback to ip */
356 BPF_STACK_BUILD_ID_IP = 2,
359 #define BPF_BUILD_ID_SIZE 20
360 struct bpf_stack_build_id {
362 unsigned char build_id[BPF_BUILD_ID_SIZE];
370 struct { /* anonymous struct used by BPF_MAP_CREATE command */
371 __u32 map_type; /* one of enum bpf_map_type */
372 __u32 key_size; /* size of key in bytes */
373 __u32 value_size; /* size of value in bytes */
374 __u32 max_entries; /* max number of entries in a map */
375 __u32 map_flags; /* BPF_MAP_CREATE related
376 * flags defined above.
378 __u32 inner_map_fd; /* fd pointing to the inner map */
379 __u32 numa_node; /* numa node (effective only if
380 * BPF_F_NUMA_NODE is set).
382 char map_name[BPF_OBJ_NAME_LEN];
383 __u32 map_ifindex; /* ifindex of netdev to create on */
384 __u32 btf_fd; /* fd pointing to a BTF type data */
385 __u32 btf_key_type_id; /* BTF type_id of the key */
386 __u32 btf_value_type_id; /* BTF type_id of the value */
389 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
394 __aligned_u64 next_key;
399 struct { /* anonymous struct used by BPF_PROG_LOAD command */
400 __u32 prog_type; /* one of enum bpf_prog_type */
403 __aligned_u64 license;
404 __u32 log_level; /* verbosity level of verifier */
405 __u32 log_size; /* size of user buffer */
406 __aligned_u64 log_buf; /* user supplied buffer */
407 __u32 kern_version; /* not used */
409 char prog_name[BPF_OBJ_NAME_LEN];
410 __u32 prog_ifindex; /* ifindex of netdev to prep for */
411 /* For some prog types expected attach type must be known at
412 * load time to verify attach type specific parts of prog
413 * (context accesses, allowed helpers, etc).
415 __u32 expected_attach_type;
416 __u32 prog_btf_fd; /* fd pointing to BTF type data */
417 __u32 func_info_rec_size; /* userspace bpf_func_info size */
418 __aligned_u64 func_info; /* func info */
419 __u32 func_info_cnt; /* number of bpf_func_info records */
420 __u32 line_info_rec_size; /* userspace bpf_line_info size */
421 __aligned_u64 line_info; /* line info */
422 __u32 line_info_cnt; /* number of bpf_line_info records */
425 struct { /* anonymous struct used by BPF_OBJ_* commands */
426 __aligned_u64 pathname;
431 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
432 __u32 target_fd; /* container object to attach to */
433 __u32 attach_bpf_fd; /* eBPF program to attach */
438 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
441 __u32 data_size_in; /* input: len of data_in */
442 __u32 data_size_out; /* input/output: len of data_out
443 * returns ENOSPC if data_out
446 __aligned_u64 data_in;
447 __aligned_u64 data_out;
450 __u32 ctx_size_in; /* input: len of ctx_in */
451 __u32 ctx_size_out; /* input/output: len of ctx_out
452 * returns ENOSPC if ctx_out
455 __aligned_u64 ctx_in;
456 __aligned_u64 ctx_out;
459 struct { /* anonymous struct used by BPF_*_GET_*_ID */
470 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
476 struct { /* anonymous struct used by BPF_PROG_QUERY command */
477 __u32 target_fd; /* container object to query */
481 __aligned_u64 prog_ids;
490 struct { /* anonymous struct for BPF_BTF_LOAD */
492 __aligned_u64 btf_log_buf;
499 __u32 pid; /* input: pid */
500 __u32 fd; /* input: fd */
501 __u32 flags; /* input: flags */
502 __u32 buf_len; /* input/output: buf len */
503 __aligned_u64 buf; /* input/output:
504 * tp_name for tracepoint
506 * filename for uprobe
508 __u32 prog_id; /* output: prod_id */
509 __u32 fd_type; /* output: BPF_FD_TYPE_* */
510 __u64 probe_offset; /* output: probe_offset */
511 __u64 probe_addr; /* output: probe_addr */
513 } __attribute__((aligned(8)));
515 /* The description below is an attempt at providing documentation to eBPF
516 * developers about the multiple available eBPF helper functions. It can be
517 * parsed and used to produce a manual page. The workflow is the following,
518 * and requires the rst2man utility:
520 * $ ./scripts/bpf_helpers_doc.py \
521 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
522 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
523 * $ man /tmp/bpf-helpers.7
525 * Note that in order to produce this external documentation, some RST
526 * formatting is used in the descriptions to get "bold" and "italics" in
527 * manual pages. Also note that the few trailing white spaces are
528 * intentional, removing them would break paragraphs for rst2man.
530 * Start of BPF helper function descriptions:
532 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
534 * Perform a lookup in *map* for an entry associated to *key*.
536 * Map value associated to *key*, or **NULL** if no entry was
539 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
541 * Add or update the value of the entry associated to *key* in
542 * *map* with *value*. *flags* is one of:
545 * The entry for *key* must not exist in the map.
547 * The entry for *key* must already exist in the map.
549 * No condition on the existence of the entry for *key*.
551 * Flag value **BPF_NOEXIST** cannot be used for maps of types
552 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
553 * elements always exist), the helper would return an error.
555 * 0 on success, or a negative error in case of failure.
557 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
559 * Delete entry with *key* from *map*.
561 * 0 on success, or a negative error in case of failure.
563 * int bpf_probe_read(void *dst, u32 size, const void *src)
565 * For tracing programs, safely attempt to read *size* bytes from
566 * address *src* and store the data in *dst*.
568 * 0 on success, or a negative error in case of failure.
570 * u64 bpf_ktime_get_ns(void)
572 * Return the time elapsed since system boot, in nanoseconds.
576 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
578 * This helper is a "printk()-like" facility for debugging. It
579 * prints a message defined by format *fmt* (of size *fmt_size*)
580 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
581 * available. It can take up to three additional **u64**
582 * arguments (as an eBPF helpers, the total number of arguments is
585 * Each time the helper is called, it appends a line to the trace.
586 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
587 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
588 * The format of the trace is customizable, and the exact output
589 * one will get depends on the options set in
590 * *\/sys/kernel/debug/tracing/trace_options* (see also the
591 * *README* file under the same directory). However, it usually
592 * defaults to something like:
596 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
600 * * ``telnet`` is the name of the current task.
601 * * ``470`` is the PID of the current task.
602 * * ``001`` is the CPU number on which the task is
604 * * In ``.N..``, each character refers to a set of
605 * options (whether irqs are enabled, scheduling
606 * options, whether hard/softirqs are running, level of
607 * preempt_disabled respectively). **N** means that
608 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
610 * * ``419421.045894`` is a timestamp.
611 * * ``0x00000001`` is a fake value used by BPF for the
612 * instruction pointer register.
613 * * ``<formatted msg>`` is the message formatted with
616 * The conversion specifiers supported by *fmt* are similar, but
617 * more limited than for printk(). They are **%d**, **%i**,
618 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
619 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
620 * of field, padding with zeroes, etc.) is available, and the
621 * helper will return **-EINVAL** (but print nothing) if it
622 * encounters an unknown specifier.
624 * Also, note that **bpf_trace_printk**\ () is slow, and should
625 * only be used for debugging purposes. For this reason, a notice
626 * bloc (spanning several lines) is printed to kernel logs and
627 * states that the helper should not be used "for production use"
628 * the first time this helper is used (or more precisely, when
629 * **trace_printk**\ () buffers are allocated). For passing values
630 * to user space, perf events should be preferred.
632 * The number of bytes written to the buffer, or a negative error
633 * in case of failure.
635 * u32 bpf_get_prandom_u32(void)
637 * Get a pseudo-random number.
639 * From a security point of view, this helper uses its own
640 * pseudo-random internal state, and cannot be used to infer the
641 * seed of other random functions in the kernel. However, it is
642 * essential to note that the generator used by the helper is not
643 * cryptographically secure.
645 * A random 32-bit unsigned value.
647 * u32 bpf_get_smp_processor_id(void)
649 * Get the SMP (symmetric multiprocessing) processor id. Note that
650 * all programs run with preemption disabled, which means that the
651 * SMP processor id is stable during all the execution of the
654 * The SMP id of the processor running the program.
656 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
658 * Store *len* bytes from address *from* into the packet
659 * associated to *skb*, at *offset*. *flags* are a combination of
660 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
661 * checksum for the packet after storing the bytes) and
662 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
663 * **->swhash** and *skb*\ **->l4hash** to 0).
665 * A call to this helper is susceptible to change the underlying
666 * packet buffer. Therefore, at load time, all checks on pointers
667 * previously done by the verifier are invalidated and must be
668 * performed again, if the helper is used in combination with
669 * direct packet access.
671 * 0 on success, or a negative error in case of failure.
673 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
675 * Recompute the layer 3 (e.g. IP) checksum for the packet
676 * associated to *skb*. Computation is incremental, so the helper
677 * must know the former value of the header field that was
678 * modified (*from*), the new value of this field (*to*), and the
679 * number of bytes (2 or 4) for this field, stored in *size*.
680 * Alternatively, it is possible to store the difference between
681 * the previous and the new values of the header field in *to*, by
682 * setting *from* and *size* to 0. For both methods, *offset*
683 * indicates the location of the IP checksum within the packet.
685 * This helper works in combination with **bpf_csum_diff**\ (),
686 * which does not update the checksum in-place, but offers more
687 * flexibility and can handle sizes larger than 2 or 4 for the
688 * checksum to update.
690 * A call to this helper is susceptible to change the underlying
691 * packet buffer. Therefore, at load time, all checks on pointers
692 * previously done by the verifier are invalidated and must be
693 * performed again, if the helper is used in combination with
694 * direct packet access.
696 * 0 on success, or a negative error in case of failure.
698 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
700 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
701 * packet associated to *skb*. Computation is incremental, so the
702 * helper must know the former value of the header field that was
703 * modified (*from*), the new value of this field (*to*), and the
704 * number of bytes (2 or 4) for this field, stored on the lowest
705 * four bits of *flags*. Alternatively, it is possible to store
706 * the difference between the previous and the new values of the
707 * header field in *to*, by setting *from* and the four lowest
708 * bits of *flags* to 0. For both methods, *offset* indicates the
709 * location of the IP checksum within the packet. In addition to
710 * the size of the field, *flags* can be added (bitwise OR) actual
711 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
712 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
713 * for updates resulting in a null checksum the value is set to
714 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
715 * the checksum is to be computed against a pseudo-header.
717 * This helper works in combination with **bpf_csum_diff**\ (),
718 * which does not update the checksum in-place, but offers more
719 * flexibility and can handle sizes larger than 2 or 4 for the
720 * checksum to update.
722 * A call to this helper is susceptible to change the underlying
723 * packet buffer. Therefore, at load time, all checks on pointers
724 * previously done by the verifier are invalidated and must be
725 * performed again, if the helper is used in combination with
726 * direct packet access.
728 * 0 on success, or a negative error in case of failure.
730 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
732 * This special helper is used to trigger a "tail call", or in
733 * other words, to jump into another eBPF program. The same stack
734 * frame is used (but values on stack and in registers for the
735 * caller are not accessible to the callee). This mechanism allows
736 * for program chaining, either for raising the maximum number of
737 * available eBPF instructions, or to execute given programs in
738 * conditional blocks. For security reasons, there is an upper
739 * limit to the number of successive tail calls that can be
742 * Upon call of this helper, the program attempts to jump into a
743 * program referenced at index *index* in *prog_array_map*, a
744 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
745 * *ctx*, a pointer to the context.
747 * If the call succeeds, the kernel immediately runs the first
748 * instruction of the new program. This is not a function call,
749 * and it never returns to the previous program. If the call
750 * fails, then the helper has no effect, and the caller continues
751 * to run its subsequent instructions. A call can fail if the
752 * destination program for the jump does not exist (i.e. *index*
753 * is superior to the number of entries in *prog_array_map*), or
754 * if the maximum number of tail calls has been reached for this
755 * chain of programs. This limit is defined in the kernel by the
756 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
757 * which is currently set to 32.
759 * 0 on success, or a negative error in case of failure.
761 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
763 * Clone and redirect the packet associated to *skb* to another
764 * net device of index *ifindex*. Both ingress and egress
765 * interfaces can be used for redirection. The **BPF_F_INGRESS**
766 * value in *flags* is used to make the distinction (ingress path
767 * is selected if the flag is present, egress path otherwise).
768 * This is the only flag supported for now.
770 * In comparison with **bpf_redirect**\ () helper,
771 * **bpf_clone_redirect**\ () has the associated cost of
772 * duplicating the packet buffer, but this can be executed out of
773 * the eBPF program. Conversely, **bpf_redirect**\ () is more
774 * efficient, but it is handled through an action code where the
775 * redirection happens only after the eBPF program has returned.
777 * A call to this helper is susceptible to change the underlying
778 * packet buffer. Therefore, at load time, all checks on pointers
779 * previously done by the verifier are invalidated and must be
780 * performed again, if the helper is used in combination with
781 * direct packet access.
783 * 0 on success, or a negative error in case of failure.
785 * u64 bpf_get_current_pid_tgid(void)
787 * A 64-bit integer containing the current tgid and pid, and
789 * *current_task*\ **->tgid << 32 \|**
790 * *current_task*\ **->pid**.
792 * u64 bpf_get_current_uid_gid(void)
794 * A 64-bit integer containing the current GID and UID, and
795 * created as such: *current_gid* **<< 32 \|** *current_uid*.
797 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
799 * Copy the **comm** attribute of the current task into *buf* of
800 * *size_of_buf*. The **comm** attribute contains the name of
801 * the executable (excluding the path) for the current task. The
802 * *size_of_buf* must be strictly positive. On success, the
803 * helper makes sure that the *buf* is NUL-terminated. On failure,
804 * it is filled with zeroes.
806 * 0 on success, or a negative error in case of failure.
808 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
810 * Retrieve the classid for the current task, i.e. for the net_cls
811 * cgroup to which *skb* belongs.
813 * This helper can be used on TC egress path, but not on ingress.
815 * The net_cls cgroup provides an interface to tag network packets
816 * based on a user-provided identifier for all traffic coming from
817 * the tasks belonging to the related cgroup. See also the related
818 * kernel documentation, available from the Linux sources in file
819 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
821 * The Linux kernel has two versions for cgroups: there are
822 * cgroups v1 and cgroups v2. Both are available to users, who can
823 * use a mixture of them, but note that the net_cls cgroup is for
824 * cgroup v1 only. This makes it incompatible with BPF programs
825 * run on cgroups, which is a cgroup-v2-only feature (a socket can
826 * only hold data for one version of cgroups at a time).
828 * This helper is only available is the kernel was compiled with
829 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
830 * "**y**" or to "**m**".
832 * The classid, or 0 for the default unconfigured classid.
834 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
836 * Push a *vlan_tci* (VLAN tag control information) of protocol
837 * *vlan_proto* to the packet associated to *skb*, then update
838 * the checksum. Note that if *vlan_proto* is different from
839 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
840 * be **ETH_P_8021Q**.
842 * A call to this helper is susceptible to change the underlying
843 * packet buffer. Therefore, at load time, all checks on pointers
844 * previously done by the verifier are invalidated and must be
845 * performed again, if the helper is used in combination with
846 * direct packet access.
848 * 0 on success, or a negative error in case of failure.
850 * int bpf_skb_vlan_pop(struct sk_buff *skb)
852 * Pop a VLAN header from the packet associated to *skb*.
854 * A call to this helper is susceptible to change the underlying
855 * packet buffer. Therefore, at load time, all checks on pointers
856 * previously done by the verifier are invalidated and must be
857 * performed again, if the helper is used in combination with
858 * direct packet access.
860 * 0 on success, or a negative error in case of failure.
862 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
864 * Get tunnel metadata. This helper takes a pointer *key* to an
865 * empty **struct bpf_tunnel_key** of **size**, that will be
866 * filled with tunnel metadata for the packet associated to *skb*.
867 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
868 * indicates that the tunnel is based on IPv6 protocol instead of
871 * The **struct bpf_tunnel_key** is an object that generalizes the
872 * principal parameters used by various tunneling protocols into a
873 * single struct. This way, it can be used to easily make a
874 * decision based on the contents of the encapsulation header,
875 * "summarized" in this struct. In particular, it holds the IP
876 * address of the remote end (IPv4 or IPv6, depending on the case)
877 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
878 * this struct exposes the *key*\ **->tunnel_id**, which is
879 * generally mapped to a VNI (Virtual Network Identifier), making
880 * it programmable together with the **bpf_skb_set_tunnel_key**\
883 * Let's imagine that the following code is part of a program
884 * attached to the TC ingress interface, on one end of a GRE
885 * tunnel, and is supposed to filter out all messages coming from
886 * remote ends with IPv4 address other than 10.0.0.1:
891 * struct bpf_tunnel_key key = {};
893 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
895 * return TC_ACT_SHOT; // drop packet
897 * if (key.remote_ipv4 != 0x0a000001)
898 * return TC_ACT_SHOT; // drop packet
900 * return TC_ACT_OK; // accept packet
902 * This interface can also be used with all encapsulation devices
903 * that can operate in "collect metadata" mode: instead of having
904 * one network device per specific configuration, the "collect
905 * metadata" mode only requires a single device where the
906 * configuration can be extracted from this helper.
908 * This can be used together with various tunnels such as VXLan,
909 * Geneve, GRE or IP in IP (IPIP).
911 * 0 on success, or a negative error in case of failure.
913 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
915 * Populate tunnel metadata for packet associated to *skb.* The
916 * tunnel metadata is set to the contents of *key*, of *size*. The
917 * *flags* can be set to a combination of the following values:
919 * **BPF_F_TUNINFO_IPV6**
920 * Indicate that the tunnel is based on IPv6 protocol
922 * **BPF_F_ZERO_CSUM_TX**
923 * For IPv4 packets, add a flag to tunnel metadata
924 * indicating that checksum computation should be skipped
925 * and checksum set to zeroes.
926 * **BPF_F_DONT_FRAGMENT**
927 * Add a flag to tunnel metadata indicating that the
928 * packet should not be fragmented.
929 * **BPF_F_SEQ_NUMBER**
930 * Add a flag to tunnel metadata indicating that a
931 * sequence number should be added to tunnel header before
932 * sending the packet. This flag was added for GRE
933 * encapsulation, but might be used with other protocols
934 * as well in the future.
936 * Here is a typical usage on the transmit path:
940 * struct bpf_tunnel_key key;
942 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
943 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
945 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
946 * helper for additional information.
948 * 0 on success, or a negative error in case of failure.
950 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
952 * Read the value of a perf event counter. This helper relies on a
953 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
954 * the perf event counter is selected when *map* is updated with
955 * perf event file descriptors. The *map* is an array whose size
956 * is the number of available CPUs, and each cell contains a value
957 * relative to one CPU. The value to retrieve is indicated by
958 * *flags*, that contains the index of the CPU to look up, masked
959 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
960 * **BPF_F_CURRENT_CPU** to indicate that the value for the
961 * current CPU should be retrieved.
963 * Note that before Linux 4.13, only hardware perf event can be
966 * Also, be aware that the newer helper
967 * **bpf_perf_event_read_value**\ () is recommended over
968 * **bpf_perf_event_read**\ () in general. The latter has some ABI
969 * quirks where error and counter value are used as a return code
970 * (which is wrong to do since ranges may overlap). This issue is
971 * fixed with **bpf_perf_event_read_value**\ (), which at the same
972 * time provides more features over the **bpf_perf_event_read**\
973 * () interface. Please refer to the description of
974 * **bpf_perf_event_read_value**\ () for details.
976 * The value of the perf event counter read from the map, or a
977 * negative error code in case of failure.
979 * int bpf_redirect(u32 ifindex, u64 flags)
981 * Redirect the packet to another net device of index *ifindex*.
982 * This helper is somewhat similar to **bpf_clone_redirect**\
983 * (), except that the packet is not cloned, which provides
984 * increased performance.
986 * Except for XDP, both ingress and egress interfaces can be used
987 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
988 * to make the distinction (ingress path is selected if the flag
989 * is present, egress path otherwise). Currently, XDP only
990 * supports redirection to the egress interface, and accepts no
993 * The same effect can be attained with the more generic
994 * **bpf_redirect_map**\ (), which requires specific maps to be
995 * used but offers better performance.
997 * For XDP, the helper returns **XDP_REDIRECT** on success or
998 * **XDP_ABORTED** on error. For other program types, the values
999 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1002 * u32 bpf_get_route_realm(struct sk_buff *skb)
1004 * Retrieve the realm or the route, that is to say the
1005 * **tclassid** field of the destination for the *skb*. The
1006 * indentifier retrieved is a user-provided tag, similar to the
1007 * one used with the net_cls cgroup (see description for
1008 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1009 * held by a route (a destination entry), not by a task.
1011 * Retrieving this identifier works with the clsact TC egress hook
1012 * (see also **tc-bpf(8)**), or alternatively on conventional
1013 * classful egress qdiscs, but not on TC ingress path. In case of
1014 * clsact TC egress hook, this has the advantage that, internally,
1015 * the destination entry has not been dropped yet in the transmit
1016 * path. Therefore, the destination entry does not need to be
1017 * artificially held via **netif_keep_dst**\ () for a classful
1018 * qdisc until the *skb* is freed.
1020 * This helper is available only if the kernel was compiled with
1021 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1023 * The realm of the route for the packet associated to *skb*, or 0
1024 * if none was found.
1026 * int bpf_perf_event_output(struct pt_regs *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1028 * Write raw *data* blob into a special BPF perf event held by
1029 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1030 * event must have the following attributes: **PERF_SAMPLE_RAW**
1031 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1032 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1034 * The *flags* are used to indicate the index in *map* for which
1035 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1036 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1037 * to indicate that the index of the current CPU core should be
1040 * The value to write, of *size*, is passed through eBPF stack and
1041 * pointed by *data*.
1043 * The context of the program *ctx* needs also be passed to the
1046 * On user space, a program willing to read the values needs to
1047 * call **perf_event_open**\ () on the perf event (either for
1048 * one or for all CPUs) and to store the file descriptor into the
1049 * *map*. This must be done before the eBPF program can send data
1050 * into it. An example is available in file
1051 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1052 * tree (the eBPF program counterpart is in
1053 * *samples/bpf/trace_output_kern.c*).
1055 * **bpf_perf_event_output**\ () achieves better performance
1056 * than **bpf_trace_printk**\ () for sharing data with user
1057 * space, and is much better suitable for streaming data from eBPF
1060 * Note that this helper is not restricted to tracing use cases
1061 * and can be used with programs attached to TC or XDP as well,
1062 * where it allows for passing data to user space listeners. Data
1065 * * Only custom structs,
1066 * * Only the packet payload, or
1067 * * A combination of both.
1069 * 0 on success, or a negative error in case of failure.
1071 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1073 * This helper was provided as an easy way to load data from a
1074 * packet. It can be used to load *len* bytes from *offset* from
1075 * the packet associated to *skb*, into the buffer pointed by
1078 * Since Linux 4.7, usage of this helper has mostly been replaced
1079 * by "direct packet access", enabling packet data to be
1080 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1081 * pointing respectively to the first byte of packet data and to
1082 * the byte after the last byte of packet data. However, it
1083 * remains useful if one wishes to read large quantities of data
1084 * at once from a packet into the eBPF stack.
1086 * 0 on success, or a negative error in case of failure.
1088 * int bpf_get_stackid(struct pt_regs *ctx, struct bpf_map *map, u64 flags)
1090 * Walk a user or a kernel stack and return its id. To achieve
1091 * this, the helper needs *ctx*, which is a pointer to the context
1092 * on which the tracing program is executed, and a pointer to a
1093 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1095 * The last argument, *flags*, holds the number of stack frames to
1096 * skip (from 0 to 255), masked with
1097 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1098 * a combination of the following flags:
1100 * **BPF_F_USER_STACK**
1101 * Collect a user space stack instead of a kernel stack.
1102 * **BPF_F_FAST_STACK_CMP**
1103 * Compare stacks by hash only.
1104 * **BPF_F_REUSE_STACKID**
1105 * If two different stacks hash into the same *stackid*,
1106 * discard the old one.
1108 * The stack id retrieved is a 32 bit long integer handle which
1109 * can be further combined with other data (including other stack
1110 * ids) and used as a key into maps. This can be useful for
1111 * generating a variety of graphs (such as flame graphs or off-cpu
1114 * For walking a stack, this helper is an improvement over
1115 * **bpf_probe_read**\ (), which can be used with unrolled loops
1116 * but is not efficient and consumes a lot of eBPF instructions.
1117 * Instead, **bpf_get_stackid**\ () can collect up to
1118 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1119 * this limit can be controlled with the **sysctl** program, and
1120 * that it should be manually increased in order to profile long
1121 * user stacks (such as stacks for Java programs). To do so, use:
1125 * # sysctl kernel.perf_event_max_stack=<new value>
1127 * The positive or null stack id on success, or a negative error
1128 * in case of failure.
1130 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1132 * Compute a checksum difference, from the raw buffer pointed by
1133 * *from*, of length *from_size* (that must be a multiple of 4),
1134 * towards the raw buffer pointed by *to*, of size *to_size*
1135 * (same remark). An optional *seed* can be added to the value
1136 * (this can be cascaded, the seed may come from a previous call
1139 * This is flexible enough to be used in several ways:
1141 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1142 * checksum, it can be used when pushing new data.
1143 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1144 * checksum, it can be used when removing data from a packet.
1145 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1146 * can be used to compute a diff. Note that *from_size* and
1147 * *to_size* do not need to be equal.
1149 * This helper can be used in combination with
1150 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1151 * which one can feed in the difference computed with
1152 * **bpf_csum_diff**\ ().
1154 * The checksum result, or a negative error code in case of
1157 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1159 * Retrieve tunnel options metadata for the packet associated to
1160 * *skb*, and store the raw tunnel option data to the buffer *opt*
1163 * This helper can be used with encapsulation devices that can
1164 * operate in "collect metadata" mode (please refer to the related
1165 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1166 * more details). A particular example where this can be used is
1167 * in combination with the Geneve encapsulation protocol, where it
1168 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1169 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1170 * the eBPF program. This allows for full customization of these
1173 * The size of the option data retrieved.
1175 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1177 * Set tunnel options metadata for the packet associated to *skb*
1178 * to the option data contained in the raw buffer *opt* of *size*.
1180 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1181 * helper for additional information.
1183 * 0 on success, or a negative error in case of failure.
1185 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1187 * Change the protocol of the *skb* to *proto*. Currently
1188 * supported are transition from IPv4 to IPv6, and from IPv6 to
1189 * IPv4. The helper takes care of the groundwork for the
1190 * transition, including resizing the socket buffer. The eBPF
1191 * program is expected to fill the new headers, if any, via
1192 * **skb_store_bytes**\ () and to recompute the checksums with
1193 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1194 * (). The main case for this helper is to perform NAT64
1195 * operations out of an eBPF program.
1197 * Internally, the GSO type is marked as dodgy so that headers are
1198 * checked and segments are recalculated by the GSO/GRO engine.
1199 * The size for GSO target is adapted as well.
1201 * All values for *flags* are reserved for future usage, and must
1204 * A call to this helper is susceptible to change the underlying
1205 * packet buffer. Therefore, at load time, all checks on pointers
1206 * previously done by the verifier are invalidated and must be
1207 * performed again, if the helper is used in combination with
1208 * direct packet access.
1210 * 0 on success, or a negative error in case of failure.
1212 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1214 * Change the packet type for the packet associated to *skb*. This
1215 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1216 * the eBPF program does not have a write access to *skb*\
1217 * **->pkt_type** beside this helper. Using a helper here allows
1218 * for graceful handling of errors.
1220 * The major use case is to change incoming *skb*s to
1221 * **PACKET_HOST** in a programmatic way instead of having to
1222 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1225 * Note that *type* only allows certain values. At this time, they
1230 * **PACKET_BROADCAST**
1231 * Send packet to all.
1232 * **PACKET_MULTICAST**
1233 * Send packet to group.
1234 * **PACKET_OTHERHOST**
1235 * Send packet to someone else.
1237 * 0 on success, or a negative error in case of failure.
1239 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1241 * Check whether *skb* is a descendant of the cgroup2 held by
1242 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1244 * The return value depends on the result of the test, and can be:
1246 * * 0, if the *skb* failed the cgroup2 descendant test.
1247 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1248 * * A negative error code, if an error occurred.
1250 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1252 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1253 * not set, in particular if the hash was cleared due to mangling,
1254 * recompute this hash. Later accesses to the hash can be done
1255 * directly with *skb*\ **->hash**.
1257 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1258 * prototype with **bpf_skb_change_proto**\ (), or calling
1259 * **bpf_skb_store_bytes**\ () with the
1260 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1261 * the hash and to trigger a new computation for the next call to
1262 * **bpf_get_hash_recalc**\ ().
1266 * u64 bpf_get_current_task(void)
1268 * A pointer to the current task struct.
1270 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1272 * Attempt in a safe way to write *len* bytes from the buffer
1273 * *src* to *dst* in memory. It only works for threads that are in
1274 * user context, and *dst* must be a valid user space address.
1276 * This helper should not be used to implement any kind of
1277 * security mechanism because of TOC-TOU attacks, but rather to
1278 * debug, divert, and manipulate execution of semi-cooperative
1281 * Keep in mind that this feature is meant for experiments, and it
1282 * has a risk of crashing the system and running programs.
1283 * Therefore, when an eBPF program using this helper is attached,
1284 * a warning including PID and process name is printed to kernel
1287 * 0 on success, or a negative error in case of failure.
1289 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1291 * Check whether the probe is being run is the context of a given
1292 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1293 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1295 * The return value depends on the result of the test, and can be:
1297 * * 0, if the *skb* task belongs to the cgroup2.
1298 * * 1, if the *skb* task does not belong to the cgroup2.
1299 * * A negative error code, if an error occurred.
1301 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1303 * Resize (trim or grow) the packet associated to *skb* to the
1304 * new *len*. The *flags* are reserved for future usage, and must
1307 * The basic idea is that the helper performs the needed work to
1308 * change the size of the packet, then the eBPF program rewrites
1309 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1310 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1311 * and others. This helper is a slow path utility intended for
1312 * replies with control messages. And because it is targeted for
1313 * slow path, the helper itself can afford to be slow: it
1314 * implicitly linearizes, unclones and drops offloads from the
1317 * A call to this helper is susceptible to change the underlying
1318 * packet buffer. Therefore, at load time, all checks on pointers
1319 * previously done by the verifier are invalidated and must be
1320 * performed again, if the helper is used in combination with
1321 * direct packet access.
1323 * 0 on success, or a negative error in case of failure.
1325 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1327 * Pull in non-linear data in case the *skb* is non-linear and not
1328 * all of *len* are part of the linear section. Make *len* bytes
1329 * from *skb* readable and writable. If a zero value is passed for
1330 * *len*, then the whole length of the *skb* is pulled.
1332 * This helper is only needed for reading and writing with direct
1335 * For direct packet access, testing that offsets to access
1336 * are within packet boundaries (test on *skb*\ **->data_end**) is
1337 * susceptible to fail if offsets are invalid, or if the requested
1338 * data is in non-linear parts of the *skb*. On failure the
1339 * program can just bail out, or in the case of a non-linear
1340 * buffer, use a helper to make the data available. The
1341 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1342 * the data. Another one consists in using **bpf_skb_pull_data**
1343 * to pull in once the non-linear parts, then retesting and
1344 * eventually access the data.
1346 * At the same time, this also makes sure the *skb* is uncloned,
1347 * which is a necessary condition for direct write. As this needs
1348 * to be an invariant for the write part only, the verifier
1349 * detects writes and adds a prologue that is calling
1350 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1351 * the very beginning in case it is indeed cloned.
1353 * A call to this helper is susceptible to change the underlying
1354 * packet buffer. Therefore, at load time, all checks on pointers
1355 * previously done by the verifier are invalidated and must be
1356 * performed again, if the helper is used in combination with
1357 * direct packet access.
1359 * 0 on success, or a negative error in case of failure.
1361 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1363 * Add the checksum *csum* into *skb*\ **->csum** in case the
1364 * driver has supplied a checksum for the entire packet into that
1365 * field. Return an error otherwise. This helper is intended to be
1366 * used in combination with **bpf_csum_diff**\ (), in particular
1367 * when the checksum needs to be updated after data has been
1368 * written into the packet through direct packet access.
1370 * The checksum on success, or a negative error code in case of
1373 * void bpf_set_hash_invalid(struct sk_buff *skb)
1375 * Invalidate the current *skb*\ **->hash**. It can be used after
1376 * mangling on headers through direct packet access, in order to
1377 * indicate that the hash is outdated and to trigger a
1378 * recalculation the next time the kernel tries to access this
1379 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1381 * int bpf_get_numa_node_id(void)
1383 * Return the id of the current NUMA node. The primary use case
1384 * for this helper is the selection of sockets for the local NUMA
1385 * node, when the program is attached to sockets using the
1386 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1387 * but the helper is also available to other eBPF program types,
1388 * similarly to **bpf_get_smp_processor_id**\ ().
1390 * The id of current NUMA node.
1392 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1394 * Grows headroom of packet associated to *skb* and adjusts the
1395 * offset of the MAC header accordingly, adding *len* bytes of
1396 * space. It automatically extends and reallocates memory as
1399 * This helper can be used on a layer 3 *skb* to push a MAC header
1400 * for redirection into a layer 2 device.
1402 * All values for *flags* are reserved for future usage, and must
1405 * A call to this helper is susceptible to change the underlying
1406 * packet buffer. Therefore, at load time, all checks on pointers
1407 * previously done by the verifier are invalidated and must be
1408 * performed again, if the helper is used in combination with
1409 * direct packet access.
1411 * 0 on success, or a negative error in case of failure.
1413 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1415 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1416 * it is possible to use a negative value for *delta*. This helper
1417 * can be used to prepare the packet for pushing or popping
1420 * A call to this helper is susceptible to change the underlying
1421 * packet buffer. Therefore, at load time, all checks on pointers
1422 * previously done by the verifier are invalidated and must be
1423 * performed again, if the helper is used in combination with
1424 * direct packet access.
1426 * 0 on success, or a negative error in case of failure.
1428 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1430 * Copy a NUL terminated string from an unsafe address
1431 * *unsafe_ptr* to *dst*. The *size* should include the
1432 * terminating NUL byte. In case the string length is smaller than
1433 * *size*, the target is not padded with further NUL bytes. If the
1434 * string length is larger than *size*, just *size*-1 bytes are
1435 * copied and the last byte is set to NUL.
1437 * On success, the length of the copied string is returned. This
1438 * makes this helper useful in tracing programs for reading
1439 * strings, and more importantly to get its length at runtime. See
1440 * the following snippet:
1444 * SEC("kprobe/sys_open")
1445 * void bpf_sys_open(struct pt_regs *ctx)
1447 * char buf[PATHLEN]; // PATHLEN is defined to 256
1448 * int res = bpf_probe_read_str(buf, sizeof(buf),
1451 * // Consume buf, for example push it to
1452 * // userspace via bpf_perf_event_output(); we
1453 * // can use res (the string length) as event
1454 * // size, after checking its boundaries.
1457 * In comparison, using **bpf_probe_read()** helper here instead
1458 * to read the string would require to estimate the length at
1459 * compile time, and would often result in copying more memory
1462 * Another useful use case is when parsing individual process
1463 * arguments or individual environment variables navigating
1464 * *current*\ **->mm->arg_start** and *current*\
1465 * **->mm->env_start**: using this helper and the return value,
1466 * one can quickly iterate at the right offset of the memory area.
1468 * On success, the strictly positive length of the string,
1469 * including the trailing NUL character. On error, a negative
1472 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1474 * If the **struct sk_buff** pointed by *skb* has a known socket,
1475 * retrieve the cookie (generated by the kernel) of this socket.
1476 * If no cookie has been set yet, generate a new cookie. Once
1477 * generated, the socket cookie remains stable for the life of the
1478 * socket. This helper can be useful for monitoring per socket
1479 * networking traffic statistics as it provides a global socket
1480 * identifier that can be assumed unique.
1482 * A 8-byte long non-decreasing number on success, or 0 if the
1483 * socket field is missing inside *skb*.
1485 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1487 * Equivalent to bpf_get_socket_cookie() helper that accepts
1488 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1490 * A 8-byte long non-decreasing number.
1492 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1494 * Equivalent to bpf_get_socket_cookie() helper that accepts
1495 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1497 * A 8-byte long non-decreasing number.
1499 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1501 * The owner UID of the socket associated to *skb*. If the socket
1502 * is **NULL**, or if it is not a full socket (i.e. if it is a
1503 * time-wait or a request socket instead), **overflowuid** value
1504 * is returned (note that **overflowuid** might also be the actual
1505 * UID value for the socket).
1507 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1509 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1514 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1516 * Emulate a call to **setsockopt()** on the socket associated to
1517 * *bpf_socket*, which must be a full socket. The *level* at
1518 * which the option resides and the name *optname* of the option
1519 * must be specified, see **setsockopt(2)** for more information.
1520 * The option value of length *optlen* is pointed by *optval*.
1522 * This helper actually implements a subset of **setsockopt()**.
1523 * It supports the following *level*\ s:
1525 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1526 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1527 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1528 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1529 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1530 * **TCP_BPF_SNDCWND_CLAMP**.
1531 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1532 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1534 * 0 on success, or a negative error in case of failure.
1536 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1538 * Grow or shrink the room for data in the packet associated to
1539 * *skb* by *len_diff*, and according to the selected *mode*.
1541 * There are two supported modes at this time:
1543 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1544 * (room space is added or removed below the layer 2 header).
1546 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1547 * (room space is added or removed below the layer 3 header).
1549 * The following flags are supported at this time:
1551 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1552 * Adjusting mss in this way is not allowed for datagrams.
1554 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1555 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1556 * Any new space is reserved to hold a tunnel header.
1557 * Configure skb offsets and other fields accordingly.
1559 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1560 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1561 * Use with ENCAP_L3 flags to further specify the tunnel type.
1563 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1564 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1565 * type; *len* is the length of the inner MAC header.
1567 * A call to this helper is susceptible to change the underlying
1568 * packet buffer. Therefore, at load time, all checks on pointers
1569 * previously done by the verifier are invalidated and must be
1570 * performed again, if the helper is used in combination with
1571 * direct packet access.
1573 * 0 on success, or a negative error in case of failure.
1575 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1577 * Redirect the packet to the endpoint referenced by *map* at
1578 * index *key*. Depending on its type, this *map* can contain
1579 * references to net devices (for forwarding packets through other
1580 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1581 * but this is only implemented for native XDP (with driver
1582 * support) as of this writing).
1584 * The lower two bits of *flags* are used as the return code if
1585 * the map lookup fails. This is so that the return value can be
1586 * one of the XDP program return codes up to XDP_TX, as chosen by
1587 * the caller. Any higher bits in the *flags* argument must be
1590 * When used to redirect packets to net devices, this helper
1591 * provides a high performance increase over **bpf_redirect**\ ().
1592 * This is due to various implementation details of the underlying
1593 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1594 * () tries to send packet as a "bulk" to the device.
1596 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1598 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1600 * Redirect the packet to the socket referenced by *map* (of type
1601 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1602 * egress interfaces can be used for redirection. The
1603 * **BPF_F_INGRESS** value in *flags* is used to make the
1604 * distinction (ingress path is selected if the flag is present,
1605 * egress path otherwise). This is the only flag supported for now.
1607 * **SK_PASS** on success, or **SK_DROP** on error.
1609 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1611 * Add an entry to, or update a *map* referencing sockets. The
1612 * *skops* is used as a new value for the entry associated to
1613 * *key*. *flags* is one of:
1616 * The entry for *key* must not exist in the map.
1618 * The entry for *key* must already exist in the map.
1620 * No condition on the existence of the entry for *key*.
1622 * If the *map* has eBPF programs (parser and verdict), those will
1623 * be inherited by the socket being added. If the socket is
1624 * already attached to eBPF programs, this results in an error.
1626 * 0 on success, or a negative error in case of failure.
1628 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1630 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1631 * *delta* (which can be positive or negative). Note that this
1632 * operation modifies the address stored in *xdp_md*\ **->data**,
1633 * so the latter must be loaded only after the helper has been
1636 * The use of *xdp_md*\ **->data_meta** is optional and programs
1637 * are not required to use it. The rationale is that when the
1638 * packet is processed with XDP (e.g. as DoS filter), it is
1639 * possible to push further meta data along with it before passing
1640 * to the stack, and to give the guarantee that an ingress eBPF
1641 * program attached as a TC classifier on the same device can pick
1642 * this up for further post-processing. Since TC works with socket
1643 * buffers, it remains possible to set from XDP the **mark** or
1644 * **priority** pointers, or other pointers for the socket buffer.
1645 * Having this scratch space generic and programmable allows for
1646 * more flexibility as the user is free to store whatever meta
1649 * A call to this helper is susceptible to change the underlying
1650 * packet buffer. Therefore, at load time, all checks on pointers
1651 * previously done by the verifier are invalidated and must be
1652 * performed again, if the helper is used in combination with
1653 * direct packet access.
1655 * 0 on success, or a negative error in case of failure.
1657 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1659 * Read the value of a perf event counter, and store it into *buf*
1660 * of size *buf_size*. This helper relies on a *map* of type
1661 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1662 * counter is selected when *map* is updated with perf event file
1663 * descriptors. The *map* is an array whose size is the number of
1664 * available CPUs, and each cell contains a value relative to one
1665 * CPU. The value to retrieve is indicated by *flags*, that
1666 * contains the index of the CPU to look up, masked with
1667 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1668 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1669 * current CPU should be retrieved.
1671 * This helper behaves in a way close to
1672 * **bpf_perf_event_read**\ () helper, save that instead of
1673 * just returning the value observed, it fills the *buf*
1674 * structure. This allows for additional data to be retrieved: in
1675 * particular, the enabled and running times (in *buf*\
1676 * **->enabled** and *buf*\ **->running**, respectively) are
1677 * copied. In general, **bpf_perf_event_read_value**\ () is
1678 * recommended over **bpf_perf_event_read**\ (), which has some
1679 * ABI issues and provides fewer functionalities.
1681 * These values are interesting, because hardware PMU (Performance
1682 * Monitoring Unit) counters are limited resources. When there are
1683 * more PMU based perf events opened than available counters,
1684 * kernel will multiplex these events so each event gets certain
1685 * percentage (but not all) of the PMU time. In case that
1686 * multiplexing happens, the number of samples or counter value
1687 * will not reflect the case compared to when no multiplexing
1688 * occurs. This makes comparison between different runs difficult.
1689 * Typically, the counter value should be normalized before
1690 * comparing to other experiments. The usual normalization is done
1695 * normalized_counter = counter * t_enabled / t_running
1697 * Where t_enabled is the time enabled for event and t_running is
1698 * the time running for event since last normalization. The
1699 * enabled and running times are accumulated since the perf event
1700 * open. To achieve scaling factor between two invocations of an
1701 * eBPF program, users can can use CPU id as the key (which is
1702 * typical for perf array usage model) to remember the previous
1703 * value and do the calculation inside the eBPF program.
1705 * 0 on success, or a negative error in case of failure.
1707 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1709 * For en eBPF program attached to a perf event, retrieve the
1710 * value of the event counter associated to *ctx* and store it in
1711 * the structure pointed by *buf* and of size *buf_size*. Enabled
1712 * and running times are also stored in the structure (see
1713 * description of helper **bpf_perf_event_read_value**\ () for
1716 * 0 on success, or a negative error in case of failure.
1718 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1720 * Emulate a call to **getsockopt()** on the socket associated to
1721 * *bpf_socket*, which must be a full socket. The *level* at
1722 * which the option resides and the name *optname* of the option
1723 * must be specified, see **getsockopt(2)** for more information.
1724 * The retrieved value is stored in the structure pointed by
1725 * *opval* and of length *optlen*.
1727 * This helper actually implements a subset of **getsockopt()**.
1728 * It supports the following *level*\ s:
1730 * * **IPPROTO_TCP**, which supports *optname*
1731 * **TCP_CONGESTION**.
1732 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1733 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1735 * 0 on success, or a negative error in case of failure.
1737 * int bpf_override_return(struct pt_regs *regs, u64 rc)
1739 * Used for error injection, this helper uses kprobes to override
1740 * the return value of the probed function, and to set it to *rc*.
1741 * The first argument is the context *regs* on which the kprobe
1744 * This helper works by setting setting the PC (program counter)
1745 * to an override function which is run in place of the original
1746 * probed function. This means the probed function is not run at
1747 * all. The replacement function just returns with the required
1750 * This helper has security implications, and thus is subject to
1751 * restrictions. It is only available if the kernel was compiled
1752 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1753 * option, and in this case it only works on functions tagged with
1754 * **ALLOW_ERROR_INJECTION** in the kernel code.
1756 * Also, the helper is only available for the architectures having
1757 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1758 * x86 architecture is the only one to support this feature.
1762 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1764 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1765 * for the full TCP socket associated to *bpf_sock_ops* to
1768 * The primary use of this field is to determine if there should
1769 * be calls to eBPF programs of type
1770 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1771 * code. A program of the same type can change its value, per
1772 * connection and as necessary, when the connection is
1773 * established. This field is directly accessible for reading, but
1774 * this helper must be used for updates in order to return an
1775 * error if an eBPF program tries to set a callback that is not
1776 * supported in the current kernel.
1778 * *argval* is a flag array which can combine these flags:
1780 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1781 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1782 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1783 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1785 * Therefore, this function can be used to clear a callback flag by
1786 * setting the appropriate bit to zero. e.g. to disable the RTO
1789 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1790 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1792 * Here are some examples of where one could call such eBPF
1796 * * When a packet is retransmitted.
1797 * * When the connection terminates.
1798 * * When a packet is sent.
1799 * * When a packet is received.
1801 * Code **-EINVAL** if the socket is not a full TCP socket;
1802 * otherwise, a positive number containing the bits that could not
1803 * be set is returned (which comes down to 0 if all bits were set
1806 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1808 * This helper is used in programs implementing policies at the
1809 * socket level. If the message *msg* is allowed to pass (i.e. if
1810 * the verdict eBPF program returns **SK_PASS**), redirect it to
1811 * the socket referenced by *map* (of type
1812 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1813 * egress interfaces can be used for redirection. The
1814 * **BPF_F_INGRESS** value in *flags* is used to make the
1815 * distinction (ingress path is selected if the flag is present,
1816 * egress path otherwise). This is the only flag supported for now.
1818 * **SK_PASS** on success, or **SK_DROP** on error.
1820 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1822 * For socket policies, apply the verdict of the eBPF program to
1823 * the next *bytes* (number of bytes) of message *msg*.
1825 * For example, this helper can be used in the following cases:
1827 * * A single **sendmsg**\ () or **sendfile**\ () system call
1828 * contains multiple logical messages that the eBPF program is
1829 * supposed to read and for which it should apply a verdict.
1830 * * An eBPF program only cares to read the first *bytes* of a
1831 * *msg*. If the message has a large payload, then setting up
1832 * and calling the eBPF program repeatedly for all bytes, even
1833 * though the verdict is already known, would create unnecessary
1836 * When called from within an eBPF program, the helper sets a
1837 * counter internal to the BPF infrastructure, that is used to
1838 * apply the last verdict to the next *bytes*. If *bytes* is
1839 * smaller than the current data being processed from a
1840 * **sendmsg**\ () or **sendfile**\ () system call, the first
1841 * *bytes* will be sent and the eBPF program will be re-run with
1842 * the pointer for start of data pointing to byte number *bytes*
1843 * **+ 1**. If *bytes* is larger than the current data being
1844 * processed, then the eBPF verdict will be applied to multiple
1845 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1848 * Note that if a socket closes with the internal counter holding
1849 * a non-zero value, this is not a problem because data is not
1850 * being buffered for *bytes* and is sent as it is received.
1854 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1856 * For socket policies, prevent the execution of the verdict eBPF
1857 * program for message *msg* until *bytes* (byte number) have been
1860 * This can be used when one needs a specific number of bytes
1861 * before a verdict can be assigned, even if the data spans
1862 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1863 * case would be a user calling **sendmsg**\ () repeatedly with
1864 * 1-byte long message segments. Obviously, this is bad for
1865 * performance, but it is still valid. If the eBPF program needs
1866 * *bytes* bytes to validate a header, this helper can be used to
1867 * prevent the eBPF program to be called again until *bytes* have
1872 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1874 * For socket policies, pull in non-linear data from user space
1875 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1876 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1879 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1880 * *msg* it can only parse data that the (**data**, **data_end**)
1881 * pointers have already consumed. For **sendmsg**\ () hooks this
1882 * is likely the first scatterlist element. But for calls relying
1883 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1884 * be the range (**0**, **0**) because the data is shared with
1885 * user space and by default the objective is to avoid allowing
1886 * user space to modify data while (or after) eBPF verdict is
1887 * being decided. This helper can be used to pull in data and to
1888 * set the start and end pointer to given values. Data will be
1889 * copied if necessary (i.e. if data was not linear and if start
1890 * and end pointers do not point to the same chunk).
1892 * A call to this helper is susceptible to change the underlying
1893 * packet buffer. Therefore, at load time, all checks on pointers
1894 * previously done by the verifier are invalidated and must be
1895 * performed again, if the helper is used in combination with
1896 * direct packet access.
1898 * All values for *flags* are reserved for future usage, and must
1901 * 0 on success, or a negative error in case of failure.
1903 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1905 * Bind the socket associated to *ctx* to the address pointed by
1906 * *addr*, of length *addr_len*. This allows for making outgoing
1907 * connection from the desired IP address, which can be useful for
1908 * example when all processes inside a cgroup should use one
1909 * single IP address on a host that has multiple IP configured.
1911 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1912 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1913 * **AF_INET6**). Looking for a free port to bind to can be
1914 * expensive, therefore binding to port is not permitted by the
1915 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1916 * must be set to zero.
1918 * 0 on success, or a negative error in case of failure.
1920 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1922 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1923 * only possible to shrink the packet as of this writing,
1924 * therefore *delta* must be a negative integer.
1926 * A call to this helper is susceptible to change the underlying
1927 * packet buffer. Therefore, at load time, all checks on pointers
1928 * previously done by the verifier are invalidated and must be
1929 * performed again, if the helper is used in combination with
1930 * direct packet access.
1932 * 0 on success, or a negative error in case of failure.
1934 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1936 * Retrieve the XFRM state (IP transform framework, see also
1937 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1939 * The retrieved value is stored in the **struct bpf_xfrm_state**
1940 * pointed by *xfrm_state* and of length *size*.
1942 * All values for *flags* are reserved for future usage, and must
1945 * This helper is available only if the kernel was compiled with
1946 * **CONFIG_XFRM** configuration option.
1948 * 0 on success, or a negative error in case of failure.
1950 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1952 * Return a user or a kernel stack in bpf program provided buffer.
1953 * To achieve this, the helper needs *ctx*, which is a pointer
1954 * to the context on which the tracing program is executed.
1955 * To store the stacktrace, the bpf program provides *buf* with
1956 * a nonnegative *size*.
1958 * The last argument, *flags*, holds the number of stack frames to
1959 * skip (from 0 to 255), masked with
1960 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1961 * the following flags:
1963 * **BPF_F_USER_STACK**
1964 * Collect a user space stack instead of a kernel stack.
1965 * **BPF_F_USER_BUILD_ID**
1966 * Collect buildid+offset instead of ips for user stack,
1967 * only valid if **BPF_F_USER_STACK** is also specified.
1969 * **bpf_get_stack**\ () can collect up to
1970 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1971 * to sufficient large buffer size. Note that
1972 * this limit can be controlled with the **sysctl** program, and
1973 * that it should be manually increased in order to profile long
1974 * user stacks (such as stacks for Java programs). To do so, use:
1978 * # sysctl kernel.perf_event_max_stack=<new value>
1980 * A non-negative value equal to or less than *size* on success,
1981 * or a negative error in case of failure.
1983 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1985 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1986 * it provides an easy way to load *len* bytes from *offset*
1987 * from the packet associated to *skb*, into the buffer pointed
1988 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1989 * a fifth argument *start_header* exists in order to select a
1990 * base offset to start from. *start_header* can be one of:
1992 * **BPF_HDR_START_MAC**
1993 * Base offset to load data from is *skb*'s mac header.
1994 * **BPF_HDR_START_NET**
1995 * Base offset to load data from is *skb*'s network header.
1997 * In general, "direct packet access" is the preferred method to
1998 * access packet data, however, this helper is in particular useful
1999 * in socket filters where *skb*\ **->data** does not always point
2000 * to the start of the mac header and where "direct packet access"
2003 * 0 on success, or a negative error in case of failure.
2005 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2007 * Do FIB lookup in kernel tables using parameters in *params*.
2008 * If lookup is successful and result shows packet is to be
2009 * forwarded, the neighbor tables are searched for the nexthop.
2010 * If successful (ie., FIB lookup shows forwarding and nexthop
2011 * is resolved), the nexthop address is returned in ipv4_dst
2012 * or ipv6_dst based on family, smac is set to mac address of
2013 * egress device, dmac is set to nexthop mac address, rt_metric
2014 * is set to metric from route (IPv4/IPv6 only), and ifindex
2015 * is set to the device index of the nexthop from the FIB lookup.
2017 * *plen* argument is the size of the passed in struct.
2018 * *flags* argument can be a combination of one or more of the
2021 * **BPF_FIB_LOOKUP_DIRECT**
2022 * Do a direct table lookup vs full lookup using FIB
2024 * **BPF_FIB_LOOKUP_OUTPUT**
2025 * Perform lookup from an egress perspective (default is
2028 * *ctx* is either **struct xdp_md** for XDP programs or
2029 * **struct sk_buff** tc cls_act programs.
2031 * * < 0 if any input argument is invalid
2032 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2033 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2034 * packet is not forwarded or needs assist from full stack
2036 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
2038 * Add an entry to, or update a sockhash *map* referencing sockets.
2039 * The *skops* is used as a new value for the entry associated to
2040 * *key*. *flags* is one of:
2043 * The entry for *key* must not exist in the map.
2045 * The entry for *key* must already exist in the map.
2047 * No condition on the existence of the entry for *key*.
2049 * If the *map* has eBPF programs (parser and verdict), those will
2050 * be inherited by the socket being added. If the socket is
2051 * already attached to eBPF programs, this results in an error.
2053 * 0 on success, or a negative error in case of failure.
2055 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2057 * This helper is used in programs implementing policies at the
2058 * socket level. If the message *msg* is allowed to pass (i.e. if
2059 * the verdict eBPF program returns **SK_PASS**), redirect it to
2060 * the socket referenced by *map* (of type
2061 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2062 * egress interfaces can be used for redirection. The
2063 * **BPF_F_INGRESS** value in *flags* is used to make the
2064 * distinction (ingress path is selected if the flag is present,
2065 * egress path otherwise). This is the only flag supported for now.
2067 * **SK_PASS** on success, or **SK_DROP** on error.
2069 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2071 * This helper is used in programs implementing policies at the
2072 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2073 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2074 * to the socket referenced by *map* (of type
2075 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2076 * egress interfaces can be used for redirection. The
2077 * **BPF_F_INGRESS** value in *flags* is used to make the
2078 * distinction (ingress path is selected if the flag is present,
2079 * egress otherwise). This is the only flag supported for now.
2081 * **SK_PASS** on success, or **SK_DROP** on error.
2083 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2085 * Encapsulate the packet associated to *skb* within a Layer 3
2086 * protocol header. This header is provided in the buffer at
2087 * address *hdr*, with *len* its size in bytes. *type* indicates
2088 * the protocol of the header and can be one of:
2090 * **BPF_LWT_ENCAP_SEG6**
2091 * IPv6 encapsulation with Segment Routing Header
2092 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2093 * the IPv6 header is computed by the kernel.
2094 * **BPF_LWT_ENCAP_SEG6_INLINE**
2095 * Only works if *skb* contains an IPv6 packet. Insert a
2096 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2098 * **BPF_LWT_ENCAP_IP**
2099 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2100 * must be IPv4 or IPv6, followed by zero or more
2101 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2102 * total bytes in all prepended headers. Please note that
2103 * if **skb_is_gso**\ (*skb*) is true, no more than two
2104 * headers can be prepended, and the inner header, if
2105 * present, should be either GRE or UDP/GUE.
2107 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2108 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2109 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2110 * **BPF_PROG_TYPE_LWT_XMIT**.
2112 * A call to this helper is susceptible to change the underlying
2113 * packet buffer. Therefore, at load time, all checks on pointers
2114 * previously done by the verifier are invalidated and must be
2115 * performed again, if the helper is used in combination with
2116 * direct packet access.
2118 * 0 on success, or a negative error in case of failure.
2120 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2122 * Store *len* bytes from address *from* into the packet
2123 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2124 * inside the outermost IPv6 Segment Routing Header can be
2125 * modified through this helper.
2127 * A call to this helper is susceptible to change the underlying
2128 * packet buffer. Therefore, at load time, all checks on pointers
2129 * previously done by the verifier are invalidated and must be
2130 * performed again, if the helper is used in combination with
2131 * direct packet access.
2133 * 0 on success, or a negative error in case of failure.
2135 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2137 * Adjust the size allocated to TLVs in the outermost IPv6
2138 * Segment Routing Header contained in the packet associated to
2139 * *skb*, at position *offset* by *delta* bytes. Only offsets
2140 * after the segments are accepted. *delta* can be as well
2141 * positive (growing) as negative (shrinking).
2143 * A call to this helper is susceptible to change the underlying
2144 * packet buffer. Therefore, at load time, all checks on pointers
2145 * previously done by the verifier are invalidated and must be
2146 * performed again, if the helper is used in combination with
2147 * direct packet access.
2149 * 0 on success, or a negative error in case of failure.
2151 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2153 * Apply an IPv6 Segment Routing action of type *action* to the
2154 * packet associated to *skb*. Each action takes a parameter
2155 * contained at address *param*, and of length *param_len* bytes.
2156 * *action* can be one of:
2158 * **SEG6_LOCAL_ACTION_END_X**
2159 * End.X action: Endpoint with Layer-3 cross-connect.
2160 * Type of *param*: **struct in6_addr**.
2161 * **SEG6_LOCAL_ACTION_END_T**
2162 * End.T action: Endpoint with specific IPv6 table lookup.
2163 * Type of *param*: **int**.
2164 * **SEG6_LOCAL_ACTION_END_B6**
2165 * End.B6 action: Endpoint bound to an SRv6 policy.
2166 * Type of *param*: **struct ipv6_sr_hdr**.
2167 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2168 * End.B6.Encap action: Endpoint bound to an SRv6
2169 * encapsulation policy.
2170 * Type of *param*: **struct ipv6_sr_hdr**.
2172 * A call to this helper is susceptible to change the underlying
2173 * packet buffer. Therefore, at load time, all checks on pointers
2174 * previously done by the verifier are invalidated and must be
2175 * performed again, if the helper is used in combination with
2176 * direct packet access.
2178 * 0 on success, or a negative error in case of failure.
2180 * int bpf_rc_repeat(void *ctx)
2182 * This helper is used in programs implementing IR decoding, to
2183 * report a successfully decoded repeat key message. This delays
2184 * the generation of a key up event for previously generated
2187 * Some IR protocols like NEC have a special IR message for
2188 * repeating last button, for when a button is held down.
2190 * The *ctx* should point to the lirc sample as passed into
2193 * This helper is only available is the kernel was compiled with
2194 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2199 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2201 * This helper is used in programs implementing IR decoding, to
2202 * report a successfully decoded key press with *scancode*,
2203 * *toggle* value in the given *protocol*. The scancode will be
2204 * translated to a keycode using the rc keymap, and reported as
2205 * an input key down event. After a period a key up event is
2206 * generated. This period can be extended by calling either
2207 * **bpf_rc_keydown**\ () again with the same values, or calling
2208 * **bpf_rc_repeat**\ ().
2210 * Some protocols include a toggle bit, in case the button was
2211 * released and pressed again between consecutive scancodes.
2213 * The *ctx* should point to the lirc sample as passed into
2216 * The *protocol* is the decoded protocol number (see
2217 * **enum rc_proto** for some predefined values).
2219 * This helper is only available is the kernel was compiled with
2220 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2225 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2227 * Return the cgroup v2 id of the socket associated with the *skb*.
2228 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2229 * helper for cgroup v1 by providing a tag resp. identifier that
2230 * can be matched on or used for map lookups e.g. to implement
2231 * policy. The cgroup v2 id of a given path in the hierarchy is
2232 * exposed in user space through the f_handle API in order to get
2233 * to the same 64-bit id.
2235 * This helper can be used on TC egress path, but not on ingress,
2236 * and is available only if the kernel was compiled with the
2237 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2239 * The id is returned or 0 in case the id could not be retrieved.
2241 * u64 bpf_get_current_cgroup_id(void)
2243 * A 64-bit integer containing the current cgroup id based
2244 * on the cgroup within which the current task is running.
2246 * void *bpf_get_local_storage(void *map, u64 flags)
2248 * Get the pointer to the local storage area.
2249 * The type and the size of the local storage is defined
2250 * by the *map* argument.
2251 * The *flags* meaning is specific for each map type,
2252 * and has to be 0 for cgroup local storage.
2254 * Depending on the BPF program type, a local storage area
2255 * can be shared between multiple instances of the BPF program,
2256 * running simultaneously.
2258 * A user should care about the synchronization by himself.
2259 * For example, by using the **BPF_STX_XADD** instruction to alter
2262 * A pointer to the local storage area.
2264 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2266 * Select a **SO_REUSEPORT** socket from a
2267 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2268 * It checks the selected socket is matching the incoming
2269 * request in the socket buffer.
2271 * 0 on success, or a negative error in case of failure.
2273 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2275 * Return id of cgroup v2 that is ancestor of cgroup associated
2276 * with the *skb* at the *ancestor_level*. The root cgroup is at
2277 * *ancestor_level* zero and each step down the hierarchy
2278 * increments the level. If *ancestor_level* == level of cgroup
2279 * associated with *skb*, then return value will be same as that
2280 * of **bpf_skb_cgroup_id**\ ().
2282 * The helper is useful to implement policies based on cgroups
2283 * that are upper in hierarchy than immediate cgroup associated
2286 * The format of returned id and helper limitations are same as in
2287 * **bpf_skb_cgroup_id**\ ().
2289 * The id is returned or 0 in case the id could not be retrieved.
2291 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2293 * Look for TCP socket matching *tuple*, optionally in a child
2294 * network namespace *netns*. The return value must be checked,
2295 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2297 * The *ctx* should point to the context of the program, such as
2298 * the skb or socket (depending on the hook in use). This is used
2299 * to determine the base network namespace for the lookup.
2301 * *tuple_size* must be one of:
2303 * **sizeof**\ (*tuple*\ **->ipv4**)
2304 * Look for an IPv4 socket.
2305 * **sizeof**\ (*tuple*\ **->ipv6**)
2306 * Look for an IPv6 socket.
2308 * If the *netns* is a negative signed 32-bit integer, then the
2309 * socket lookup table in the netns associated with the *ctx* will
2310 * will be used. For the TC hooks, this is the netns of the device
2311 * in the skb. For socket hooks, this is the netns of the socket.
2312 * If *netns* is any other signed 32-bit value greater than or
2313 * equal to zero then it specifies the ID of the netns relative to
2314 * the netns associated with the *ctx*. *netns* values beyond the
2315 * range of 32-bit integers are reserved for future use.
2317 * All values for *flags* are reserved for future usage, and must
2320 * This helper is available only if the kernel was compiled with
2321 * **CONFIG_NET** configuration option.
2323 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2324 * For sockets with reuseport option, the **struct bpf_sock**
2325 * result is from *reuse*\ **->socks**\ [] using the hash of the
2328 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2330 * Look for UDP socket matching *tuple*, optionally in a child
2331 * network namespace *netns*. The return value must be checked,
2332 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2334 * The *ctx* should point to the context of the program, such as
2335 * the skb or socket (depending on the hook in use). This is used
2336 * to determine the base network namespace for the lookup.
2338 * *tuple_size* must be one of:
2340 * **sizeof**\ (*tuple*\ **->ipv4**)
2341 * Look for an IPv4 socket.
2342 * **sizeof**\ (*tuple*\ **->ipv6**)
2343 * Look for an IPv6 socket.
2345 * If the *netns* is a negative signed 32-bit integer, then the
2346 * socket lookup table in the netns associated with the *ctx* will
2347 * will be used. For the TC hooks, this is the netns of the device
2348 * in the skb. For socket hooks, this is the netns of the socket.
2349 * If *netns* is any other signed 32-bit value greater than or
2350 * equal to zero then it specifies the ID of the netns relative to
2351 * the netns associated with the *ctx*. *netns* values beyond the
2352 * range of 32-bit integers are reserved for future use.
2354 * All values for *flags* are reserved for future usage, and must
2357 * This helper is available only if the kernel was compiled with
2358 * **CONFIG_NET** configuration option.
2360 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2361 * For sockets with reuseport option, the **struct bpf_sock**
2362 * result is from *reuse*\ **->socks**\ [] using the hash of the
2365 * int bpf_sk_release(struct bpf_sock *sock)
2367 * Release the reference held by *sock*. *sock* must be a
2368 * non-**NULL** pointer that was returned from
2369 * **bpf_sk_lookup_xxx**\ ().
2371 * 0 on success, or a negative error in case of failure.
2373 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2375 * Push an element *value* in *map*. *flags* is one of:
2378 * If the queue/stack is full, the oldest element is
2379 * removed to make room for this.
2381 * 0 on success, or a negative error in case of failure.
2383 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2385 * Pop an element from *map*.
2387 * 0 on success, or a negative error in case of failure.
2389 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2391 * Get an element from *map* without removing it.
2393 * 0 on success, or a negative error in case of failure.
2395 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2397 * For socket policies, insert *len* bytes into *msg* at offset
2400 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2401 * *msg* it may want to insert metadata or options into the *msg*.
2402 * This can later be read and used by any of the lower layer BPF
2405 * This helper may fail if under memory pressure (a malloc
2406 * fails) in these cases BPF programs will get an appropriate
2407 * error and BPF programs will need to handle them.
2409 * 0 on success, or a negative error in case of failure.
2411 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2413 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2414 * This may result in **ENOMEM** errors under certain situations if
2415 * an allocation and copy are required due to a full ring buffer.
2416 * However, the helper will try to avoid doing the allocation
2417 * if possible. Other errors can occur if input parameters are
2418 * invalid either due to *start* byte not being valid part of *msg*
2419 * payload and/or *pop* value being to large.
2421 * 0 on success, or a negative error in case of failure.
2423 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2425 * This helper is used in programs implementing IR decoding, to
2426 * report a successfully decoded pointer movement.
2428 * The *ctx* should point to the lirc sample as passed into
2431 * This helper is only available is the kernel was compiled with
2432 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2437 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2439 * Acquire a spinlock represented by the pointer *lock*, which is
2440 * stored as part of a value of a map. Taking the lock allows to
2441 * safely update the rest of the fields in that value. The
2442 * spinlock can (and must) later be released with a call to
2443 * **bpf_spin_unlock**\ (\ *lock*\ ).
2445 * Spinlocks in BPF programs come with a number of restrictions
2448 * * **bpf_spin_lock** objects are only allowed inside maps of
2449 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2450 * list could be extended in the future).
2451 * * BTF description of the map is mandatory.
2452 * * The BPF program can take ONE lock at a time, since taking two
2453 * or more could cause dead locks.
2454 * * Only one **struct bpf_spin_lock** is allowed per map element.
2455 * * When the lock is taken, calls (either BPF to BPF or helpers)
2457 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2458 * allowed inside a spinlock-ed region.
2459 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2460 * the lock, on all execution paths, before it returns.
2461 * * The BPF program can access **struct bpf_spin_lock** only via
2462 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2463 * helpers. Loading or storing data into the **struct
2464 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2465 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2466 * of the map value must be a struct and have **struct
2467 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2468 * Nested lock inside another struct is not allowed.
2469 * * The **struct bpf_spin_lock** *lock* field in a map value must
2470 * be aligned on a multiple of 4 bytes in that value.
2471 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2472 * the **bpf_spin_lock** field to user space.
2473 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2474 * a BPF program, do not update the **bpf_spin_lock** field.
2475 * * **bpf_spin_lock** cannot be on the stack or inside a
2476 * networking packet (it can only be inside of a map values).
2477 * * **bpf_spin_lock** is available to root only.
2478 * * Tracing programs and socket filter programs cannot use
2479 * **bpf_spin_lock**\ () due to insufficient preemption checks
2480 * (but this may change in the future).
2481 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2485 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2487 * Release the *lock* previously locked by a call to
2488 * **bpf_spin_lock**\ (\ *lock*\ ).
2492 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2494 * This helper gets a **struct bpf_sock** pointer such
2495 * that all the fields in this **bpf_sock** can be accessed.
2497 * A **struct bpf_sock** pointer on success, or **NULL** in
2500 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2502 * This helper gets a **struct bpf_tcp_sock** pointer from a
2503 * **struct bpf_sock** pointer.
2505 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2508 * int bpf_skb_ecn_set_ce(struct sk_buf *skb)
2510 * Set ECN (Explicit Congestion Notification) field of IP header
2511 * to **CE** (Congestion Encountered) if current value is **ECT**
2512 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2515 * 1 if the **CE** flag is set (either by the current helper call
2516 * or because it was already present), 0 if it is not set.
2518 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2520 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2521 * **bpf_sk_release**\ () is unnecessary and not allowed.
2523 * A **struct bpf_sock** pointer on success, or **NULL** in
2526 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2528 * Look for TCP socket matching *tuple*, optionally in a child
2529 * network namespace *netns*. The return value must be checked,
2530 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2532 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2533 * that it also returns timewait or request sockets. Use
2534 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2537 * This helper is available only if the kernel was compiled with
2538 * **CONFIG_NET** configuration option.
2540 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2541 * For sockets with reuseport option, the **struct bpf_sock**
2542 * result is from *reuse*\ **->socks**\ [] using the hash of the
2545 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2547 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2548 * the listening socket in *sk*.
2550 * *iph* points to the start of the IPv4 or IPv6 header, while
2551 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2552 * **sizeof**\ (**struct ip6hdr**).
2554 * *th* points to the start of the TCP header, while *th_len*
2555 * contains **sizeof**\ (**struct tcphdr**).
2558 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2561 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2563 * Get name of sysctl in /proc/sys/ and copy it into provided by
2564 * program buffer *buf* of size *buf_len*.
2566 * The buffer is always NUL terminated, unless it's zero-sized.
2568 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2569 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2570 * only (e.g. "tcp_mem").
2572 * Number of character copied (not including the trailing NUL).
2574 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2575 * truncated name in this case).
2577 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2579 * Get current value of sysctl as it is presented in /proc/sys
2580 * (incl. newline, etc), and copy it as a string into provided
2581 * by program buffer *buf* of size *buf_len*.
2583 * The whole value is copied, no matter what file position user
2584 * space issued e.g. sys_read at.
2586 * The buffer is always NUL terminated, unless it's zero-sized.
2588 * Number of character copied (not including the trailing NUL).
2590 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2591 * truncated name in this case).
2593 * **-EINVAL** if current value was unavailable, e.g. because
2594 * sysctl is uninitialized and read returns -EIO for it.
2596 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2598 * Get new value being written by user space to sysctl (before
2599 * the actual write happens) and copy it as a string into
2600 * provided by program buffer *buf* of size *buf_len*.
2602 * User space may write new value at file position > 0.
2604 * The buffer is always NUL terminated, unless it's zero-sized.
2606 * Number of character copied (not including the trailing NUL).
2608 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2609 * truncated name in this case).
2611 * **-EINVAL** if sysctl is being read.
2613 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2615 * Override new value being written by user space to sysctl with
2616 * value provided by program in buffer *buf* of size *buf_len*.
2618 * *buf* should contain a string in same form as provided by user
2619 * space on sysctl write.
2621 * User space may write new value at file position > 0. To override
2622 * the whole sysctl value file position should be set to zero.
2626 * **-E2BIG** if the *buf_len* is too big.
2628 * **-EINVAL** if sysctl is being read.
2630 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2632 * Convert the initial part of the string from buffer *buf* of
2633 * size *buf_len* to a long integer according to the given base
2634 * and save the result in *res*.
2636 * The string may begin with an arbitrary amount of white space
2637 * (as determined by **isspace**\ (3)) followed by a single
2638 * optional '**-**' sign.
2640 * Five least significant bits of *flags* encode base, other bits
2641 * are currently unused.
2643 * Base must be either 8, 10, 16 or 0 to detect it automatically
2644 * similar to user space **strtol**\ (3).
2646 * Number of characters consumed on success. Must be positive but
2647 * no more than *buf_len*.
2649 * **-EINVAL** if no valid digits were found or unsupported base
2652 * **-ERANGE** if resulting value was out of range.
2654 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2656 * Convert the initial part of the string from buffer *buf* of
2657 * size *buf_len* to an unsigned long integer according to the
2658 * given base and save the result in *res*.
2660 * The string may begin with an arbitrary amount of white space
2661 * (as determined by **isspace**\ (3)).
2663 * Five least significant bits of *flags* encode base, other bits
2664 * are currently unused.
2666 * Base must be either 8, 10, 16 or 0 to detect it automatically
2667 * similar to user space **strtoul**\ (3).
2669 * Number of characters consumed on success. Must be positive but
2670 * no more than *buf_len*.
2672 * **-EINVAL** if no valid digits were found or unsupported base
2675 * **-ERANGE** if resulting value was out of range.
2677 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2679 * Get a bpf-local-storage from a *sk*.
2681 * Logically, it could be thought of getting the value from
2682 * a *map* with *sk* as the **key**. From this
2683 * perspective, the usage is not much different from
2684 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2685 * helper enforces the key must be a full socket and the map must
2686 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2688 * Underneath, the value is stored locally at *sk* instead of
2689 * the *map*. The *map* is used as the bpf-local-storage
2690 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2691 * searched against all bpf-local-storages residing at *sk*.
2693 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2694 * used such that a new bpf-local-storage will be
2695 * created if one does not exist. *value* can be used
2696 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2697 * the initial value of a bpf-local-storage. If *value* is
2698 * **NULL**, the new bpf-local-storage will be zero initialized.
2700 * A bpf-local-storage pointer is returned on success.
2702 * **NULL** if not found or there was an error in adding
2703 * a new bpf-local-storage.
2705 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2707 * Delete a bpf-local-storage from a *sk*.
2711 * **-ENOENT** if the bpf-local-storage cannot be found.
2713 * int bpf_send_signal(u32 sig)
2715 * Send signal *sig* to the current task.
2717 * 0 on success or successfully queued.
2719 * **-EBUSY** if work queue under nmi is full.
2721 * **-EINVAL** if *sig* is invalid.
2723 * **-EPERM** if no permission to send the *sig*.
2725 * **-EAGAIN** if bpf program can try again.
2727 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2729 * Try to issue a SYN cookie for the packet with corresponding
2730 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2732 * *iph* points to the start of the IPv4 or IPv6 header, while
2733 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2734 * **sizeof**\ (**struct ip6hdr**).
2736 * *th* points to the start of the TCP header, while *th_len*
2737 * contains the length of the TCP header.
2740 * On success, lower 32 bits hold the generated SYN cookie in
2741 * followed by 16 bits which hold the MSS value for that cookie,
2742 * and the top 16 bits are unused.
2744 * On failure, the returned value is one of the following:
2746 * **-EINVAL** SYN cookie cannot be issued due to error
2748 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2750 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2752 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2754 #define __BPF_FUNC_MAPPER(FN) \
2756 FN(map_lookup_elem), \
2757 FN(map_update_elem), \
2758 FN(map_delete_elem), \
2762 FN(get_prandom_u32), \
2763 FN(get_smp_processor_id), \
2764 FN(skb_store_bytes), \
2765 FN(l3_csum_replace), \
2766 FN(l4_csum_replace), \
2768 FN(clone_redirect), \
2769 FN(get_current_pid_tgid), \
2770 FN(get_current_uid_gid), \
2771 FN(get_current_comm), \
2772 FN(get_cgroup_classid), \
2773 FN(skb_vlan_push), \
2775 FN(skb_get_tunnel_key), \
2776 FN(skb_set_tunnel_key), \
2777 FN(perf_event_read), \
2779 FN(get_route_realm), \
2780 FN(perf_event_output), \
2781 FN(skb_load_bytes), \
2784 FN(skb_get_tunnel_opt), \
2785 FN(skb_set_tunnel_opt), \
2786 FN(skb_change_proto), \
2787 FN(skb_change_type), \
2788 FN(skb_under_cgroup), \
2789 FN(get_hash_recalc), \
2790 FN(get_current_task), \
2791 FN(probe_write_user), \
2792 FN(current_task_under_cgroup), \
2793 FN(skb_change_tail), \
2794 FN(skb_pull_data), \
2796 FN(set_hash_invalid), \
2797 FN(get_numa_node_id), \
2798 FN(skb_change_head), \
2799 FN(xdp_adjust_head), \
2800 FN(probe_read_str), \
2801 FN(get_socket_cookie), \
2802 FN(get_socket_uid), \
2805 FN(skb_adjust_room), \
2807 FN(sk_redirect_map), \
2808 FN(sock_map_update), \
2809 FN(xdp_adjust_meta), \
2810 FN(perf_event_read_value), \
2811 FN(perf_prog_read_value), \
2813 FN(override_return), \
2814 FN(sock_ops_cb_flags_set), \
2815 FN(msg_redirect_map), \
2816 FN(msg_apply_bytes), \
2817 FN(msg_cork_bytes), \
2818 FN(msg_pull_data), \
2820 FN(xdp_adjust_tail), \
2821 FN(skb_get_xfrm_state), \
2823 FN(skb_load_bytes_relative), \
2825 FN(sock_hash_update), \
2826 FN(msg_redirect_hash), \
2827 FN(sk_redirect_hash), \
2828 FN(lwt_push_encap), \
2829 FN(lwt_seg6_store_bytes), \
2830 FN(lwt_seg6_adjust_srh), \
2831 FN(lwt_seg6_action), \
2834 FN(skb_cgroup_id), \
2835 FN(get_current_cgroup_id), \
2836 FN(get_local_storage), \
2837 FN(sk_select_reuseport), \
2838 FN(skb_ancestor_cgroup_id), \
2839 FN(sk_lookup_tcp), \
2840 FN(sk_lookup_udp), \
2842 FN(map_push_elem), \
2844 FN(map_peek_elem), \
2845 FN(msg_push_data), \
2847 FN(rc_pointer_rel), \
2852 FN(skb_ecn_set_ce), \
2853 FN(get_listener_sock), \
2854 FN(skc_lookup_tcp), \
2855 FN(tcp_check_syncookie), \
2856 FN(sysctl_get_name), \
2857 FN(sysctl_get_current_value), \
2858 FN(sysctl_get_new_value), \
2859 FN(sysctl_set_new_value), \
2862 FN(sk_storage_get), \
2863 FN(sk_storage_delete), \
2865 FN(tcp_gen_syncookie),
2867 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2868 * function eBPF program intends to call
2870 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2872 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2875 #undef __BPF_ENUM_FN
2877 /* All flags used by eBPF helper functions, placed here. */
2879 /* BPF_FUNC_skb_store_bytes flags. */
2880 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2881 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2883 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2884 * First 4 bits are for passing the header field size.
2886 #define BPF_F_HDR_FIELD_MASK 0xfULL
2888 /* BPF_FUNC_l4_csum_replace flags. */
2889 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2890 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2891 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2893 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2894 #define BPF_F_INGRESS (1ULL << 0)
2896 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2897 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2899 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2900 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2901 #define BPF_F_USER_STACK (1ULL << 8)
2902 /* flags used by BPF_FUNC_get_stackid only. */
2903 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2904 #define BPF_F_REUSE_STACKID (1ULL << 10)
2905 /* flags used by BPF_FUNC_get_stack only. */
2906 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2908 /* BPF_FUNC_skb_set_tunnel_key flags. */
2909 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2910 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2911 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2913 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2914 * BPF_FUNC_perf_event_read_value flags.
2916 #define BPF_F_INDEX_MASK 0xffffffffULL
2917 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2918 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2919 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2921 /* Current network namespace */
2922 #define BPF_F_CURRENT_NETNS (-1L)
2924 /* BPF_FUNC_skb_adjust_room flags. */
2925 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
2927 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
2928 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
2930 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
2931 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
2932 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
2933 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
2934 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
2935 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
2936 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
2938 /* BPF_FUNC_sysctl_get_name flags. */
2939 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
2941 /* BPF_FUNC_sk_storage_get flags */
2942 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0)
2944 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2945 enum bpf_adj_room_mode {
2950 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2951 enum bpf_hdr_start_off {
2956 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2957 enum bpf_lwt_encap_mode {
2959 BPF_LWT_ENCAP_SEG6_INLINE,
2963 #define __bpf_md_ptr(type, name) \
2967 } __attribute__((aligned(8)))
2969 /* user accessible mirror of in-kernel sk_buff.
2970 * new fields can only be added to the end of this structure
2976 __u32 queue_mapping;
2982 __u32 ingress_ifindex;
2992 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2994 __u32 remote_ip4; /* Stored in network byte order */
2995 __u32 local_ip4; /* Stored in network byte order */
2996 __u32 remote_ip6[4]; /* Stored in network byte order */
2997 __u32 local_ip6[4]; /* Stored in network byte order */
2998 __u32 remote_port; /* Stored in network byte order */
2999 __u32 local_port; /* stored in host byte order */
3003 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
3007 __bpf_md_ptr(struct bpf_sock *, sk);
3010 struct bpf_tunnel_key {
3014 __u32 remote_ipv6[4];
3018 __u16 tunnel_ext; /* Padding, future use. */
3022 /* user accessible mirror of in-kernel xfrm_state.
3023 * new fields can only be added to the end of this structure
3025 struct bpf_xfrm_state {
3027 __u32 spi; /* Stored in network byte order */
3029 __u16 ext; /* Padding, future use. */
3031 __u32 remote_ipv4; /* Stored in network byte order */
3032 __u32 remote_ipv6[4]; /* Stored in network byte order */
3036 /* Generic BPF return codes which all BPF program types may support.
3037 * The values are binary compatible with their TC_ACT_* counter-part to
3038 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3041 * XDP is handled seprately, see XDP_*.
3049 /* >127 are reserved for prog type specific return codes.
3051 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3052 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3053 * changed and should be routed based on its new L3 header.
3054 * (This is an L3 redirect, as opposed to L2 redirect
3055 * represented by BPF_REDIRECT above).
3057 BPF_LWT_REROUTE = 128,
3067 /* IP address also allows 1 and 2 bytes access */
3070 __u32 src_port; /* host byte order */
3071 __u32 dst_port; /* network byte order */
3077 struct bpf_tcp_sock {
3078 __u32 snd_cwnd; /* Sending congestion window */
3079 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3081 __u32 snd_ssthresh; /* Slow start size threshold */
3082 __u32 rcv_nxt; /* What we want to receive next */
3083 __u32 snd_nxt; /* Next sequence we send */
3084 __u32 snd_una; /* First byte we want an ack for */
3085 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3086 __u32 ecn_flags; /* ECN status bits. */
3087 __u32 rate_delivered; /* saved rate sample: packets delivered */
3088 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3089 __u32 packets_out; /* Packets which are "in flight" */
3090 __u32 retrans_out; /* Retransmitted packets out */
3091 __u32 total_retrans; /* Total retransmits for entire connection */
3092 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3093 * total number of segments in.
3095 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3096 * total number of data segments in.
3098 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3099 * The total number of segments sent.
3101 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3102 * total number of data segments sent.
3104 __u32 lost_out; /* Lost packets */
3105 __u32 sacked_out; /* SACK'd packets */
3106 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3107 * sum(delta(rcv_nxt)), or how many bytes
3110 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3111 * sum(delta(snd_una)), or how many bytes
3114 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3115 * total number of DSACK blocks received
3117 __u32 delivered; /* Total data packets delivered incl. rexmits */
3118 __u32 delivered_ce; /* Like the above but only ECE marked packets */
3119 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
3122 struct bpf_sock_tuple {
3139 struct bpf_xdp_sock {
3143 #define XDP_PACKET_HEADROOM 256
3145 /* User return codes for XDP prog type.
3146 * A valid XDP program must return one of these defined values. All other
3147 * return codes are reserved for future use. Unknown return codes will
3148 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3158 /* user accessible metadata for XDP packet hook
3159 * new fields must be added to the end of this structure
3165 /* Below access go through struct xdp_rxq_info */
3166 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3167 __u32 rx_queue_index; /* rxq->queue_index */
3175 /* user accessible metadata for SK_MSG packet hook, new fields must
3176 * be added to the end of this structure
3179 __bpf_md_ptr(void *, data);
3180 __bpf_md_ptr(void *, data_end);
3183 __u32 remote_ip4; /* Stored in network byte order */
3184 __u32 local_ip4; /* Stored in network byte order */
3185 __u32 remote_ip6[4]; /* Stored in network byte order */
3186 __u32 local_ip6[4]; /* Stored in network byte order */
3187 __u32 remote_port; /* Stored in network byte order */
3188 __u32 local_port; /* stored in host byte order */
3189 __u32 size; /* Total size of sk_msg */
3192 struct sk_reuseport_md {
3194 * Start of directly accessible data. It begins from
3195 * the tcp/udp header.
3197 __bpf_md_ptr(void *, data);
3198 /* End of directly accessible data */
3199 __bpf_md_ptr(void *, data_end);
3201 * Total length of packet (starting from the tcp/udp header).
3202 * Note that the directly accessible bytes (data_end - data)
3203 * could be less than this "len". Those bytes could be
3204 * indirectly read by a helper "bpf_skb_load_bytes()".
3208 * Eth protocol in the mac header (network byte order). e.g.
3209 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3212 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3213 __u32 bind_inany; /* Is sock bound to an INANY address? */
3214 __u32 hash; /* A hash of the packet 4 tuples */
3217 #define BPF_TAG_SIZE 8
3219 struct bpf_prog_info {
3222 __u8 tag[BPF_TAG_SIZE];
3223 __u32 jited_prog_len;
3224 __u32 xlated_prog_len;
3225 __aligned_u64 jited_prog_insns;
3226 __aligned_u64 xlated_prog_insns;
3227 __u64 load_time; /* ns since boottime */
3228 __u32 created_by_uid;
3230 __aligned_u64 map_ids;
3231 char name[BPF_OBJ_NAME_LEN];
3233 __u32 gpl_compatible:1;
3234 __u32 :31; /* alignment pad */
3237 __u32 nr_jited_ksyms;
3238 __u32 nr_jited_func_lens;
3239 __aligned_u64 jited_ksyms;
3240 __aligned_u64 jited_func_lens;
3242 __u32 func_info_rec_size;
3243 __aligned_u64 func_info;
3246 __aligned_u64 line_info;
3247 __aligned_u64 jited_line_info;
3248 __u32 nr_jited_line_info;
3249 __u32 line_info_rec_size;
3250 __u32 jited_line_info_rec_size;
3252 __aligned_u64 prog_tags;
3255 } __attribute__((aligned(8)));
3257 struct bpf_map_info {
3264 char name[BPF_OBJ_NAME_LEN];
3270 __u32 btf_key_type_id;
3271 __u32 btf_value_type_id;
3272 } __attribute__((aligned(8)));
3274 struct bpf_btf_info {
3278 } __attribute__((aligned(8)));
3280 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3281 * by user and intended to be used by socket (e.g. to bind to, depends on
3282 * attach attach type).
3284 struct bpf_sock_addr {
3285 __u32 user_family; /* Allows 4-byte read, but no write. */
3286 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3287 * Stored in network byte order.
3289 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3290 * Stored in network byte order.
3292 __u32 user_port; /* Allows 4-byte read and write.
3293 * Stored in network byte order
3295 __u32 family; /* Allows 4-byte read, but no write */
3296 __u32 type; /* Allows 4-byte read, but no write */
3297 __u32 protocol; /* Allows 4-byte read, but no write */
3298 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3299 * Stored in network byte order.
3301 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3302 * Stored in network byte order.
3304 __bpf_md_ptr(struct bpf_sock *, sk);
3307 /* User bpf_sock_ops struct to access socket values and specify request ops
3308 * and their replies.
3309 * Some of this fields are in network (bigendian) byte order and may need
3310 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3311 * New fields can only be added at the end of this structure
3313 struct bpf_sock_ops {
3316 __u32 args[4]; /* Optionally passed to bpf program */
3317 __u32 reply; /* Returned by bpf program */
3318 __u32 replylong[4]; /* Optionally returned by bpf prog */
3321 __u32 remote_ip4; /* Stored in network byte order */
3322 __u32 local_ip4; /* Stored in network byte order */
3323 __u32 remote_ip6[4]; /* Stored in network byte order */
3324 __u32 local_ip6[4]; /* Stored in network byte order */
3325 __u32 remote_port; /* Stored in network byte order */
3326 __u32 local_port; /* stored in host byte order */
3327 __u32 is_fullsock; /* Some TCP fields are only valid if
3328 * there is a full socket. If not, the
3329 * fields read as zero.
3332 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
3333 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3342 __u32 rate_delivered;
3343 __u32 rate_interval_us;
3346 __u32 total_retrans;
3350 __u32 data_segs_out;
3354 __u64 bytes_received;
3356 __bpf_md_ptr(struct bpf_sock *, sk);
3359 /* Definitions for bpf_sock_ops_cb_flags */
3360 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
3361 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
3362 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3363 #define BPF_SOCK_OPS_RTT_CB_FLAG (1<<3)
3364 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0xF /* Mask of all currently
3365 * supported cb flags
3368 /* List of known BPF sock_ops operators.
3369 * New entries can only be added at the end
3373 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3374 * -1 if default value should be used
3376 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3377 * window (in packets) or -1 if default
3378 * value should be used
3380 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3381 * active connection is initialized
3383 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3384 * active connection is
3387 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3388 * passive connection is
3391 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3394 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3395 * based on the path and may be
3396 * dependent on the congestion control
3397 * algorithm. In general it indicates
3398 * a congestion threshold. RTTs above
3399 * this indicate congestion
3401 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3402 * Arg1: value of icsk_retransmits
3403 * Arg2: value of icsk_rto
3404 * Arg3: whether RTO has expired
3406 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3407 * Arg1: sequence number of 1st byte
3409 * Arg3: return value of
3410 * tcp_transmit_skb (0 => success)
3412 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3416 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3417 * socket transition to LISTEN state.
3419 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3423 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3424 * changes between the TCP and BPF versions. Ideally this should never happen.
3425 * If it does, we need to add code to convert them before calling
3426 * the BPF sock_ops function.
3429 BPF_TCP_ESTABLISHED = 1,
3439 BPF_TCP_CLOSING, /* Now a valid state */
3440 BPF_TCP_NEW_SYN_RECV,
3442 BPF_TCP_MAX_STATES /* Leave at the end! */
3445 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3446 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3448 struct bpf_perf_event_value {
3454 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3455 #define BPF_DEVCG_ACC_READ (1ULL << 1)
3456 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3458 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3459 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3461 struct bpf_cgroup_dev_ctx {
3462 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3468 struct bpf_raw_tracepoint_args {
3472 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3473 * OUTPUT: Do lookup from egress perspective; default is ingress
3475 #define BPF_FIB_LOOKUP_DIRECT (1U << 0)
3476 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
3479 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3480 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3481 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3482 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3483 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3484 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3485 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3486 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3487 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3490 struct bpf_fib_lookup {
3491 /* input: network family for lookup (AF_INET, AF_INET6)
3492 * output: network family of egress nexthop
3496 /* set if lookup is to consider L4 data - e.g., FIB rules */
3501 /* total length of packet from network header - used for MTU check */
3504 /* input: L3 device index for lookup
3505 * output: device index from FIB lookup
3510 /* inputs to lookup */
3511 __u8 tos; /* AF_INET */
3512 __be32 flowinfo; /* AF_INET6, flow_label + priority */
3514 /* output: metric of fib result (IPv4/IPv6 only) */
3520 __u32 ipv6_src[4]; /* in6_addr; network order */
3523 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3524 * network header. output: bpf_fib_lookup sets to gateway address
3525 * if FIB lookup returns gateway route
3529 __u32 ipv6_dst[4]; /* in6_addr; network order */
3533 __be16 h_vlan_proto;
3535 __u8 smac[6]; /* ETH_ALEN */
3536 __u8 dmac[6]; /* ETH_ALEN */
3539 enum bpf_task_fd_type {
3540 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3541 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3542 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3543 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3544 BPF_FD_TYPE_UPROBE, /* filename + offset */
3545 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3548 #define BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG (1U << 0)
3549 #define BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL (1U << 1)
3550 #define BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP (1U << 2)
3552 struct bpf_flow_keys {
3555 __u16 addr_proto; /* ETH_P_* of valid addrs */
3569 __u32 ipv6_src[4]; /* in6_addr; network order */
3570 __u32 ipv6_dst[4]; /* in6_addr; network order */
3577 struct bpf_func_info {
3582 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3583 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3585 struct bpf_line_info {
3587 __u32 file_name_off;
3592 struct bpf_spin_lock {
3597 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3598 * Allows 1,2,4-byte read, but no write.
3600 __u32 file_pos; /* Sysctl file position to read from, write to.
3601 * Allows 1,2,4-byte read an 4-byte write.
3605 struct bpf_sockopt {
3606 __bpf_md_ptr(struct bpf_sock *, sk);
3607 __bpf_md_ptr(void *, optval);
3608 __bpf_md_ptr(void *, optval_end);
3616 #endif /* _UAPI__LINUX_BPF_H__ */