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[]; /* 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,
110 BPF_MAP_LOOKUP_BATCH,
111 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
112 BPF_MAP_UPDATE_BATCH,
113 BPF_MAP_DELETE_BATCH,
116 BPF_LINK_GET_FD_BY_ID,
117 BPF_LINK_GET_NEXT_ID,
126 BPF_MAP_TYPE_PROG_ARRAY,
127 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
128 BPF_MAP_TYPE_PERCPU_HASH,
129 BPF_MAP_TYPE_PERCPU_ARRAY,
130 BPF_MAP_TYPE_STACK_TRACE,
131 BPF_MAP_TYPE_CGROUP_ARRAY,
132 BPF_MAP_TYPE_LRU_HASH,
133 BPF_MAP_TYPE_LRU_PERCPU_HASH,
134 BPF_MAP_TYPE_LPM_TRIE,
135 BPF_MAP_TYPE_ARRAY_OF_MAPS,
136 BPF_MAP_TYPE_HASH_OF_MAPS,
138 BPF_MAP_TYPE_SOCKMAP,
141 BPF_MAP_TYPE_SOCKHASH,
142 BPF_MAP_TYPE_CGROUP_STORAGE,
143 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
144 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
147 BPF_MAP_TYPE_SK_STORAGE,
148 BPF_MAP_TYPE_DEVMAP_HASH,
149 BPF_MAP_TYPE_STRUCT_OPS,
152 /* Note that tracing related programs such as
153 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
154 * are not subject to a stable API since kernel internal data
155 * structures can change from release to release and may
156 * therefore break existing tracing BPF programs. Tracing BPF
157 * programs correspond to /a/ specific kernel which is to be
158 * analyzed, and not /a/ specific kernel /and/ all future ones.
161 BPF_PROG_TYPE_UNSPEC,
162 BPF_PROG_TYPE_SOCKET_FILTER,
163 BPF_PROG_TYPE_KPROBE,
164 BPF_PROG_TYPE_SCHED_CLS,
165 BPF_PROG_TYPE_SCHED_ACT,
166 BPF_PROG_TYPE_TRACEPOINT,
168 BPF_PROG_TYPE_PERF_EVENT,
169 BPF_PROG_TYPE_CGROUP_SKB,
170 BPF_PROG_TYPE_CGROUP_SOCK,
171 BPF_PROG_TYPE_LWT_IN,
172 BPF_PROG_TYPE_LWT_OUT,
173 BPF_PROG_TYPE_LWT_XMIT,
174 BPF_PROG_TYPE_SOCK_OPS,
175 BPF_PROG_TYPE_SK_SKB,
176 BPF_PROG_TYPE_CGROUP_DEVICE,
177 BPF_PROG_TYPE_SK_MSG,
178 BPF_PROG_TYPE_RAW_TRACEPOINT,
179 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
180 BPF_PROG_TYPE_LWT_SEG6LOCAL,
181 BPF_PROG_TYPE_LIRC_MODE2,
182 BPF_PROG_TYPE_SK_REUSEPORT,
183 BPF_PROG_TYPE_FLOW_DISSECTOR,
184 BPF_PROG_TYPE_CGROUP_SYSCTL,
185 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
186 BPF_PROG_TYPE_CGROUP_SOCKOPT,
187 BPF_PROG_TYPE_TRACING,
188 BPF_PROG_TYPE_STRUCT_OPS,
193 enum bpf_attach_type {
194 BPF_CGROUP_INET_INGRESS,
195 BPF_CGROUP_INET_EGRESS,
196 BPF_CGROUP_INET_SOCK_CREATE,
198 BPF_SK_SKB_STREAM_PARSER,
199 BPF_SK_SKB_STREAM_VERDICT,
202 BPF_CGROUP_INET4_BIND,
203 BPF_CGROUP_INET6_BIND,
204 BPF_CGROUP_INET4_CONNECT,
205 BPF_CGROUP_INET6_CONNECT,
206 BPF_CGROUP_INET4_POST_BIND,
207 BPF_CGROUP_INET6_POST_BIND,
208 BPF_CGROUP_UDP4_SENDMSG,
209 BPF_CGROUP_UDP6_SENDMSG,
213 BPF_CGROUP_UDP4_RECVMSG,
214 BPF_CGROUP_UDP6_RECVMSG,
215 BPF_CGROUP_GETSOCKOPT,
216 BPF_CGROUP_SETSOCKOPT,
223 __MAX_BPF_ATTACH_TYPE
226 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
229 BPF_LINK_TYPE_UNSPEC = 0,
230 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
231 BPF_LINK_TYPE_TRACING = 2,
232 BPF_LINK_TYPE_CGROUP = 3,
233 BPF_LINK_TYPE_ITER = 4,
238 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
240 * NONE(default): No further bpf programs allowed in the subtree.
242 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
243 * the program in this cgroup yields to sub-cgroup program.
245 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
246 * that cgroup program gets run in addition to the program in this cgroup.
248 * Only one program is allowed to be attached to a cgroup with
249 * NONE or BPF_F_ALLOW_OVERRIDE flag.
250 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
251 * release old program and attach the new one. Attach flags has to match.
253 * Multiple programs are allowed to be attached to a cgroup with
254 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
255 * (those that were attached first, run first)
256 * The programs of sub-cgroup are executed first, then programs of
257 * this cgroup and then programs of parent cgroup.
258 * When children program makes decision (like picking TCP CA or sock bind)
259 * parent program has a chance to override it.
261 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
262 * programs for a cgroup. Though it's possible to replace an old program at
263 * any position by also specifying BPF_F_REPLACE flag and position itself in
264 * replace_bpf_fd attribute. Old program at this position will be released.
266 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
267 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
269 * cgrp1 (MULTI progs A, B) ->
270 * cgrp2 (OVERRIDE prog C) ->
271 * cgrp3 (MULTI prog D) ->
272 * cgrp4 (OVERRIDE prog E) ->
273 * cgrp5 (NONE prog F)
274 * the event in cgrp5 triggers execution of F,D,A,B in that order.
275 * if prog F is detached, the execution is E,D,A,B
276 * if prog F and D are detached, the execution is E,A,B
277 * if prog F, E and D are detached, the execution is C,A,B
279 * All eligible programs are executed regardless of return code from
282 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
283 #define BPF_F_ALLOW_MULTI (1U << 1)
284 #define BPF_F_REPLACE (1U << 2)
286 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
287 * verifier will perform strict alignment checking as if the kernel
288 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
289 * and NET_IP_ALIGN defined to 2.
291 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
293 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
294 * verifier will allow any alignment whatsoever. On platforms
295 * with strict alignment requirements for loads ands stores (such
296 * as sparc and mips) the verifier validates that all loads and
297 * stores provably follow this requirement. This flag turns that
298 * checking and enforcement off.
300 * It is mostly used for testing when we want to validate the
301 * context and memory access aspects of the verifier, but because
302 * of an unaligned access the alignment check would trigger before
303 * the one we are interested in.
305 #define BPF_F_ANY_ALIGNMENT (1U << 1)
307 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
308 * Verifier does sub-register def/use analysis and identifies instructions whose
309 * def only matters for low 32-bit, high 32-bit is never referenced later
310 * through implicit zero extension. Therefore verifier notifies JIT back-ends
311 * that it is safe to ignore clearing high 32-bit for these instructions. This
312 * saves some back-ends a lot of code-gen. However such optimization is not
313 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
314 * hence hasn't used verifier's analysis result. But, we really want to have a
315 * way to be able to verify the correctness of the described optimization on
316 * x86_64 on which testsuites are frequently exercised.
318 * So, this flag is introduced. Once it is set, verifier will randomize high
319 * 32-bit for those instructions who has been identified as safe to ignore them.
320 * Then, if verifier is not doing correct analysis, such randomization will
321 * regress tests to expose bugs.
323 #define BPF_F_TEST_RND_HI32 (1U << 2)
325 /* The verifier internal test flag. Behavior is undefined */
326 #define BPF_F_TEST_STATE_FREQ (1U << 3)
328 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
331 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
332 * insn[0].imm: map fd map fd
333 * insn[1].imm: 0 offset into value
336 * ldimm64 rewrite: address of map address of map[0]+offset
337 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
339 #define BPF_PSEUDO_MAP_FD 1
340 #define BPF_PSEUDO_MAP_VALUE 2
342 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
343 * offset to another bpf function
345 #define BPF_PSEUDO_CALL 1
347 /* flags for BPF_MAP_UPDATE_ELEM command */
349 BPF_ANY = 0, /* create new element or update existing */
350 BPF_NOEXIST = 1, /* create new element if it didn't exist */
351 BPF_EXIST = 2, /* update existing element */
352 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
355 /* flags for BPF_MAP_CREATE command */
357 BPF_F_NO_PREALLOC = (1U << 0),
358 /* Instead of having one common LRU list in the
359 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
360 * which can scale and perform better.
361 * Note, the LRU nodes (including free nodes) cannot be moved
362 * across different LRU lists.
364 BPF_F_NO_COMMON_LRU = (1U << 1),
365 /* Specify numa node during map creation */
366 BPF_F_NUMA_NODE = (1U << 2),
368 /* Flags for accessing BPF object from syscall side. */
369 BPF_F_RDONLY = (1U << 3),
370 BPF_F_WRONLY = (1U << 4),
372 /* Flag for stack_map, store build_id+offset instead of pointer */
373 BPF_F_STACK_BUILD_ID = (1U << 5),
375 /* Zero-initialize hash function seed. This should only be used for testing. */
376 BPF_F_ZERO_SEED = (1U << 6),
378 /* Flags for accessing BPF object from program side. */
379 BPF_F_RDONLY_PROG = (1U << 7),
380 BPF_F_WRONLY_PROG = (1U << 8),
382 /* Clone map from listener for newly accepted socket */
383 BPF_F_CLONE = (1U << 9),
385 /* Enable memory-mapping BPF map */
386 BPF_F_MMAPABLE = (1U << 10),
389 /* Flags for BPF_PROG_QUERY. */
391 /* Query effective (directly attached + inherited from ancestor cgroups)
392 * programs that will be executed for events within a cgroup.
393 * attach_flags with this flag are returned only for directly attached programs.
395 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
397 /* type for BPF_ENABLE_STATS */
398 enum bpf_stats_type {
399 /* enabled run_time_ns and run_cnt */
400 BPF_STATS_RUN_TIME = 0,
403 enum bpf_stack_build_id_status {
404 /* user space need an empty entry to identify end of a trace */
405 BPF_STACK_BUILD_ID_EMPTY = 0,
406 /* with valid build_id and offset */
407 BPF_STACK_BUILD_ID_VALID = 1,
408 /* couldn't get build_id, fallback to ip */
409 BPF_STACK_BUILD_ID_IP = 2,
412 #define BPF_BUILD_ID_SIZE 20
413 struct bpf_stack_build_id {
415 unsigned char build_id[BPF_BUILD_ID_SIZE];
422 #define BPF_OBJ_NAME_LEN 16U
425 struct { /* anonymous struct used by BPF_MAP_CREATE command */
426 __u32 map_type; /* one of enum bpf_map_type */
427 __u32 key_size; /* size of key in bytes */
428 __u32 value_size; /* size of value in bytes */
429 __u32 max_entries; /* max number of entries in a map */
430 __u32 map_flags; /* BPF_MAP_CREATE related
431 * flags defined above.
433 __u32 inner_map_fd; /* fd pointing to the inner map */
434 __u32 numa_node; /* numa node (effective only if
435 * BPF_F_NUMA_NODE is set).
437 char map_name[BPF_OBJ_NAME_LEN];
438 __u32 map_ifindex; /* ifindex of netdev to create on */
439 __u32 btf_fd; /* fd pointing to a BTF type data */
440 __u32 btf_key_type_id; /* BTF type_id of the key */
441 __u32 btf_value_type_id; /* BTF type_id of the value */
442 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
443 * struct stored as the
448 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
453 __aligned_u64 next_key;
458 struct { /* struct used by BPF_MAP_*_BATCH commands */
459 __aligned_u64 in_batch; /* start batch,
460 * NULL to start from beginning
462 __aligned_u64 out_batch; /* output: next start batch */
464 __aligned_u64 values;
465 __u32 count; /* input/output:
466 * input: # of key/value
468 * output: # of filled elements
475 struct { /* anonymous struct used by BPF_PROG_LOAD command */
476 __u32 prog_type; /* one of enum bpf_prog_type */
479 __aligned_u64 license;
480 __u32 log_level; /* verbosity level of verifier */
481 __u32 log_size; /* size of user buffer */
482 __aligned_u64 log_buf; /* user supplied buffer */
483 __u32 kern_version; /* not used */
485 char prog_name[BPF_OBJ_NAME_LEN];
486 __u32 prog_ifindex; /* ifindex of netdev to prep for */
487 /* For some prog types expected attach type must be known at
488 * load time to verify attach type specific parts of prog
489 * (context accesses, allowed helpers, etc).
491 __u32 expected_attach_type;
492 __u32 prog_btf_fd; /* fd pointing to BTF type data */
493 __u32 func_info_rec_size; /* userspace bpf_func_info size */
494 __aligned_u64 func_info; /* func info */
495 __u32 func_info_cnt; /* number of bpf_func_info records */
496 __u32 line_info_rec_size; /* userspace bpf_line_info size */
497 __aligned_u64 line_info; /* line info */
498 __u32 line_info_cnt; /* number of bpf_line_info records */
499 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
500 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
503 struct { /* anonymous struct used by BPF_OBJ_* commands */
504 __aligned_u64 pathname;
509 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
510 __u32 target_fd; /* container object to attach to */
511 __u32 attach_bpf_fd; /* eBPF program to attach */
514 __u32 replace_bpf_fd; /* previously attached eBPF
515 * program to replace if
516 * BPF_F_REPLACE is used
520 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
523 __u32 data_size_in; /* input: len of data_in */
524 __u32 data_size_out; /* input/output: len of data_out
525 * returns ENOSPC if data_out
528 __aligned_u64 data_in;
529 __aligned_u64 data_out;
532 __u32 ctx_size_in; /* input: len of ctx_in */
533 __u32 ctx_size_out; /* input/output: len of ctx_out
534 * returns ENOSPC if ctx_out
537 __aligned_u64 ctx_in;
538 __aligned_u64 ctx_out;
541 struct { /* anonymous struct used by BPF_*_GET_*_ID */
553 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
559 struct { /* anonymous struct used by BPF_PROG_QUERY command */
560 __u32 target_fd; /* container object to query */
564 __aligned_u64 prog_ids;
568 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
573 struct { /* anonymous struct for BPF_BTF_LOAD */
575 __aligned_u64 btf_log_buf;
582 __u32 pid; /* input: pid */
583 __u32 fd; /* input: fd */
584 __u32 flags; /* input: flags */
585 __u32 buf_len; /* input/output: buf len */
586 __aligned_u64 buf; /* input/output:
587 * tp_name for tracepoint
589 * filename for uprobe
591 __u32 prog_id; /* output: prod_id */
592 __u32 fd_type; /* output: BPF_FD_TYPE_* */
593 __u64 probe_offset; /* output: probe_offset */
594 __u64 probe_addr; /* output: probe_addr */
597 struct { /* struct used by BPF_LINK_CREATE command */
598 __u32 prog_fd; /* eBPF program to attach */
599 __u32 target_fd; /* object to attach to */
600 __u32 attach_type; /* attach type */
601 __u32 flags; /* extra flags */
604 struct { /* struct used by BPF_LINK_UPDATE command */
605 __u32 link_fd; /* link fd */
606 /* new program fd to update link with */
608 __u32 flags; /* extra flags */
609 /* expected link's program fd; is specified only if
610 * BPF_F_REPLACE flag is set in flags */
614 struct { /* struct used by BPF_ENABLE_STATS command */
618 struct { /* struct used by BPF_ITER_CREATE command */
623 } __attribute__((aligned(8)));
625 /* The description below is an attempt at providing documentation to eBPF
626 * developers about the multiple available eBPF helper functions. It can be
627 * parsed and used to produce a manual page. The workflow is the following,
628 * and requires the rst2man utility:
630 * $ ./scripts/bpf_helpers_doc.py \
631 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
632 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
633 * $ man /tmp/bpf-helpers.7
635 * Note that in order to produce this external documentation, some RST
636 * formatting is used in the descriptions to get "bold" and "italics" in
637 * manual pages. Also note that the few trailing white spaces are
638 * intentional, removing them would break paragraphs for rst2man.
640 * Start of BPF helper function descriptions:
642 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
644 * Perform a lookup in *map* for an entry associated to *key*.
646 * Map value associated to *key*, or **NULL** if no entry was
649 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
651 * Add or update the value of the entry associated to *key* in
652 * *map* with *value*. *flags* is one of:
655 * The entry for *key* must not exist in the map.
657 * The entry for *key* must already exist in the map.
659 * No condition on the existence of the entry for *key*.
661 * Flag value **BPF_NOEXIST** cannot be used for maps of types
662 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
663 * elements always exist), the helper would return an error.
665 * 0 on success, or a negative error in case of failure.
667 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
669 * Delete entry with *key* from *map*.
671 * 0 on success, or a negative error in case of failure.
673 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
675 * For tracing programs, safely attempt to read *size* bytes from
676 * kernel space address *unsafe_ptr* and store the data in *dst*.
678 * Generally, use **bpf_probe_read_user**\ () or
679 * **bpf_probe_read_kernel**\ () instead.
681 * 0 on success, or a negative error in case of failure.
683 * u64 bpf_ktime_get_ns(void)
685 * Return the time elapsed since system boot, in nanoseconds.
686 * Does not include time the system was suspended.
687 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
691 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
693 * This helper is a "printk()-like" facility for debugging. It
694 * prints a message defined by format *fmt* (of size *fmt_size*)
695 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
696 * available. It can take up to three additional **u64**
697 * arguments (as an eBPF helpers, the total number of arguments is
700 * Each time the helper is called, it appends a line to the trace.
701 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
702 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
703 * The format of the trace is customizable, and the exact output
704 * one will get depends on the options set in
705 * *\/sys/kernel/debug/tracing/trace_options* (see also the
706 * *README* file under the same directory). However, it usually
707 * defaults to something like:
711 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
715 * * ``telnet`` is the name of the current task.
716 * * ``470`` is the PID of the current task.
717 * * ``001`` is the CPU number on which the task is
719 * * In ``.N..``, each character refers to a set of
720 * options (whether irqs are enabled, scheduling
721 * options, whether hard/softirqs are running, level of
722 * preempt_disabled respectively). **N** means that
723 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
725 * * ``419421.045894`` is a timestamp.
726 * * ``0x00000001`` is a fake value used by BPF for the
727 * instruction pointer register.
728 * * ``<formatted msg>`` is the message formatted with
731 * The conversion specifiers supported by *fmt* are similar, but
732 * more limited than for printk(). They are **%d**, **%i**,
733 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
734 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
735 * of field, padding with zeroes, etc.) is available, and the
736 * helper will return **-EINVAL** (but print nothing) if it
737 * encounters an unknown specifier.
739 * Also, note that **bpf_trace_printk**\ () is slow, and should
740 * only be used for debugging purposes. For this reason, a notice
741 * bloc (spanning several lines) is printed to kernel logs and
742 * states that the helper should not be used "for production use"
743 * the first time this helper is used (or more precisely, when
744 * **trace_printk**\ () buffers are allocated). For passing values
745 * to user space, perf events should be preferred.
747 * The number of bytes written to the buffer, or a negative error
748 * in case of failure.
750 * u32 bpf_get_prandom_u32(void)
752 * Get a pseudo-random number.
754 * From a security point of view, this helper uses its own
755 * pseudo-random internal state, and cannot be used to infer the
756 * seed of other random functions in the kernel. However, it is
757 * essential to note that the generator used by the helper is not
758 * cryptographically secure.
760 * A random 32-bit unsigned value.
762 * u32 bpf_get_smp_processor_id(void)
764 * Get the SMP (symmetric multiprocessing) processor id. Note that
765 * all programs run with preemption disabled, which means that the
766 * SMP processor id is stable during all the execution of the
769 * The SMP id of the processor running the program.
771 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
773 * Store *len* bytes from address *from* into the packet
774 * associated to *skb*, at *offset*. *flags* are a combination of
775 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
776 * checksum for the packet after storing the bytes) and
777 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
778 * **->swhash** and *skb*\ **->l4hash** to 0).
780 * A call to this helper is susceptible to change the underlying
781 * packet buffer. Therefore, at load time, all checks on pointers
782 * previously done by the verifier are invalidated and must be
783 * performed again, if the helper is used in combination with
784 * direct packet access.
786 * 0 on success, or a negative error in case of failure.
788 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
790 * Recompute the layer 3 (e.g. IP) checksum for the packet
791 * associated to *skb*. Computation is incremental, so the helper
792 * must know the former value of the header field that was
793 * modified (*from*), the new value of this field (*to*), and the
794 * number of bytes (2 or 4) for this field, stored in *size*.
795 * Alternatively, it is possible to store the difference between
796 * the previous and the new values of the header field in *to*, by
797 * setting *from* and *size* to 0. For both methods, *offset*
798 * indicates the location of the IP checksum within the packet.
800 * This helper works in combination with **bpf_csum_diff**\ (),
801 * which does not update the checksum in-place, but offers more
802 * flexibility and can handle sizes larger than 2 or 4 for the
803 * checksum to update.
805 * A call to this helper is susceptible to change the underlying
806 * packet buffer. Therefore, at load time, all checks on pointers
807 * previously done by the verifier are invalidated and must be
808 * performed again, if the helper is used in combination with
809 * direct packet access.
811 * 0 on success, or a negative error in case of failure.
813 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
815 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
816 * packet associated to *skb*. Computation is incremental, so the
817 * helper must know the former value of the header field that was
818 * modified (*from*), the new value of this field (*to*), and the
819 * number of bytes (2 or 4) for this field, stored on the lowest
820 * four bits of *flags*. Alternatively, it is possible to store
821 * the difference between the previous and the new values of the
822 * header field in *to*, by setting *from* and the four lowest
823 * bits of *flags* to 0. For both methods, *offset* indicates the
824 * location of the IP checksum within the packet. In addition to
825 * the size of the field, *flags* can be added (bitwise OR) actual
826 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
827 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
828 * for updates resulting in a null checksum the value is set to
829 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
830 * the checksum is to be computed against a pseudo-header.
832 * This helper works in combination with **bpf_csum_diff**\ (),
833 * which does not update the checksum in-place, but offers more
834 * flexibility and can handle sizes larger than 2 or 4 for the
835 * checksum to update.
837 * A call to this helper is susceptible to change the underlying
838 * packet buffer. Therefore, at load time, all checks on pointers
839 * previously done by the verifier are invalidated and must be
840 * performed again, if the helper is used in combination with
841 * direct packet access.
843 * 0 on success, or a negative error in case of failure.
845 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
847 * This special helper is used to trigger a "tail call", or in
848 * other words, to jump into another eBPF program. The same stack
849 * frame is used (but values on stack and in registers for the
850 * caller are not accessible to the callee). This mechanism allows
851 * for program chaining, either for raising the maximum number of
852 * available eBPF instructions, or to execute given programs in
853 * conditional blocks. For security reasons, there is an upper
854 * limit to the number of successive tail calls that can be
857 * Upon call of this helper, the program attempts to jump into a
858 * program referenced at index *index* in *prog_array_map*, a
859 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
860 * *ctx*, a pointer to the context.
862 * If the call succeeds, the kernel immediately runs the first
863 * instruction of the new program. This is not a function call,
864 * and it never returns to the previous program. If the call
865 * fails, then the helper has no effect, and the caller continues
866 * to run its subsequent instructions. A call can fail if the
867 * destination program for the jump does not exist (i.e. *index*
868 * is superior to the number of entries in *prog_array_map*), or
869 * if the maximum number of tail calls has been reached for this
870 * chain of programs. This limit is defined in the kernel by the
871 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
872 * which is currently set to 32.
874 * 0 on success, or a negative error in case of failure.
876 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
878 * Clone and redirect the packet associated to *skb* to another
879 * net device of index *ifindex*. Both ingress and egress
880 * interfaces can be used for redirection. The **BPF_F_INGRESS**
881 * value in *flags* is used to make the distinction (ingress path
882 * is selected if the flag is present, egress path otherwise).
883 * This is the only flag supported for now.
885 * In comparison with **bpf_redirect**\ () helper,
886 * **bpf_clone_redirect**\ () has the associated cost of
887 * duplicating the packet buffer, but this can be executed out of
888 * the eBPF program. Conversely, **bpf_redirect**\ () is more
889 * efficient, but it is handled through an action code where the
890 * redirection happens only after the eBPF program has returned.
892 * A call to this helper is susceptible to change the underlying
893 * packet buffer. Therefore, at load time, all checks on pointers
894 * previously done by the verifier are invalidated and must be
895 * performed again, if the helper is used in combination with
896 * direct packet access.
898 * 0 on success, or a negative error in case of failure.
900 * u64 bpf_get_current_pid_tgid(void)
902 * A 64-bit integer containing the current tgid and pid, and
904 * *current_task*\ **->tgid << 32 \|**
905 * *current_task*\ **->pid**.
907 * u64 bpf_get_current_uid_gid(void)
909 * A 64-bit integer containing the current GID and UID, and
910 * created as such: *current_gid* **<< 32 \|** *current_uid*.
912 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
914 * Copy the **comm** attribute of the current task into *buf* of
915 * *size_of_buf*. The **comm** attribute contains the name of
916 * the executable (excluding the path) for the current task. The
917 * *size_of_buf* must be strictly positive. On success, the
918 * helper makes sure that the *buf* is NUL-terminated. On failure,
919 * it is filled with zeroes.
921 * 0 on success, or a negative error in case of failure.
923 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
925 * Retrieve the classid for the current task, i.e. for the net_cls
926 * cgroup to which *skb* belongs.
928 * This helper can be used on TC egress path, but not on ingress.
930 * The net_cls cgroup provides an interface to tag network packets
931 * based on a user-provided identifier for all traffic coming from
932 * the tasks belonging to the related cgroup. See also the related
933 * kernel documentation, available from the Linux sources in file
934 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
936 * The Linux kernel has two versions for cgroups: there are
937 * cgroups v1 and cgroups v2. Both are available to users, who can
938 * use a mixture of them, but note that the net_cls cgroup is for
939 * cgroup v1 only. This makes it incompatible with BPF programs
940 * run on cgroups, which is a cgroup-v2-only feature (a socket can
941 * only hold data for one version of cgroups at a time).
943 * This helper is only available is the kernel was compiled with
944 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
945 * "**y**" or to "**m**".
947 * The classid, or 0 for the default unconfigured classid.
949 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
951 * Push a *vlan_tci* (VLAN tag control information) of protocol
952 * *vlan_proto* to the packet associated to *skb*, then update
953 * the checksum. Note that if *vlan_proto* is different from
954 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
955 * be **ETH_P_8021Q**.
957 * A call to this helper is susceptible to change the underlying
958 * packet buffer. Therefore, at load time, all checks on pointers
959 * previously done by the verifier are invalidated and must be
960 * performed again, if the helper is used in combination with
961 * direct packet access.
963 * 0 on success, or a negative error in case of failure.
965 * int bpf_skb_vlan_pop(struct sk_buff *skb)
967 * Pop a VLAN header from the packet associated to *skb*.
969 * A call to this helper is susceptible to change the underlying
970 * packet buffer. Therefore, at load time, all checks on pointers
971 * previously done by the verifier are invalidated and must be
972 * performed again, if the helper is used in combination with
973 * direct packet access.
975 * 0 on success, or a negative error in case of failure.
977 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
979 * Get tunnel metadata. This helper takes a pointer *key* to an
980 * empty **struct bpf_tunnel_key** of **size**, that will be
981 * filled with tunnel metadata for the packet associated to *skb*.
982 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
983 * indicates that the tunnel is based on IPv6 protocol instead of
986 * The **struct bpf_tunnel_key** is an object that generalizes the
987 * principal parameters used by various tunneling protocols into a
988 * single struct. This way, it can be used to easily make a
989 * decision based on the contents of the encapsulation header,
990 * "summarized" in this struct. In particular, it holds the IP
991 * address of the remote end (IPv4 or IPv6, depending on the case)
992 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
993 * this struct exposes the *key*\ **->tunnel_id**, which is
994 * generally mapped to a VNI (Virtual Network Identifier), making
995 * it programmable together with the **bpf_skb_set_tunnel_key**\
998 * Let's imagine that the following code is part of a program
999 * attached to the TC ingress interface, on one end of a GRE
1000 * tunnel, and is supposed to filter out all messages coming from
1001 * remote ends with IPv4 address other than 10.0.0.1:
1006 * struct bpf_tunnel_key key = {};
1008 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1010 * return TC_ACT_SHOT; // drop packet
1012 * if (key.remote_ipv4 != 0x0a000001)
1013 * return TC_ACT_SHOT; // drop packet
1015 * return TC_ACT_OK; // accept packet
1017 * This interface can also be used with all encapsulation devices
1018 * that can operate in "collect metadata" mode: instead of having
1019 * one network device per specific configuration, the "collect
1020 * metadata" mode only requires a single device where the
1021 * configuration can be extracted from this helper.
1023 * This can be used together with various tunnels such as VXLan,
1024 * Geneve, GRE or IP in IP (IPIP).
1026 * 0 on success, or a negative error in case of failure.
1028 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1030 * Populate tunnel metadata for packet associated to *skb.* The
1031 * tunnel metadata is set to the contents of *key*, of *size*. The
1032 * *flags* can be set to a combination of the following values:
1034 * **BPF_F_TUNINFO_IPV6**
1035 * Indicate that the tunnel is based on IPv6 protocol
1037 * **BPF_F_ZERO_CSUM_TX**
1038 * For IPv4 packets, add a flag to tunnel metadata
1039 * indicating that checksum computation should be skipped
1040 * and checksum set to zeroes.
1041 * **BPF_F_DONT_FRAGMENT**
1042 * Add a flag to tunnel metadata indicating that the
1043 * packet should not be fragmented.
1044 * **BPF_F_SEQ_NUMBER**
1045 * Add a flag to tunnel metadata indicating that a
1046 * sequence number should be added to tunnel header before
1047 * sending the packet. This flag was added for GRE
1048 * encapsulation, but might be used with other protocols
1049 * as well in the future.
1051 * Here is a typical usage on the transmit path:
1055 * struct bpf_tunnel_key key;
1057 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1058 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1060 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1061 * helper for additional information.
1063 * 0 on success, or a negative error in case of failure.
1065 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1067 * Read the value of a perf event counter. This helper relies on a
1068 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1069 * the perf event counter is selected when *map* is updated with
1070 * perf event file descriptors. The *map* is an array whose size
1071 * is the number of available CPUs, and each cell contains a value
1072 * relative to one CPU. The value to retrieve is indicated by
1073 * *flags*, that contains the index of the CPU to look up, masked
1074 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1075 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1076 * current CPU should be retrieved.
1078 * Note that before Linux 4.13, only hardware perf event can be
1081 * Also, be aware that the newer helper
1082 * **bpf_perf_event_read_value**\ () is recommended over
1083 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1084 * quirks where error and counter value are used as a return code
1085 * (which is wrong to do since ranges may overlap). This issue is
1086 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1087 * time provides more features over the **bpf_perf_event_read**\
1088 * () interface. Please refer to the description of
1089 * **bpf_perf_event_read_value**\ () for details.
1091 * The value of the perf event counter read from the map, or a
1092 * negative error code in case of failure.
1094 * int bpf_redirect(u32 ifindex, u64 flags)
1096 * Redirect the packet to another net device of index *ifindex*.
1097 * This helper is somewhat similar to **bpf_clone_redirect**\
1098 * (), except that the packet is not cloned, which provides
1099 * increased performance.
1101 * Except for XDP, both ingress and egress interfaces can be used
1102 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1103 * to make the distinction (ingress path is selected if the flag
1104 * is present, egress path otherwise). Currently, XDP only
1105 * supports redirection to the egress interface, and accepts no
1108 * The same effect can also be attained with the more generic
1109 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1110 * redirect target instead of providing it directly to the helper.
1112 * For XDP, the helper returns **XDP_REDIRECT** on success or
1113 * **XDP_ABORTED** on error. For other program types, the values
1114 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1117 * u32 bpf_get_route_realm(struct sk_buff *skb)
1119 * Retrieve the realm or the route, that is to say the
1120 * **tclassid** field of the destination for the *skb*. The
1121 * indentifier retrieved is a user-provided tag, similar to the
1122 * one used with the net_cls cgroup (see description for
1123 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1124 * held by a route (a destination entry), not by a task.
1126 * Retrieving this identifier works with the clsact TC egress hook
1127 * (see also **tc-bpf(8)**), or alternatively on conventional
1128 * classful egress qdiscs, but not on TC ingress path. In case of
1129 * clsact TC egress hook, this has the advantage that, internally,
1130 * the destination entry has not been dropped yet in the transmit
1131 * path. Therefore, the destination entry does not need to be
1132 * artificially held via **netif_keep_dst**\ () for a classful
1133 * qdisc until the *skb* is freed.
1135 * This helper is available only if the kernel was compiled with
1136 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1138 * The realm of the route for the packet associated to *skb*, or 0
1139 * if none was found.
1141 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1143 * Write raw *data* blob into a special BPF perf event held by
1144 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1145 * event must have the following attributes: **PERF_SAMPLE_RAW**
1146 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1147 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1149 * The *flags* are used to indicate the index in *map* for which
1150 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1151 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1152 * to indicate that the index of the current CPU core should be
1155 * The value to write, of *size*, is passed through eBPF stack and
1156 * pointed by *data*.
1158 * The context of the program *ctx* needs also be passed to the
1161 * On user space, a program willing to read the values needs to
1162 * call **perf_event_open**\ () on the perf event (either for
1163 * one or for all CPUs) and to store the file descriptor into the
1164 * *map*. This must be done before the eBPF program can send data
1165 * into it. An example is available in file
1166 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1167 * tree (the eBPF program counterpart is in
1168 * *samples/bpf/trace_output_kern.c*).
1170 * **bpf_perf_event_output**\ () achieves better performance
1171 * than **bpf_trace_printk**\ () for sharing data with user
1172 * space, and is much better suitable for streaming data from eBPF
1175 * Note that this helper is not restricted to tracing use cases
1176 * and can be used with programs attached to TC or XDP as well,
1177 * where it allows for passing data to user space listeners. Data
1180 * * Only custom structs,
1181 * * Only the packet payload, or
1182 * * A combination of both.
1184 * 0 on success, or a negative error in case of failure.
1186 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1188 * This helper was provided as an easy way to load data from a
1189 * packet. It can be used to load *len* bytes from *offset* from
1190 * the packet associated to *skb*, into the buffer pointed by
1193 * Since Linux 4.7, usage of this helper has mostly been replaced
1194 * by "direct packet access", enabling packet data to be
1195 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1196 * pointing respectively to the first byte of packet data and to
1197 * the byte after the last byte of packet data. However, it
1198 * remains useful if one wishes to read large quantities of data
1199 * at once from a packet into the eBPF stack.
1201 * 0 on success, or a negative error in case of failure.
1203 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1205 * Walk a user or a kernel stack and return its id. To achieve
1206 * this, the helper needs *ctx*, which is a pointer to the context
1207 * on which the tracing program is executed, and a pointer to a
1208 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1210 * The last argument, *flags*, holds the number of stack frames to
1211 * skip (from 0 to 255), masked with
1212 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1213 * a combination of the following flags:
1215 * **BPF_F_USER_STACK**
1216 * Collect a user space stack instead of a kernel stack.
1217 * **BPF_F_FAST_STACK_CMP**
1218 * Compare stacks by hash only.
1219 * **BPF_F_REUSE_STACKID**
1220 * If two different stacks hash into the same *stackid*,
1221 * discard the old one.
1223 * The stack id retrieved is a 32 bit long integer handle which
1224 * can be further combined with other data (including other stack
1225 * ids) and used as a key into maps. This can be useful for
1226 * generating a variety of graphs (such as flame graphs or off-cpu
1229 * For walking a stack, this helper is an improvement over
1230 * **bpf_probe_read**\ (), which can be used with unrolled loops
1231 * but is not efficient and consumes a lot of eBPF instructions.
1232 * Instead, **bpf_get_stackid**\ () can collect up to
1233 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1234 * this limit can be controlled with the **sysctl** program, and
1235 * that it should be manually increased in order to profile long
1236 * user stacks (such as stacks for Java programs). To do so, use:
1240 * # sysctl kernel.perf_event_max_stack=<new value>
1242 * The positive or null stack id on success, or a negative error
1243 * in case of failure.
1245 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1247 * Compute a checksum difference, from the raw buffer pointed by
1248 * *from*, of length *from_size* (that must be a multiple of 4),
1249 * towards the raw buffer pointed by *to*, of size *to_size*
1250 * (same remark). An optional *seed* can be added to the value
1251 * (this can be cascaded, the seed may come from a previous call
1254 * This is flexible enough to be used in several ways:
1256 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1257 * checksum, it can be used when pushing new data.
1258 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1259 * checksum, it can be used when removing data from a packet.
1260 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1261 * can be used to compute a diff. Note that *from_size* and
1262 * *to_size* do not need to be equal.
1264 * This helper can be used in combination with
1265 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1266 * which one can feed in the difference computed with
1267 * **bpf_csum_diff**\ ().
1269 * The checksum result, or a negative error code in case of
1272 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1274 * Retrieve tunnel options metadata for the packet associated to
1275 * *skb*, and store the raw tunnel option data to the buffer *opt*
1278 * This helper can be used with encapsulation devices that can
1279 * operate in "collect metadata" mode (please refer to the related
1280 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1281 * more details). A particular example where this can be used is
1282 * in combination with the Geneve encapsulation protocol, where it
1283 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1284 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1285 * the eBPF program. This allows for full customization of these
1288 * The size of the option data retrieved.
1290 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1292 * Set tunnel options metadata for the packet associated to *skb*
1293 * to the option data contained in the raw buffer *opt* of *size*.
1295 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1296 * helper for additional information.
1298 * 0 on success, or a negative error in case of failure.
1300 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1302 * Change the protocol of the *skb* to *proto*. Currently
1303 * supported are transition from IPv4 to IPv6, and from IPv6 to
1304 * IPv4. The helper takes care of the groundwork for the
1305 * transition, including resizing the socket buffer. The eBPF
1306 * program is expected to fill the new headers, if any, via
1307 * **skb_store_bytes**\ () and to recompute the checksums with
1308 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1309 * (). The main case for this helper is to perform NAT64
1310 * operations out of an eBPF program.
1312 * Internally, the GSO type is marked as dodgy so that headers are
1313 * checked and segments are recalculated by the GSO/GRO engine.
1314 * The size for GSO target is adapted as well.
1316 * All values for *flags* are reserved for future usage, and must
1319 * A call to this helper is susceptible to change the underlying
1320 * packet buffer. Therefore, at load time, all checks on pointers
1321 * previously done by the verifier are invalidated and must be
1322 * performed again, if the helper is used in combination with
1323 * direct packet access.
1325 * 0 on success, or a negative error in case of failure.
1327 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1329 * Change the packet type for the packet associated to *skb*. This
1330 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1331 * the eBPF program does not have a write access to *skb*\
1332 * **->pkt_type** beside this helper. Using a helper here allows
1333 * for graceful handling of errors.
1335 * The major use case is to change incoming *skb*s to
1336 * **PACKET_HOST** in a programmatic way instead of having to
1337 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1340 * Note that *type* only allows certain values. At this time, they
1345 * **PACKET_BROADCAST**
1346 * Send packet to all.
1347 * **PACKET_MULTICAST**
1348 * Send packet to group.
1349 * **PACKET_OTHERHOST**
1350 * Send packet to someone else.
1352 * 0 on success, or a negative error in case of failure.
1354 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1356 * Check whether *skb* is a descendant of the cgroup2 held by
1357 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1359 * The return value depends on the result of the test, and can be:
1361 * * 0, if the *skb* failed the cgroup2 descendant test.
1362 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1363 * * A negative error code, if an error occurred.
1365 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1367 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1368 * not set, in particular if the hash was cleared due to mangling,
1369 * recompute this hash. Later accesses to the hash can be done
1370 * directly with *skb*\ **->hash**.
1372 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1373 * prototype with **bpf_skb_change_proto**\ (), or calling
1374 * **bpf_skb_store_bytes**\ () with the
1375 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1376 * the hash and to trigger a new computation for the next call to
1377 * **bpf_get_hash_recalc**\ ().
1381 * u64 bpf_get_current_task(void)
1383 * A pointer to the current task struct.
1385 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1387 * Attempt in a safe way to write *len* bytes from the buffer
1388 * *src* to *dst* in memory. It only works for threads that are in
1389 * user context, and *dst* must be a valid user space address.
1391 * This helper should not be used to implement any kind of
1392 * security mechanism because of TOC-TOU attacks, but rather to
1393 * debug, divert, and manipulate execution of semi-cooperative
1396 * Keep in mind that this feature is meant for experiments, and it
1397 * has a risk of crashing the system and running programs.
1398 * Therefore, when an eBPF program using this helper is attached,
1399 * a warning including PID and process name is printed to kernel
1402 * 0 on success, or a negative error in case of failure.
1404 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1406 * Check whether the probe is being run is the context of a given
1407 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1408 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1410 * The return value depends on the result of the test, and can be:
1412 * * 0, if the *skb* task belongs to the cgroup2.
1413 * * 1, if the *skb* task does not belong to the cgroup2.
1414 * * A negative error code, if an error occurred.
1416 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1418 * Resize (trim or grow) the packet associated to *skb* to the
1419 * new *len*. The *flags* are reserved for future usage, and must
1422 * The basic idea is that the helper performs the needed work to
1423 * change the size of the packet, then the eBPF program rewrites
1424 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1425 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1426 * and others. This helper is a slow path utility intended for
1427 * replies with control messages. And because it is targeted for
1428 * slow path, the helper itself can afford to be slow: it
1429 * implicitly linearizes, unclones and drops offloads from the
1432 * A call to this helper is susceptible to change the underlying
1433 * packet buffer. Therefore, at load time, all checks on pointers
1434 * previously done by the verifier are invalidated and must be
1435 * performed again, if the helper is used in combination with
1436 * direct packet access.
1438 * 0 on success, or a negative error in case of failure.
1440 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1442 * Pull in non-linear data in case the *skb* is non-linear and not
1443 * all of *len* are part of the linear section. Make *len* bytes
1444 * from *skb* readable and writable. If a zero value is passed for
1445 * *len*, then the whole length of the *skb* is pulled.
1447 * This helper is only needed for reading and writing with direct
1450 * For direct packet access, testing that offsets to access
1451 * are within packet boundaries (test on *skb*\ **->data_end**) is
1452 * susceptible to fail if offsets are invalid, or if the requested
1453 * data is in non-linear parts of the *skb*. On failure the
1454 * program can just bail out, or in the case of a non-linear
1455 * buffer, use a helper to make the data available. The
1456 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1457 * the data. Another one consists in using **bpf_skb_pull_data**
1458 * to pull in once the non-linear parts, then retesting and
1459 * eventually access the data.
1461 * At the same time, this also makes sure the *skb* is uncloned,
1462 * which is a necessary condition for direct write. As this needs
1463 * to be an invariant for the write part only, the verifier
1464 * detects writes and adds a prologue that is calling
1465 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1466 * the very beginning in case it is indeed cloned.
1468 * A call to this helper is susceptible to change the underlying
1469 * packet buffer. Therefore, at load time, all checks on pointers
1470 * previously done by the verifier are invalidated and must be
1471 * performed again, if the helper is used in combination with
1472 * direct packet access.
1474 * 0 on success, or a negative error in case of failure.
1476 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1478 * Add the checksum *csum* into *skb*\ **->csum** in case the
1479 * driver has supplied a checksum for the entire packet into that
1480 * field. Return an error otherwise. This helper is intended to be
1481 * used in combination with **bpf_csum_diff**\ (), in particular
1482 * when the checksum needs to be updated after data has been
1483 * written into the packet through direct packet access.
1485 * The checksum on success, or a negative error code in case of
1488 * void bpf_set_hash_invalid(struct sk_buff *skb)
1490 * Invalidate the current *skb*\ **->hash**. It can be used after
1491 * mangling on headers through direct packet access, in order to
1492 * indicate that the hash is outdated and to trigger a
1493 * recalculation the next time the kernel tries to access this
1494 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1496 * int bpf_get_numa_node_id(void)
1498 * Return the id of the current NUMA node. The primary use case
1499 * for this helper is the selection of sockets for the local NUMA
1500 * node, when the program is attached to sockets using the
1501 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1502 * but the helper is also available to other eBPF program types,
1503 * similarly to **bpf_get_smp_processor_id**\ ().
1505 * The id of current NUMA node.
1507 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1509 * Grows headroom of packet associated to *skb* and adjusts the
1510 * offset of the MAC header accordingly, adding *len* bytes of
1511 * space. It automatically extends and reallocates memory as
1514 * This helper can be used on a layer 3 *skb* to push a MAC header
1515 * for redirection into a layer 2 device.
1517 * All values for *flags* are reserved for future usage, and must
1520 * A call to this helper is susceptible to change the underlying
1521 * packet buffer. Therefore, at load time, all checks on pointers
1522 * previously done by the verifier are invalidated and must be
1523 * performed again, if the helper is used in combination with
1524 * direct packet access.
1526 * 0 on success, or a negative error in case of failure.
1528 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1530 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1531 * it is possible to use a negative value for *delta*. This helper
1532 * can be used to prepare the packet for pushing or popping
1535 * A call to this helper is susceptible to change the underlying
1536 * packet buffer. Therefore, at load time, all checks on pointers
1537 * previously done by the verifier are invalidated and must be
1538 * performed again, if the helper is used in combination with
1539 * direct packet access.
1541 * 0 on success, or a negative error in case of failure.
1543 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1545 * Copy a NUL terminated string from an unsafe kernel address
1546 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
1549 * Generally, use **bpf_probe_read_user_str**\ () or
1550 * **bpf_probe_read_kernel_str**\ () instead.
1552 * On success, the strictly positive length of the string,
1553 * including the trailing NUL character. On error, a negative
1556 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1558 * If the **struct sk_buff** pointed by *skb* has a known socket,
1559 * retrieve the cookie (generated by the kernel) of this socket.
1560 * If no cookie has been set yet, generate a new cookie. Once
1561 * generated, the socket cookie remains stable for the life of the
1562 * socket. This helper can be useful for monitoring per socket
1563 * networking traffic statistics as it provides a global socket
1564 * identifier that can be assumed unique.
1566 * A 8-byte long non-decreasing number on success, or 0 if the
1567 * socket field is missing inside *skb*.
1569 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1571 * Equivalent to bpf_get_socket_cookie() helper that accepts
1572 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1574 * A 8-byte long non-decreasing number.
1576 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1578 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
1579 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1581 * A 8-byte long non-decreasing number.
1583 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1585 * The owner UID of the socket associated to *skb*. If the socket
1586 * is **NULL**, or if it is not a full socket (i.e. if it is a
1587 * time-wait or a request socket instead), **overflowuid** value
1588 * is returned (note that **overflowuid** might also be the actual
1589 * UID value for the socket).
1591 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1593 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1598 * int bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1600 * Emulate a call to **setsockopt()** on the socket associated to
1601 * *bpf_socket*, which must be a full socket. The *level* at
1602 * which the option resides and the name *optname* of the option
1603 * must be specified, see **setsockopt(2)** for more information.
1604 * The option value of length *optlen* is pointed by *optval*.
1606 * *bpf_socket* should be one of the following:
1608 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1609 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1610 * and **BPF_CGROUP_INET6_CONNECT**.
1612 * This helper actually implements a subset of **setsockopt()**.
1613 * It supports the following *level*\ s:
1615 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1616 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1617 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1618 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1619 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1620 * **TCP_BPF_SNDCWND_CLAMP**.
1621 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1622 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1624 * 0 on success, or a negative error in case of failure.
1626 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1628 * Grow or shrink the room for data in the packet associated to
1629 * *skb* by *len_diff*, and according to the selected *mode*.
1631 * There are two supported modes at this time:
1633 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1634 * (room space is added or removed below the layer 2 header).
1636 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1637 * (room space is added or removed below the layer 3 header).
1639 * The following flags are supported at this time:
1641 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1642 * Adjusting mss in this way is not allowed for datagrams.
1644 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1645 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1646 * Any new space is reserved to hold a tunnel header.
1647 * Configure skb offsets and other fields accordingly.
1649 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1650 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1651 * Use with ENCAP_L3 flags to further specify the tunnel type.
1653 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1654 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1655 * type; *len* is the length of the inner MAC header.
1657 * A call to this helper is susceptible to change the underlying
1658 * packet buffer. Therefore, at load time, all checks on pointers
1659 * previously done by the verifier are invalidated and must be
1660 * performed again, if the helper is used in combination with
1661 * direct packet access.
1663 * 0 on success, or a negative error in case of failure.
1665 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1667 * Redirect the packet to the endpoint referenced by *map* at
1668 * index *key*. Depending on its type, this *map* can contain
1669 * references to net devices (for forwarding packets through other
1670 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1671 * but this is only implemented for native XDP (with driver
1672 * support) as of this writing).
1674 * The lower two bits of *flags* are used as the return code if
1675 * the map lookup fails. This is so that the return value can be
1676 * one of the XDP program return codes up to **XDP_TX**, as chosen
1677 * by the caller. Any higher bits in the *flags* argument must be
1680 * See also **bpf_redirect**\ (), which only supports redirecting
1681 * to an ifindex, but doesn't require a map to do so.
1683 * **XDP_REDIRECT** on success, or the value of the two lower bits
1684 * of the *flags* argument on error.
1686 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1688 * Redirect the packet to the socket referenced by *map* (of type
1689 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1690 * egress interfaces can be used for redirection. The
1691 * **BPF_F_INGRESS** value in *flags* is used to make the
1692 * distinction (ingress path is selected if the flag is present,
1693 * egress path otherwise). This is the only flag supported for now.
1695 * **SK_PASS** on success, or **SK_DROP** on error.
1697 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1699 * Add an entry to, or update a *map* referencing sockets. The
1700 * *skops* is used as a new value for the entry associated to
1701 * *key*. *flags* is one of:
1704 * The entry for *key* must not exist in the map.
1706 * The entry for *key* must already exist in the map.
1708 * No condition on the existence of the entry for *key*.
1710 * If the *map* has eBPF programs (parser and verdict), those will
1711 * be inherited by the socket being added. If the socket is
1712 * already attached to eBPF programs, this results in an error.
1714 * 0 on success, or a negative error in case of failure.
1716 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1718 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1719 * *delta* (which can be positive or negative). Note that this
1720 * operation modifies the address stored in *xdp_md*\ **->data**,
1721 * so the latter must be loaded only after the helper has been
1724 * The use of *xdp_md*\ **->data_meta** is optional and programs
1725 * are not required to use it. The rationale is that when the
1726 * packet is processed with XDP (e.g. as DoS filter), it is
1727 * possible to push further meta data along with it before passing
1728 * to the stack, and to give the guarantee that an ingress eBPF
1729 * program attached as a TC classifier on the same device can pick
1730 * this up for further post-processing. Since TC works with socket
1731 * buffers, it remains possible to set from XDP the **mark** or
1732 * **priority** pointers, or other pointers for the socket buffer.
1733 * Having this scratch space generic and programmable allows for
1734 * more flexibility as the user is free to store whatever meta
1737 * A call to this helper is susceptible to change the underlying
1738 * packet buffer. Therefore, at load time, all checks on pointers
1739 * previously done by the verifier are invalidated and must be
1740 * performed again, if the helper is used in combination with
1741 * direct packet access.
1743 * 0 on success, or a negative error in case of failure.
1745 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1747 * Read the value of a perf event counter, and store it into *buf*
1748 * of size *buf_size*. This helper relies on a *map* of type
1749 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1750 * counter is selected when *map* is updated with perf event file
1751 * descriptors. The *map* is an array whose size is the number of
1752 * available CPUs, and each cell contains a value relative to one
1753 * CPU. The value to retrieve is indicated by *flags*, that
1754 * contains the index of the CPU to look up, masked with
1755 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1756 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1757 * current CPU should be retrieved.
1759 * This helper behaves in a way close to
1760 * **bpf_perf_event_read**\ () helper, save that instead of
1761 * just returning the value observed, it fills the *buf*
1762 * structure. This allows for additional data to be retrieved: in
1763 * particular, the enabled and running times (in *buf*\
1764 * **->enabled** and *buf*\ **->running**, respectively) are
1765 * copied. In general, **bpf_perf_event_read_value**\ () is
1766 * recommended over **bpf_perf_event_read**\ (), which has some
1767 * ABI issues and provides fewer functionalities.
1769 * These values are interesting, because hardware PMU (Performance
1770 * Monitoring Unit) counters are limited resources. When there are
1771 * more PMU based perf events opened than available counters,
1772 * kernel will multiplex these events so each event gets certain
1773 * percentage (but not all) of the PMU time. In case that
1774 * multiplexing happens, the number of samples or counter value
1775 * will not reflect the case compared to when no multiplexing
1776 * occurs. This makes comparison between different runs difficult.
1777 * Typically, the counter value should be normalized before
1778 * comparing to other experiments. The usual normalization is done
1783 * normalized_counter = counter * t_enabled / t_running
1785 * Where t_enabled is the time enabled for event and t_running is
1786 * the time running for event since last normalization. The
1787 * enabled and running times are accumulated since the perf event
1788 * open. To achieve scaling factor between two invocations of an
1789 * eBPF program, users can use CPU id as the key (which is
1790 * typical for perf array usage model) to remember the previous
1791 * value and do the calculation inside the eBPF program.
1793 * 0 on success, or a negative error in case of failure.
1795 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1797 * For en eBPF program attached to a perf event, retrieve the
1798 * value of the event counter associated to *ctx* and store it in
1799 * the structure pointed by *buf* and of size *buf_size*. Enabled
1800 * and running times are also stored in the structure (see
1801 * description of helper **bpf_perf_event_read_value**\ () for
1804 * 0 on success, or a negative error in case of failure.
1806 * int bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1808 * Emulate a call to **getsockopt()** on the socket associated to
1809 * *bpf_socket*, which must be a full socket. The *level* at
1810 * which the option resides and the name *optname* of the option
1811 * must be specified, see **getsockopt(2)** for more information.
1812 * The retrieved value is stored in the structure pointed by
1813 * *opval* and of length *optlen*.
1815 * *bpf_socket* should be one of the following:
1817 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1818 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1819 * and **BPF_CGROUP_INET6_CONNECT**.
1821 * This helper actually implements a subset of **getsockopt()**.
1822 * It supports the following *level*\ s:
1824 * * **IPPROTO_TCP**, which supports *optname*
1825 * **TCP_CONGESTION**.
1826 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1827 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1829 * 0 on success, or a negative error in case of failure.
1831 * int bpf_override_return(struct pt_regs *regs, u64 rc)
1833 * Used for error injection, this helper uses kprobes to override
1834 * the return value of the probed function, and to set it to *rc*.
1835 * The first argument is the context *regs* on which the kprobe
1838 * This helper works by setting the PC (program counter)
1839 * to an override function which is run in place of the original
1840 * probed function. This means the probed function is not run at
1841 * all. The replacement function just returns with the required
1844 * This helper has security implications, and thus is subject to
1845 * restrictions. It is only available if the kernel was compiled
1846 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1847 * option, and in this case it only works on functions tagged with
1848 * **ALLOW_ERROR_INJECTION** in the kernel code.
1850 * Also, the helper is only available for the architectures having
1851 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1852 * x86 architecture is the only one to support this feature.
1856 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1858 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1859 * for the full TCP socket associated to *bpf_sock_ops* to
1862 * The primary use of this field is to determine if there should
1863 * be calls to eBPF programs of type
1864 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1865 * code. A program of the same type can change its value, per
1866 * connection and as necessary, when the connection is
1867 * established. This field is directly accessible for reading, but
1868 * this helper must be used for updates in order to return an
1869 * error if an eBPF program tries to set a callback that is not
1870 * supported in the current kernel.
1872 * *argval* is a flag array which can combine these flags:
1874 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1875 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1876 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1877 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1879 * Therefore, this function can be used to clear a callback flag by
1880 * setting the appropriate bit to zero. e.g. to disable the RTO
1883 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1884 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1886 * Here are some examples of where one could call such eBPF
1890 * * When a packet is retransmitted.
1891 * * When the connection terminates.
1892 * * When a packet is sent.
1893 * * When a packet is received.
1895 * Code **-EINVAL** if the socket is not a full TCP socket;
1896 * otherwise, a positive number containing the bits that could not
1897 * be set is returned (which comes down to 0 if all bits were set
1900 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1902 * This helper is used in programs implementing policies at the
1903 * socket level. If the message *msg* is allowed to pass (i.e. if
1904 * the verdict eBPF program returns **SK_PASS**), redirect it to
1905 * the socket referenced by *map* (of type
1906 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1907 * egress interfaces can be used for redirection. The
1908 * **BPF_F_INGRESS** value in *flags* is used to make the
1909 * distinction (ingress path is selected if the flag is present,
1910 * egress path otherwise). This is the only flag supported for now.
1912 * **SK_PASS** on success, or **SK_DROP** on error.
1914 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1916 * For socket policies, apply the verdict of the eBPF program to
1917 * the next *bytes* (number of bytes) of message *msg*.
1919 * For example, this helper can be used in the following cases:
1921 * * A single **sendmsg**\ () or **sendfile**\ () system call
1922 * contains multiple logical messages that the eBPF program is
1923 * supposed to read and for which it should apply a verdict.
1924 * * An eBPF program only cares to read the first *bytes* of a
1925 * *msg*. If the message has a large payload, then setting up
1926 * and calling the eBPF program repeatedly for all bytes, even
1927 * though the verdict is already known, would create unnecessary
1930 * When called from within an eBPF program, the helper sets a
1931 * counter internal to the BPF infrastructure, that is used to
1932 * apply the last verdict to the next *bytes*. If *bytes* is
1933 * smaller than the current data being processed from a
1934 * **sendmsg**\ () or **sendfile**\ () system call, the first
1935 * *bytes* will be sent and the eBPF program will be re-run with
1936 * the pointer for start of data pointing to byte number *bytes*
1937 * **+ 1**. If *bytes* is larger than the current data being
1938 * processed, then the eBPF verdict will be applied to multiple
1939 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1942 * Note that if a socket closes with the internal counter holding
1943 * a non-zero value, this is not a problem because data is not
1944 * being buffered for *bytes* and is sent as it is received.
1948 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1950 * For socket policies, prevent the execution of the verdict eBPF
1951 * program for message *msg* until *bytes* (byte number) have been
1954 * This can be used when one needs a specific number of bytes
1955 * before a verdict can be assigned, even if the data spans
1956 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1957 * case would be a user calling **sendmsg**\ () repeatedly with
1958 * 1-byte long message segments. Obviously, this is bad for
1959 * performance, but it is still valid. If the eBPF program needs
1960 * *bytes* bytes to validate a header, this helper can be used to
1961 * prevent the eBPF program to be called again until *bytes* have
1966 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1968 * For socket policies, pull in non-linear data from user space
1969 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1970 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1973 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1974 * *msg* it can only parse data that the (**data**, **data_end**)
1975 * pointers have already consumed. For **sendmsg**\ () hooks this
1976 * is likely the first scatterlist element. But for calls relying
1977 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1978 * be the range (**0**, **0**) because the data is shared with
1979 * user space and by default the objective is to avoid allowing
1980 * user space to modify data while (or after) eBPF verdict is
1981 * being decided. This helper can be used to pull in data and to
1982 * set the start and end pointer to given values. Data will be
1983 * copied if necessary (i.e. if data was not linear and if start
1984 * and end pointers do not point to the same chunk).
1986 * A call to this helper is susceptible to change the underlying
1987 * packet buffer. Therefore, at load time, all checks on pointers
1988 * previously done by the verifier are invalidated and must be
1989 * performed again, if the helper is used in combination with
1990 * direct packet access.
1992 * All values for *flags* are reserved for future usage, and must
1995 * 0 on success, or a negative error in case of failure.
1997 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1999 * Bind the socket associated to *ctx* to the address pointed by
2000 * *addr*, of length *addr_len*. This allows for making outgoing
2001 * connection from the desired IP address, which can be useful for
2002 * example when all processes inside a cgroup should use one
2003 * single IP address on a host that has multiple IP configured.
2005 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2006 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2007 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2008 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2009 * behavior and lets the kernel efficiently pick up an unused
2010 * port as long as 4-tuple is unique. Passing non-zero port might
2011 * lead to degraded performance.
2013 * 0 on success, or a negative error in case of failure.
2015 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2017 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2018 * only possible to shrink the packet as of this writing,
2019 * therefore *delta* must be a negative integer.
2021 * A call to this helper is susceptible to change the underlying
2022 * packet buffer. Therefore, at load time, all checks on pointers
2023 * previously done by the verifier are invalidated and must be
2024 * performed again, if the helper is used in combination with
2025 * direct packet access.
2027 * 0 on success, or a negative error in case of failure.
2029 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2031 * Retrieve the XFRM state (IP transform framework, see also
2032 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2034 * The retrieved value is stored in the **struct bpf_xfrm_state**
2035 * pointed by *xfrm_state* and of length *size*.
2037 * All values for *flags* are reserved for future usage, and must
2040 * This helper is available only if the kernel was compiled with
2041 * **CONFIG_XFRM** configuration option.
2043 * 0 on success, or a negative error in case of failure.
2045 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2047 * Return a user or a kernel stack in bpf program provided buffer.
2048 * To achieve this, the helper needs *ctx*, which is a pointer
2049 * to the context on which the tracing program is executed.
2050 * To store the stacktrace, the bpf program provides *buf* with
2051 * a nonnegative *size*.
2053 * The last argument, *flags*, holds the number of stack frames to
2054 * skip (from 0 to 255), masked with
2055 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2056 * the following flags:
2058 * **BPF_F_USER_STACK**
2059 * Collect a user space stack instead of a kernel stack.
2060 * **BPF_F_USER_BUILD_ID**
2061 * Collect buildid+offset instead of ips for user stack,
2062 * only valid if **BPF_F_USER_STACK** is also specified.
2064 * **bpf_get_stack**\ () can collect up to
2065 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2066 * to sufficient large buffer size. Note that
2067 * this limit can be controlled with the **sysctl** program, and
2068 * that it should be manually increased in order to profile long
2069 * user stacks (such as stacks for Java programs). To do so, use:
2073 * # sysctl kernel.perf_event_max_stack=<new value>
2075 * A non-negative value equal to or less than *size* on success,
2076 * or a negative error in case of failure.
2078 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2080 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2081 * it provides an easy way to load *len* bytes from *offset*
2082 * from the packet associated to *skb*, into the buffer pointed
2083 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2084 * a fifth argument *start_header* exists in order to select a
2085 * base offset to start from. *start_header* can be one of:
2087 * **BPF_HDR_START_MAC**
2088 * Base offset to load data from is *skb*'s mac header.
2089 * **BPF_HDR_START_NET**
2090 * Base offset to load data from is *skb*'s network header.
2092 * In general, "direct packet access" is the preferred method to
2093 * access packet data, however, this helper is in particular useful
2094 * in socket filters where *skb*\ **->data** does not always point
2095 * to the start of the mac header and where "direct packet access"
2098 * 0 on success, or a negative error in case of failure.
2100 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2102 * Do FIB lookup in kernel tables using parameters in *params*.
2103 * If lookup is successful and result shows packet is to be
2104 * forwarded, the neighbor tables are searched for the nexthop.
2105 * If successful (ie., FIB lookup shows forwarding and nexthop
2106 * is resolved), the nexthop address is returned in ipv4_dst
2107 * or ipv6_dst based on family, smac is set to mac address of
2108 * egress device, dmac is set to nexthop mac address, rt_metric
2109 * is set to metric from route (IPv4/IPv6 only), and ifindex
2110 * is set to the device index of the nexthop from the FIB lookup.
2112 * *plen* argument is the size of the passed in struct.
2113 * *flags* argument can be a combination of one or more of the
2116 * **BPF_FIB_LOOKUP_DIRECT**
2117 * Do a direct table lookup vs full lookup using FIB
2119 * **BPF_FIB_LOOKUP_OUTPUT**
2120 * Perform lookup from an egress perspective (default is
2123 * *ctx* is either **struct xdp_md** for XDP programs or
2124 * **struct sk_buff** tc cls_act programs.
2126 * * < 0 if any input argument is invalid
2127 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2128 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2129 * packet is not forwarded or needs assist from full stack
2131 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2133 * Add an entry to, or update a sockhash *map* referencing sockets.
2134 * The *skops* is used as a new value for the entry associated to
2135 * *key*. *flags* is one of:
2138 * The entry for *key* must not exist in the map.
2140 * The entry for *key* must already exist in the map.
2142 * No condition on the existence of the entry for *key*.
2144 * If the *map* has eBPF programs (parser and verdict), those will
2145 * be inherited by the socket being added. If the socket is
2146 * already attached to eBPF programs, this results in an error.
2148 * 0 on success, or a negative error in case of failure.
2150 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2152 * This helper is used in programs implementing policies at the
2153 * socket level. If the message *msg* is allowed to pass (i.e. if
2154 * the verdict eBPF program returns **SK_PASS**), redirect it to
2155 * the socket referenced by *map* (of type
2156 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2157 * egress interfaces can be used for redirection. The
2158 * **BPF_F_INGRESS** value in *flags* is used to make the
2159 * distinction (ingress path is selected if the flag is present,
2160 * egress path otherwise). This is the only flag supported for now.
2162 * **SK_PASS** on success, or **SK_DROP** on error.
2164 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2166 * This helper is used in programs implementing policies at the
2167 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2168 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2169 * to the socket referenced by *map* (of type
2170 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2171 * egress interfaces can be used for redirection. The
2172 * **BPF_F_INGRESS** value in *flags* is used to make the
2173 * distinction (ingress path is selected if the flag is present,
2174 * egress otherwise). This is the only flag supported for now.
2176 * **SK_PASS** on success, or **SK_DROP** on error.
2178 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2180 * Encapsulate the packet associated to *skb* within a Layer 3
2181 * protocol header. This header is provided in the buffer at
2182 * address *hdr*, with *len* its size in bytes. *type* indicates
2183 * the protocol of the header and can be one of:
2185 * **BPF_LWT_ENCAP_SEG6**
2186 * IPv6 encapsulation with Segment Routing Header
2187 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2188 * the IPv6 header is computed by the kernel.
2189 * **BPF_LWT_ENCAP_SEG6_INLINE**
2190 * Only works if *skb* contains an IPv6 packet. Insert a
2191 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2193 * **BPF_LWT_ENCAP_IP**
2194 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2195 * must be IPv4 or IPv6, followed by zero or more
2196 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2197 * total bytes in all prepended headers. Please note that
2198 * if **skb_is_gso**\ (*skb*) is true, no more than two
2199 * headers can be prepended, and the inner header, if
2200 * present, should be either GRE or UDP/GUE.
2202 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2203 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2204 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2205 * **BPF_PROG_TYPE_LWT_XMIT**.
2207 * A call to this helper is susceptible to change the underlying
2208 * packet buffer. Therefore, at load time, all checks on pointers
2209 * previously done by the verifier are invalidated and must be
2210 * performed again, if the helper is used in combination with
2211 * direct packet access.
2213 * 0 on success, or a negative error in case of failure.
2215 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2217 * Store *len* bytes from address *from* into the packet
2218 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2219 * inside the outermost IPv6 Segment Routing Header can be
2220 * modified through this helper.
2222 * A call to this helper is susceptible to change the underlying
2223 * packet buffer. Therefore, at load time, all checks on pointers
2224 * previously done by the verifier are invalidated and must be
2225 * performed again, if the helper is used in combination with
2226 * direct packet access.
2228 * 0 on success, or a negative error in case of failure.
2230 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2232 * Adjust the size allocated to TLVs in the outermost IPv6
2233 * Segment Routing Header contained in the packet associated to
2234 * *skb*, at position *offset* by *delta* bytes. Only offsets
2235 * after the segments are accepted. *delta* can be as well
2236 * positive (growing) as negative (shrinking).
2238 * A call to this helper is susceptible to change the underlying
2239 * packet buffer. Therefore, at load time, all checks on pointers
2240 * previously done by the verifier are invalidated and must be
2241 * performed again, if the helper is used in combination with
2242 * direct packet access.
2244 * 0 on success, or a negative error in case of failure.
2246 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2248 * Apply an IPv6 Segment Routing action of type *action* to the
2249 * packet associated to *skb*. Each action takes a parameter
2250 * contained at address *param*, and of length *param_len* bytes.
2251 * *action* can be one of:
2253 * **SEG6_LOCAL_ACTION_END_X**
2254 * End.X action: Endpoint with Layer-3 cross-connect.
2255 * Type of *param*: **struct in6_addr**.
2256 * **SEG6_LOCAL_ACTION_END_T**
2257 * End.T action: Endpoint with specific IPv6 table lookup.
2258 * Type of *param*: **int**.
2259 * **SEG6_LOCAL_ACTION_END_B6**
2260 * End.B6 action: Endpoint bound to an SRv6 policy.
2261 * Type of *param*: **struct ipv6_sr_hdr**.
2262 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2263 * End.B6.Encap action: Endpoint bound to an SRv6
2264 * encapsulation policy.
2265 * Type of *param*: **struct ipv6_sr_hdr**.
2267 * A call to this helper is susceptible to change the underlying
2268 * packet buffer. Therefore, at load time, all checks on pointers
2269 * previously done by the verifier are invalidated and must be
2270 * performed again, if the helper is used in combination with
2271 * direct packet access.
2273 * 0 on success, or a negative error in case of failure.
2275 * int bpf_rc_repeat(void *ctx)
2277 * This helper is used in programs implementing IR decoding, to
2278 * report a successfully decoded repeat key message. This delays
2279 * the generation of a key up event for previously generated
2282 * Some IR protocols like NEC have a special IR message for
2283 * repeating last button, for when a button is held down.
2285 * The *ctx* should point to the lirc sample as passed into
2288 * This helper is only available is the kernel was compiled with
2289 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2294 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2296 * This helper is used in programs implementing IR decoding, to
2297 * report a successfully decoded key press with *scancode*,
2298 * *toggle* value in the given *protocol*. The scancode will be
2299 * translated to a keycode using the rc keymap, and reported as
2300 * an input key down event. After a period a key up event is
2301 * generated. This period can be extended by calling either
2302 * **bpf_rc_keydown**\ () again with the same values, or calling
2303 * **bpf_rc_repeat**\ ().
2305 * Some protocols include a toggle bit, in case the button was
2306 * released and pressed again between consecutive scancodes.
2308 * The *ctx* should point to the lirc sample as passed into
2311 * The *protocol* is the decoded protocol number (see
2312 * **enum rc_proto** for some predefined values).
2314 * This helper is only available is the kernel was compiled with
2315 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2320 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2322 * Return the cgroup v2 id of the socket associated with the *skb*.
2323 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2324 * helper for cgroup v1 by providing a tag resp. identifier that
2325 * can be matched on or used for map lookups e.g. to implement
2326 * policy. The cgroup v2 id of a given path in the hierarchy is
2327 * exposed in user space through the f_handle API in order to get
2328 * to the same 64-bit id.
2330 * This helper can be used on TC egress path, but not on ingress,
2331 * and is available only if the kernel was compiled with the
2332 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2334 * The id is returned or 0 in case the id could not be retrieved.
2336 * u64 bpf_get_current_cgroup_id(void)
2338 * A 64-bit integer containing the current cgroup id based
2339 * on the cgroup within which the current task is running.
2341 * void *bpf_get_local_storage(void *map, u64 flags)
2343 * Get the pointer to the local storage area.
2344 * The type and the size of the local storage is defined
2345 * by the *map* argument.
2346 * The *flags* meaning is specific for each map type,
2347 * and has to be 0 for cgroup local storage.
2349 * Depending on the BPF program type, a local storage area
2350 * can be shared between multiple instances of the BPF program,
2351 * running simultaneously.
2353 * A user should care about the synchronization by himself.
2354 * For example, by using the **BPF_STX_XADD** instruction to alter
2357 * A pointer to the local storage area.
2359 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2361 * Select a **SO_REUSEPORT** socket from a
2362 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2363 * It checks the selected socket is matching the incoming
2364 * request in the socket buffer.
2366 * 0 on success, or a negative error in case of failure.
2368 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2370 * Return id of cgroup v2 that is ancestor of cgroup associated
2371 * with the *skb* at the *ancestor_level*. The root cgroup is at
2372 * *ancestor_level* zero and each step down the hierarchy
2373 * increments the level. If *ancestor_level* == level of cgroup
2374 * associated with *skb*, then return value will be same as that
2375 * of **bpf_skb_cgroup_id**\ ().
2377 * The helper is useful to implement policies based on cgroups
2378 * that are upper in hierarchy than immediate cgroup associated
2381 * The format of returned id and helper limitations are same as in
2382 * **bpf_skb_cgroup_id**\ ().
2384 * The id is returned or 0 in case the id could not be retrieved.
2386 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2388 * Look for TCP socket matching *tuple*, optionally in a child
2389 * network namespace *netns*. The return value must be checked,
2390 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2392 * The *ctx* should point to the context of the program, such as
2393 * the skb or socket (depending on the hook in use). This is used
2394 * to determine the base network namespace for the lookup.
2396 * *tuple_size* must be one of:
2398 * **sizeof**\ (*tuple*\ **->ipv4**)
2399 * Look for an IPv4 socket.
2400 * **sizeof**\ (*tuple*\ **->ipv6**)
2401 * Look for an IPv6 socket.
2403 * If the *netns* is a negative signed 32-bit integer, then the
2404 * socket lookup table in the netns associated with the *ctx* will
2405 * will be used. For the TC hooks, this is the netns of the device
2406 * in the skb. For socket hooks, this is the netns of the socket.
2407 * If *netns* is any other signed 32-bit value greater than or
2408 * equal to zero then it specifies the ID of the netns relative to
2409 * the netns associated with the *ctx*. *netns* values beyond the
2410 * range of 32-bit integers are reserved for future use.
2412 * All values for *flags* are reserved for future usage, and must
2415 * This helper is available only if the kernel was compiled with
2416 * **CONFIG_NET** configuration option.
2418 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2419 * For sockets with reuseport option, the **struct bpf_sock**
2420 * result is from *reuse*\ **->socks**\ [] using the hash of the
2423 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2425 * Look for UDP socket matching *tuple*, optionally in a child
2426 * network namespace *netns*. The return value must be checked,
2427 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2429 * The *ctx* should point to the context of the program, such as
2430 * the skb or socket (depending on the hook in use). This is used
2431 * to determine the base network namespace for the lookup.
2433 * *tuple_size* must be one of:
2435 * **sizeof**\ (*tuple*\ **->ipv4**)
2436 * Look for an IPv4 socket.
2437 * **sizeof**\ (*tuple*\ **->ipv6**)
2438 * Look for an IPv6 socket.
2440 * If the *netns* is a negative signed 32-bit integer, then the
2441 * socket lookup table in the netns associated with the *ctx* will
2442 * will be used. For the TC hooks, this is the netns of the device
2443 * in the skb. For socket hooks, this is the netns of the socket.
2444 * If *netns* is any other signed 32-bit value greater than or
2445 * equal to zero then it specifies the ID of the netns relative to
2446 * the netns associated with the *ctx*. *netns* values beyond the
2447 * range of 32-bit integers are reserved for future use.
2449 * All values for *flags* are reserved for future usage, and must
2452 * This helper is available only if the kernel was compiled with
2453 * **CONFIG_NET** configuration option.
2455 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2456 * For sockets with reuseport option, the **struct bpf_sock**
2457 * result is from *reuse*\ **->socks**\ [] using the hash of the
2460 * int bpf_sk_release(struct bpf_sock *sock)
2462 * Release the reference held by *sock*. *sock* must be a
2463 * non-**NULL** pointer that was returned from
2464 * **bpf_sk_lookup_xxx**\ ().
2466 * 0 on success, or a negative error in case of failure.
2468 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2470 * Push an element *value* in *map*. *flags* is one of:
2473 * If the queue/stack is full, the oldest element is
2474 * removed to make room for this.
2476 * 0 on success, or a negative error in case of failure.
2478 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2480 * Pop an element from *map*.
2482 * 0 on success, or a negative error in case of failure.
2484 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2486 * Get an element from *map* without removing it.
2488 * 0 on success, or a negative error in case of failure.
2490 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2492 * For socket policies, insert *len* bytes into *msg* at offset
2495 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2496 * *msg* it may want to insert metadata or options into the *msg*.
2497 * This can later be read and used by any of the lower layer BPF
2500 * This helper may fail if under memory pressure (a malloc
2501 * fails) in these cases BPF programs will get an appropriate
2502 * error and BPF programs will need to handle them.
2504 * 0 on success, or a negative error in case of failure.
2506 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2508 * Will remove *len* bytes from a *msg* starting at byte *start*.
2509 * This may result in **ENOMEM** errors under certain situations if
2510 * an allocation and copy are required due to a full ring buffer.
2511 * However, the helper will try to avoid doing the allocation
2512 * if possible. Other errors can occur if input parameters are
2513 * invalid either due to *start* byte not being valid part of *msg*
2514 * payload and/or *pop* value being to large.
2516 * 0 on success, or a negative error in case of failure.
2518 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2520 * This helper is used in programs implementing IR decoding, to
2521 * report a successfully decoded pointer movement.
2523 * The *ctx* should point to the lirc sample as passed into
2526 * This helper is only available is the kernel was compiled with
2527 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2532 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2534 * Acquire a spinlock represented by the pointer *lock*, which is
2535 * stored as part of a value of a map. Taking the lock allows to
2536 * safely update the rest of the fields in that value. The
2537 * spinlock can (and must) later be released with a call to
2538 * **bpf_spin_unlock**\ (\ *lock*\ ).
2540 * Spinlocks in BPF programs come with a number of restrictions
2543 * * **bpf_spin_lock** objects are only allowed inside maps of
2544 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2545 * list could be extended in the future).
2546 * * BTF description of the map is mandatory.
2547 * * The BPF program can take ONE lock at a time, since taking two
2548 * or more could cause dead locks.
2549 * * Only one **struct bpf_spin_lock** is allowed per map element.
2550 * * When the lock is taken, calls (either BPF to BPF or helpers)
2552 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2553 * allowed inside a spinlock-ed region.
2554 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2555 * the lock, on all execution paths, before it returns.
2556 * * The BPF program can access **struct bpf_spin_lock** only via
2557 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2558 * helpers. Loading or storing data into the **struct
2559 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2560 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2561 * of the map value must be a struct and have **struct
2562 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2563 * Nested lock inside another struct is not allowed.
2564 * * The **struct bpf_spin_lock** *lock* field in a map value must
2565 * be aligned on a multiple of 4 bytes in that value.
2566 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2567 * the **bpf_spin_lock** field to user space.
2568 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2569 * a BPF program, do not update the **bpf_spin_lock** field.
2570 * * **bpf_spin_lock** cannot be on the stack or inside a
2571 * networking packet (it can only be inside of a map values).
2572 * * **bpf_spin_lock** is available to root only.
2573 * * Tracing programs and socket filter programs cannot use
2574 * **bpf_spin_lock**\ () due to insufficient preemption checks
2575 * (but this may change in the future).
2576 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2580 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2582 * Release the *lock* previously locked by a call to
2583 * **bpf_spin_lock**\ (\ *lock*\ ).
2587 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2589 * This helper gets a **struct bpf_sock** pointer such
2590 * that all the fields in this **bpf_sock** can be accessed.
2592 * A **struct bpf_sock** pointer on success, or **NULL** in
2595 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2597 * This helper gets a **struct bpf_tcp_sock** pointer from a
2598 * **struct bpf_sock** pointer.
2600 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2603 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
2605 * Set ECN (Explicit Congestion Notification) field of IP header
2606 * to **CE** (Congestion Encountered) if current value is **ECT**
2607 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2610 * 1 if the **CE** flag is set (either by the current helper call
2611 * or because it was already present), 0 if it is not set.
2613 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2615 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2616 * **bpf_sk_release**\ () is unnecessary and not allowed.
2618 * A **struct bpf_sock** pointer on success, or **NULL** in
2621 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2623 * Look for TCP socket matching *tuple*, optionally in a child
2624 * network namespace *netns*. The return value must be checked,
2625 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2627 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2628 * that it also returns timewait or request sockets. Use
2629 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2632 * This helper is available only if the kernel was compiled with
2633 * **CONFIG_NET** configuration option.
2635 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2636 * For sockets with reuseport option, the **struct bpf_sock**
2637 * result is from *reuse*\ **->socks**\ [] using the hash of the
2640 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2642 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2643 * the listening socket in *sk*.
2645 * *iph* points to the start of the IPv4 or IPv6 header, while
2646 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2647 * **sizeof**\ (**struct ip6hdr**).
2649 * *th* points to the start of the TCP header, while *th_len*
2650 * contains **sizeof**\ (**struct tcphdr**).
2652 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2655 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2657 * Get name of sysctl in /proc/sys/ and copy it into provided by
2658 * program buffer *buf* of size *buf_len*.
2660 * The buffer is always NUL terminated, unless it's zero-sized.
2662 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2663 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2664 * only (e.g. "tcp_mem").
2666 * Number of character copied (not including the trailing NUL).
2668 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2669 * truncated name in this case).
2671 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2673 * Get current value of sysctl as it is presented in /proc/sys
2674 * (incl. newline, etc), and copy it as a string into provided
2675 * by program buffer *buf* of size *buf_len*.
2677 * The whole value is copied, no matter what file position user
2678 * space issued e.g. sys_read at.
2680 * The buffer is always NUL terminated, unless it's zero-sized.
2682 * Number of character copied (not including the trailing NUL).
2684 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2685 * truncated name in this case).
2687 * **-EINVAL** if current value was unavailable, e.g. because
2688 * sysctl is uninitialized and read returns -EIO for it.
2690 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2692 * Get new value being written by user space to sysctl (before
2693 * the actual write happens) and copy it as a string into
2694 * provided by program buffer *buf* of size *buf_len*.
2696 * User space may write new value at file position > 0.
2698 * The buffer is always NUL terminated, unless it's zero-sized.
2700 * Number of character copied (not including the trailing NUL).
2702 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2703 * truncated name in this case).
2705 * **-EINVAL** if sysctl is being read.
2707 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2709 * Override new value being written by user space to sysctl with
2710 * value provided by program in buffer *buf* of size *buf_len*.
2712 * *buf* should contain a string in same form as provided by user
2713 * space on sysctl write.
2715 * User space may write new value at file position > 0. To override
2716 * the whole sysctl value file position should be set to zero.
2720 * **-E2BIG** if the *buf_len* is too big.
2722 * **-EINVAL** if sysctl is being read.
2724 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2726 * Convert the initial part of the string from buffer *buf* of
2727 * size *buf_len* to a long integer according to the given base
2728 * and save the result in *res*.
2730 * The string may begin with an arbitrary amount of white space
2731 * (as determined by **isspace**\ (3)) followed by a single
2732 * optional '**-**' sign.
2734 * Five least significant bits of *flags* encode base, other bits
2735 * are currently unused.
2737 * Base must be either 8, 10, 16 or 0 to detect it automatically
2738 * similar to user space **strtol**\ (3).
2740 * Number of characters consumed on success. Must be positive but
2741 * no more than *buf_len*.
2743 * **-EINVAL** if no valid digits were found or unsupported base
2746 * **-ERANGE** if resulting value was out of range.
2748 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2750 * Convert the initial part of the string from buffer *buf* of
2751 * size *buf_len* to an unsigned long integer according to the
2752 * given base and save the result in *res*.
2754 * The string may begin with an arbitrary amount of white space
2755 * (as determined by **isspace**\ (3)).
2757 * Five least significant bits of *flags* encode base, other bits
2758 * are currently unused.
2760 * Base must be either 8, 10, 16 or 0 to detect it automatically
2761 * similar to user space **strtoul**\ (3).
2763 * Number of characters consumed on success. Must be positive but
2764 * no more than *buf_len*.
2766 * **-EINVAL** if no valid digits were found or unsupported base
2769 * **-ERANGE** if resulting value was out of range.
2771 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2773 * Get a bpf-local-storage from a *sk*.
2775 * Logically, it could be thought of getting the value from
2776 * a *map* with *sk* as the **key**. From this
2777 * perspective, the usage is not much different from
2778 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2779 * helper enforces the key must be a full socket and the map must
2780 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2782 * Underneath, the value is stored locally at *sk* instead of
2783 * the *map*. The *map* is used as the bpf-local-storage
2784 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2785 * searched against all bpf-local-storages residing at *sk*.
2787 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2788 * used such that a new bpf-local-storage will be
2789 * created if one does not exist. *value* can be used
2790 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2791 * the initial value of a bpf-local-storage. If *value* is
2792 * **NULL**, the new bpf-local-storage will be zero initialized.
2794 * A bpf-local-storage pointer is returned on success.
2796 * **NULL** if not found or there was an error in adding
2797 * a new bpf-local-storage.
2799 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2801 * Delete a bpf-local-storage from a *sk*.
2805 * **-ENOENT** if the bpf-local-storage cannot be found.
2807 * int bpf_send_signal(u32 sig)
2809 * Send signal *sig* to the process of the current task.
2810 * The signal may be delivered to any of this process's threads.
2812 * 0 on success or successfully queued.
2814 * **-EBUSY** if work queue under nmi is full.
2816 * **-EINVAL** if *sig* is invalid.
2818 * **-EPERM** if no permission to send the *sig*.
2820 * **-EAGAIN** if bpf program can try again.
2822 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2824 * Try to issue a SYN cookie for the packet with corresponding
2825 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2827 * *iph* points to the start of the IPv4 or IPv6 header, while
2828 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2829 * **sizeof**\ (**struct ip6hdr**).
2831 * *th* points to the start of the TCP header, while *th_len*
2832 * contains the length of the TCP header.
2834 * On success, lower 32 bits hold the generated SYN cookie in
2835 * followed by 16 bits which hold the MSS value for that cookie,
2836 * and the top 16 bits are unused.
2838 * On failure, the returned value is one of the following:
2840 * **-EINVAL** SYN cookie cannot be issued due to error
2842 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2844 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2846 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2848 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2850 * Write raw *data* blob into a special BPF perf event held by
2851 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2852 * event must have the following attributes: **PERF_SAMPLE_RAW**
2853 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2854 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2856 * The *flags* are used to indicate the index in *map* for which
2857 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2858 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2859 * to indicate that the index of the current CPU core should be
2862 * The value to write, of *size*, is passed through eBPF stack and
2863 * pointed by *data*.
2865 * *ctx* is a pointer to in-kernel struct sk_buff.
2867 * This helper is similar to **bpf_perf_event_output**\ () but
2868 * restricted to raw_tracepoint bpf programs.
2870 * 0 on success, or a negative error in case of failure.
2872 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2874 * Safely attempt to read *size* bytes from user space address
2875 * *unsafe_ptr* and store the data in *dst*.
2877 * 0 on success, or a negative error in case of failure.
2879 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2881 * Safely attempt to read *size* bytes from kernel space address
2882 * *unsafe_ptr* and store the data in *dst*.
2884 * 0 on success, or a negative error in case of failure.
2886 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2888 * Copy a NUL terminated string from an unsafe user address
2889 * *unsafe_ptr* to *dst*. The *size* should include the
2890 * terminating NUL byte. In case the string length is smaller than
2891 * *size*, the target is not padded with further NUL bytes. If the
2892 * string length is larger than *size*, just *size*-1 bytes are
2893 * copied and the last byte is set to NUL.
2895 * On success, the length of the copied string is returned. This
2896 * makes this helper useful in tracing programs for reading
2897 * strings, and more importantly to get its length at runtime. See
2898 * the following snippet:
2902 * SEC("kprobe/sys_open")
2903 * void bpf_sys_open(struct pt_regs *ctx)
2905 * char buf[PATHLEN]; // PATHLEN is defined to 256
2906 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2909 * // Consume buf, for example push it to
2910 * // userspace via bpf_perf_event_output(); we
2911 * // can use res (the string length) as event
2912 * // size, after checking its boundaries.
2915 * In comparison, using **bpf_probe_read_user**\ () helper here
2916 * instead to read the string would require to estimate the length
2917 * at compile time, and would often result in copying more memory
2920 * Another useful use case is when parsing individual process
2921 * arguments or individual environment variables navigating
2922 * *current*\ **->mm->arg_start** and *current*\
2923 * **->mm->env_start**: using this helper and the return value,
2924 * one can quickly iterate at the right offset of the memory area.
2926 * On success, the strictly positive length of the string,
2927 * including the trailing NUL character. On error, a negative
2930 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2932 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
2933 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
2935 * On success, the strictly positive length of the string, including
2936 * the trailing NUL character. On error, a negative value.
2938 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2940 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
2941 * *rcv_nxt* is the ack_seq to be sent out.
2943 * 0 on success, or a negative error in case of failure.
2945 * int bpf_send_signal_thread(u32 sig)
2947 * Send signal *sig* to the thread corresponding to the current task.
2949 * 0 on success or successfully queued.
2951 * **-EBUSY** if work queue under nmi is full.
2953 * **-EINVAL** if *sig* is invalid.
2955 * **-EPERM** if no permission to send the *sig*.
2957 * **-EAGAIN** if bpf program can try again.
2959 * u64 bpf_jiffies64(void)
2961 * Obtain the 64bit jiffies
2963 * The 64 bit jiffies
2965 * int bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
2967 * For an eBPF program attached to a perf event, retrieve the
2968 * branch records (**struct perf_branch_entry**) associated to *ctx*
2969 * and store it in the buffer pointed by *buf* up to size
2972 * On success, number of bytes written to *buf*. On error, a
2975 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
2976 * instead return the number of bytes required to store all the
2977 * branch entries. If this flag is set, *buf* may be NULL.
2979 * **-EINVAL** if arguments invalid or **size** not a multiple
2980 * of **sizeof**\ (**struct perf_branch_entry**\ ).
2982 * **-ENOENT** if architecture does not support branch records.
2984 * int bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
2986 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
2987 * *namespace* will be returned in *nsdata*.
2989 * 0 on success, or one of the following in case of failure:
2991 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
2992 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
2994 * **-ENOENT** if pidns does not exists for the current task.
2996 * int bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2998 * Write raw *data* blob into a special BPF perf event held by
2999 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3000 * event must have the following attributes: **PERF_SAMPLE_RAW**
3001 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3002 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3004 * The *flags* are used to indicate the index in *map* for which
3005 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3006 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3007 * to indicate that the index of the current CPU core should be
3010 * The value to write, of *size*, is passed through eBPF stack and
3011 * pointed by *data*.
3013 * *ctx* is a pointer to in-kernel struct xdp_buff.
3015 * This helper is similar to **bpf_perf_eventoutput**\ () but
3016 * restricted to raw_tracepoint bpf programs.
3018 * 0 on success, or a negative error in case of failure.
3020 * u64 bpf_get_netns_cookie(void *ctx)
3022 * Retrieve the cookie (generated by the kernel) of the network
3023 * namespace the input *ctx* is associated with. The network
3024 * namespace cookie remains stable for its lifetime and provides
3025 * a global identifier that can be assumed unique. If *ctx* is
3026 * NULL, then the helper returns the cookie for the initial
3027 * network namespace. The cookie itself is very similar to that
3028 * of **bpf_get_socket_cookie**\ () helper, but for network
3029 * namespaces instead of sockets.
3031 * A 8-byte long opaque number.
3033 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3035 * Return id of cgroup v2 that is ancestor of the cgroup associated
3036 * with the current task at the *ancestor_level*. The root cgroup
3037 * is at *ancestor_level* zero and each step down the hierarchy
3038 * increments the level. If *ancestor_level* == level of cgroup
3039 * associated with the current task, then return value will be the
3040 * same as that of **bpf_get_current_cgroup_id**\ ().
3042 * The helper is useful to implement policies based on cgroups
3043 * that are upper in hierarchy than immediate cgroup associated
3044 * with the current task.
3046 * The format of returned id and helper limitations are same as in
3047 * **bpf_get_current_cgroup_id**\ ().
3049 * The id is returned or 0 in case the id could not be retrieved.
3051 * int bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags)
3053 * Assign the *sk* to the *skb*. When combined with appropriate
3054 * routing configuration to receive the packet towards the socket,
3055 * will cause *skb* to be delivered to the specified socket.
3056 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3057 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3058 * interfere with successful delivery to the socket.
3060 * This operation is only valid from TC ingress path.
3062 * The *flags* argument must be zero.
3064 * 0 on success, or a negative error in case of failure:
3066 * **-EINVAL** if specified *flags* are not supported.
3068 * **-ENOENT** if the socket is unavailable for assignment.
3070 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3072 * **-EOPNOTSUPP** if the operation is not supported, for example
3073 * a call from outside of TC ingress.
3075 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3078 * u64 bpf_ktime_get_boot_ns(void)
3080 * Return the time elapsed since system boot, in nanoseconds.
3081 * Does include the time the system was suspended.
3082 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
3086 * int bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3088 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3089 * out the format string.
3090 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3091 * the format string itself. The *data* and *data_len* are format string
3092 * arguments. The *data* are a **u64** array and corresponding format string
3093 * values are stored in the array. For strings and pointers where pointees
3094 * are accessed, only the pointer values are stored in the *data* array.
3095 * The *data_len* is the size of *data* in bytes.
3097 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3098 * Reading kernel memory may fail due to either invalid address or
3099 * valid address but requiring a major memory fault. If reading kernel memory
3100 * fails, the string for **%s** will be an empty string, and the ip
3101 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
3102 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
3104 * 0 on success, or a negative error in case of failure:
3106 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3107 * by returning 1 from bpf program.
3109 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3111 * **-E2BIG** if *fmt* contains too many format specifiers.
3113 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3115 * int bpf_seq_write(struct seq_file *m, const void *data, u32 len)
3117 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
3118 * The *m* represents the seq_file. The *data* and *len* represent the
3119 * data to write in bytes.
3121 * 0 on success, or a negative error in case of failure:
3123 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3125 #define __BPF_FUNC_MAPPER(FN) \
3127 FN(map_lookup_elem), \
3128 FN(map_update_elem), \
3129 FN(map_delete_elem), \
3133 FN(get_prandom_u32), \
3134 FN(get_smp_processor_id), \
3135 FN(skb_store_bytes), \
3136 FN(l3_csum_replace), \
3137 FN(l4_csum_replace), \
3139 FN(clone_redirect), \
3140 FN(get_current_pid_tgid), \
3141 FN(get_current_uid_gid), \
3142 FN(get_current_comm), \
3143 FN(get_cgroup_classid), \
3144 FN(skb_vlan_push), \
3146 FN(skb_get_tunnel_key), \
3147 FN(skb_set_tunnel_key), \
3148 FN(perf_event_read), \
3150 FN(get_route_realm), \
3151 FN(perf_event_output), \
3152 FN(skb_load_bytes), \
3155 FN(skb_get_tunnel_opt), \
3156 FN(skb_set_tunnel_opt), \
3157 FN(skb_change_proto), \
3158 FN(skb_change_type), \
3159 FN(skb_under_cgroup), \
3160 FN(get_hash_recalc), \
3161 FN(get_current_task), \
3162 FN(probe_write_user), \
3163 FN(current_task_under_cgroup), \
3164 FN(skb_change_tail), \
3165 FN(skb_pull_data), \
3167 FN(set_hash_invalid), \
3168 FN(get_numa_node_id), \
3169 FN(skb_change_head), \
3170 FN(xdp_adjust_head), \
3171 FN(probe_read_str), \
3172 FN(get_socket_cookie), \
3173 FN(get_socket_uid), \
3176 FN(skb_adjust_room), \
3178 FN(sk_redirect_map), \
3179 FN(sock_map_update), \
3180 FN(xdp_adjust_meta), \
3181 FN(perf_event_read_value), \
3182 FN(perf_prog_read_value), \
3184 FN(override_return), \
3185 FN(sock_ops_cb_flags_set), \
3186 FN(msg_redirect_map), \
3187 FN(msg_apply_bytes), \
3188 FN(msg_cork_bytes), \
3189 FN(msg_pull_data), \
3191 FN(xdp_adjust_tail), \
3192 FN(skb_get_xfrm_state), \
3194 FN(skb_load_bytes_relative), \
3196 FN(sock_hash_update), \
3197 FN(msg_redirect_hash), \
3198 FN(sk_redirect_hash), \
3199 FN(lwt_push_encap), \
3200 FN(lwt_seg6_store_bytes), \
3201 FN(lwt_seg6_adjust_srh), \
3202 FN(lwt_seg6_action), \
3205 FN(skb_cgroup_id), \
3206 FN(get_current_cgroup_id), \
3207 FN(get_local_storage), \
3208 FN(sk_select_reuseport), \
3209 FN(skb_ancestor_cgroup_id), \
3210 FN(sk_lookup_tcp), \
3211 FN(sk_lookup_udp), \
3213 FN(map_push_elem), \
3215 FN(map_peek_elem), \
3216 FN(msg_push_data), \
3218 FN(rc_pointer_rel), \
3223 FN(skb_ecn_set_ce), \
3224 FN(get_listener_sock), \
3225 FN(skc_lookup_tcp), \
3226 FN(tcp_check_syncookie), \
3227 FN(sysctl_get_name), \
3228 FN(sysctl_get_current_value), \
3229 FN(sysctl_get_new_value), \
3230 FN(sysctl_set_new_value), \
3233 FN(sk_storage_get), \
3234 FN(sk_storage_delete), \
3236 FN(tcp_gen_syncookie), \
3238 FN(probe_read_user), \
3239 FN(probe_read_kernel), \
3240 FN(probe_read_user_str), \
3241 FN(probe_read_kernel_str), \
3243 FN(send_signal_thread), \
3245 FN(read_branch_records), \
3246 FN(get_ns_current_pid_tgid), \
3248 FN(get_netns_cookie), \
3249 FN(get_current_ancestor_cgroup_id), \
3251 FN(ktime_get_boot_ns), \
3255 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
3256 * function eBPF program intends to call
3258 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
3260 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
3263 #undef __BPF_ENUM_FN
3265 /* All flags used by eBPF helper functions, placed here. */
3267 /* BPF_FUNC_skb_store_bytes flags. */
3269 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3270 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3273 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3274 * First 4 bits are for passing the header field size.
3277 BPF_F_HDR_FIELD_MASK = 0xfULL,
3280 /* BPF_FUNC_l4_csum_replace flags. */
3282 BPF_F_PSEUDO_HDR = (1ULL << 4),
3283 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3284 BPF_F_MARK_ENFORCE = (1ULL << 6),
3287 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
3289 BPF_F_INGRESS = (1ULL << 0),
3292 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
3294 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3297 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
3299 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3300 BPF_F_USER_STACK = (1ULL << 8),
3301 /* flags used by BPF_FUNC_get_stackid only. */
3302 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3303 BPF_F_REUSE_STACKID = (1ULL << 10),
3304 /* flags used by BPF_FUNC_get_stack only. */
3305 BPF_F_USER_BUILD_ID = (1ULL << 11),
3308 /* BPF_FUNC_skb_set_tunnel_key flags. */
3310 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3311 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3312 BPF_F_SEQ_NUMBER = (1ULL << 3),
3315 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3316 * BPF_FUNC_perf_event_read_value flags.
3319 BPF_F_INDEX_MASK = 0xffffffffULL,
3320 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
3321 /* BPF_FUNC_perf_event_output for sk_buff input context. */
3322 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3325 /* Current network namespace */
3327 BPF_F_CURRENT_NETNS = (-1L),
3330 /* BPF_FUNC_skb_adjust_room flags. */
3332 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3333 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3334 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3335 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3336 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
3340 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3341 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3344 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
3345 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3346 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
3348 /* BPF_FUNC_sysctl_get_name flags. */
3350 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3353 /* BPF_FUNC_sk_storage_get flags */
3355 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0),
3358 /* BPF_FUNC_read_branch_records flags. */
3360 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3363 /* Mode for BPF_FUNC_skb_adjust_room helper. */
3364 enum bpf_adj_room_mode {
3369 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3370 enum bpf_hdr_start_off {
3375 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3376 enum bpf_lwt_encap_mode {
3378 BPF_LWT_ENCAP_SEG6_INLINE,
3382 #define __bpf_md_ptr(type, name) \
3386 } __attribute__((aligned(8)))
3388 /* user accessible mirror of in-kernel sk_buff.
3389 * new fields can only be added to the end of this structure
3395 __u32 queue_mapping;
3401 __u32 ingress_ifindex;
3411 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
3413 __u32 remote_ip4; /* Stored in network byte order */
3414 __u32 local_ip4; /* Stored in network byte order */
3415 __u32 remote_ip6[4]; /* Stored in network byte order */
3416 __u32 local_ip6[4]; /* Stored in network byte order */
3417 __u32 remote_port; /* Stored in network byte order */
3418 __u32 local_port; /* stored in host byte order */
3422 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
3426 __bpf_md_ptr(struct bpf_sock *, sk);
3430 struct bpf_tunnel_key {
3434 __u32 remote_ipv6[4];
3438 __u16 tunnel_ext; /* Padding, future use. */
3442 /* user accessible mirror of in-kernel xfrm_state.
3443 * new fields can only be added to the end of this structure
3445 struct bpf_xfrm_state {
3447 __u32 spi; /* Stored in network byte order */
3449 __u16 ext; /* Padding, future use. */
3451 __u32 remote_ipv4; /* Stored in network byte order */
3452 __u32 remote_ipv6[4]; /* Stored in network byte order */
3456 /* Generic BPF return codes which all BPF program types may support.
3457 * The values are binary compatible with their TC_ACT_* counter-part to
3458 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3461 * XDP is handled seprately, see XDP_*.
3469 /* >127 are reserved for prog type specific return codes.
3471 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3472 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3473 * changed and should be routed based on its new L3 header.
3474 * (This is an L3 redirect, as opposed to L2 redirect
3475 * represented by BPF_REDIRECT above).
3477 BPF_LWT_REROUTE = 128,
3487 /* IP address also allows 1 and 2 bytes access */
3490 __u32 src_port; /* host byte order */
3491 __u32 dst_port; /* network byte order */
3497 struct bpf_tcp_sock {
3498 __u32 snd_cwnd; /* Sending congestion window */
3499 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3501 __u32 snd_ssthresh; /* Slow start size threshold */
3502 __u32 rcv_nxt; /* What we want to receive next */
3503 __u32 snd_nxt; /* Next sequence we send */
3504 __u32 snd_una; /* First byte we want an ack for */
3505 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3506 __u32 ecn_flags; /* ECN status bits. */
3507 __u32 rate_delivered; /* saved rate sample: packets delivered */
3508 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3509 __u32 packets_out; /* Packets which are "in flight" */
3510 __u32 retrans_out; /* Retransmitted packets out */
3511 __u32 total_retrans; /* Total retransmits for entire connection */
3512 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3513 * total number of segments in.
3515 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3516 * total number of data segments in.
3518 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3519 * The total number of segments sent.
3521 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3522 * total number of data segments sent.
3524 __u32 lost_out; /* Lost packets */
3525 __u32 sacked_out; /* SACK'd packets */
3526 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3527 * sum(delta(rcv_nxt)), or how many bytes
3530 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3531 * sum(delta(snd_una)), or how many bytes
3534 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3535 * total number of DSACK blocks received
3537 __u32 delivered; /* Total data packets delivered incl. rexmits */
3538 __u32 delivered_ce; /* Like the above but only ECE marked packets */
3539 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
3542 struct bpf_sock_tuple {
3559 struct bpf_xdp_sock {
3563 #define XDP_PACKET_HEADROOM 256
3565 /* User return codes for XDP prog type.
3566 * A valid XDP program must return one of these defined values. All other
3567 * return codes are reserved for future use. Unknown return codes will
3568 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3578 /* user accessible metadata for XDP packet hook
3579 * new fields must be added to the end of this structure
3585 /* Below access go through struct xdp_rxq_info */
3586 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3587 __u32 rx_queue_index; /* rxq->queue_index */
3595 /* user accessible metadata for SK_MSG packet hook, new fields must
3596 * be added to the end of this structure
3599 __bpf_md_ptr(void *, data);
3600 __bpf_md_ptr(void *, data_end);
3603 __u32 remote_ip4; /* Stored in network byte order */
3604 __u32 local_ip4; /* Stored in network byte order */
3605 __u32 remote_ip6[4]; /* Stored in network byte order */
3606 __u32 local_ip6[4]; /* Stored in network byte order */
3607 __u32 remote_port; /* Stored in network byte order */
3608 __u32 local_port; /* stored in host byte order */
3609 __u32 size; /* Total size of sk_msg */
3612 struct sk_reuseport_md {
3614 * Start of directly accessible data. It begins from
3615 * the tcp/udp header.
3617 __bpf_md_ptr(void *, data);
3618 /* End of directly accessible data */
3619 __bpf_md_ptr(void *, data_end);
3621 * Total length of packet (starting from the tcp/udp header).
3622 * Note that the directly accessible bytes (data_end - data)
3623 * could be less than this "len". Those bytes could be
3624 * indirectly read by a helper "bpf_skb_load_bytes()".
3628 * Eth protocol in the mac header (network byte order). e.g.
3629 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3632 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3633 __u32 bind_inany; /* Is sock bound to an INANY address? */
3634 __u32 hash; /* A hash of the packet 4 tuples */
3637 #define BPF_TAG_SIZE 8
3639 struct bpf_prog_info {
3642 __u8 tag[BPF_TAG_SIZE];
3643 __u32 jited_prog_len;
3644 __u32 xlated_prog_len;
3645 __aligned_u64 jited_prog_insns;
3646 __aligned_u64 xlated_prog_insns;
3647 __u64 load_time; /* ns since boottime */
3648 __u32 created_by_uid;
3650 __aligned_u64 map_ids;
3651 char name[BPF_OBJ_NAME_LEN];
3653 __u32 gpl_compatible:1;
3654 __u32 :31; /* alignment pad */
3657 __u32 nr_jited_ksyms;
3658 __u32 nr_jited_func_lens;
3659 __aligned_u64 jited_ksyms;
3660 __aligned_u64 jited_func_lens;
3662 __u32 func_info_rec_size;
3663 __aligned_u64 func_info;
3666 __aligned_u64 line_info;
3667 __aligned_u64 jited_line_info;
3668 __u32 nr_jited_line_info;
3669 __u32 line_info_rec_size;
3670 __u32 jited_line_info_rec_size;
3672 __aligned_u64 prog_tags;
3675 } __attribute__((aligned(8)));
3677 struct bpf_map_info {
3684 char name[BPF_OBJ_NAME_LEN];
3686 __u32 btf_vmlinux_value_type_id;
3690 __u32 btf_key_type_id;
3691 __u32 btf_value_type_id;
3692 } __attribute__((aligned(8)));
3694 struct bpf_btf_info {
3698 } __attribute__((aligned(8)));
3700 struct bpf_link_info {
3706 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
3707 __u32 tp_name_len; /* in/out: tp_name buffer len */
3717 } __attribute__((aligned(8)));
3719 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3720 * by user and intended to be used by socket (e.g. to bind to, depends on
3721 * attach attach type).
3723 struct bpf_sock_addr {
3724 __u32 user_family; /* Allows 4-byte read, but no write. */
3725 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3726 * Stored in network byte order.
3728 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3729 * Stored in network byte order.
3731 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
3732 * Stored in network byte order
3734 __u32 family; /* Allows 4-byte read, but no write */
3735 __u32 type; /* Allows 4-byte read, but no write */
3736 __u32 protocol; /* Allows 4-byte read, but no write */
3737 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3738 * Stored in network byte order.
3740 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3741 * Stored in network byte order.
3743 __bpf_md_ptr(struct bpf_sock *, sk);
3746 /* User bpf_sock_ops struct to access socket values and specify request ops
3747 * and their replies.
3748 * Some of this fields are in network (bigendian) byte order and may need
3749 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3750 * New fields can only be added at the end of this structure
3752 struct bpf_sock_ops {
3755 __u32 args[4]; /* Optionally passed to bpf program */
3756 __u32 reply; /* Returned by bpf program */
3757 __u32 replylong[4]; /* Optionally returned by bpf prog */
3760 __u32 remote_ip4; /* Stored in network byte order */
3761 __u32 local_ip4; /* Stored in network byte order */
3762 __u32 remote_ip6[4]; /* Stored in network byte order */
3763 __u32 local_ip6[4]; /* Stored in network byte order */
3764 __u32 remote_port; /* Stored in network byte order */
3765 __u32 local_port; /* stored in host byte order */
3766 __u32 is_fullsock; /* Some TCP fields are only valid if
3767 * there is a full socket. If not, the
3768 * fields read as zero.
3771 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
3772 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3781 __u32 rate_delivered;
3782 __u32 rate_interval_us;
3785 __u32 total_retrans;
3789 __u32 data_segs_out;
3793 __u64 bytes_received;
3795 __bpf_md_ptr(struct bpf_sock *, sk);
3798 /* Definitions for bpf_sock_ops_cb_flags */
3800 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
3801 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
3802 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
3803 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
3804 /* Mask of all currently supported cb flags */
3805 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF,
3808 /* List of known BPF sock_ops operators.
3809 * New entries can only be added at the end
3813 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3814 * -1 if default value should be used
3816 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3817 * window (in packets) or -1 if default
3818 * value should be used
3820 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3821 * active connection is initialized
3823 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3824 * active connection is
3827 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3828 * passive connection is
3831 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3834 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3835 * based on the path and may be
3836 * dependent on the congestion control
3837 * algorithm. In general it indicates
3838 * a congestion threshold. RTTs above
3839 * this indicate congestion
3841 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3842 * Arg1: value of icsk_retransmits
3843 * Arg2: value of icsk_rto
3844 * Arg3: whether RTO has expired
3846 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3847 * Arg1: sequence number of 1st byte
3849 * Arg3: return value of
3850 * tcp_transmit_skb (0 => success)
3852 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3856 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3857 * socket transition to LISTEN state.
3859 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3863 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3864 * changes between the TCP and BPF versions. Ideally this should never happen.
3865 * If it does, we need to add code to convert them before calling
3866 * the BPF sock_ops function.
3869 BPF_TCP_ESTABLISHED = 1,
3879 BPF_TCP_CLOSING, /* Now a valid state */
3880 BPF_TCP_NEW_SYN_RECV,
3882 BPF_TCP_MAX_STATES /* Leave at the end! */
3886 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
3887 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
3890 struct bpf_perf_event_value {
3897 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
3898 BPF_DEVCG_ACC_READ = (1ULL << 1),
3899 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
3903 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
3904 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
3907 struct bpf_cgroup_dev_ctx {
3908 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3914 struct bpf_raw_tracepoint_args {
3918 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3919 * OUTPUT: Do lookup from egress perspective; default is ingress
3922 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
3923 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
3927 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3928 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3929 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3930 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3931 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3932 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3933 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3934 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3935 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3938 struct bpf_fib_lookup {
3939 /* input: network family for lookup (AF_INET, AF_INET6)
3940 * output: network family of egress nexthop
3944 /* set if lookup is to consider L4 data - e.g., FIB rules */
3949 /* total length of packet from network header - used for MTU check */
3952 /* input: L3 device index for lookup
3953 * output: device index from FIB lookup
3958 /* inputs to lookup */
3959 __u8 tos; /* AF_INET */
3960 __be32 flowinfo; /* AF_INET6, flow_label + priority */
3962 /* output: metric of fib result (IPv4/IPv6 only) */
3968 __u32 ipv6_src[4]; /* in6_addr; network order */
3971 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3972 * network header. output: bpf_fib_lookup sets to gateway address
3973 * if FIB lookup returns gateway route
3977 __u32 ipv6_dst[4]; /* in6_addr; network order */
3981 __be16 h_vlan_proto;
3983 __u8 smac[6]; /* ETH_ALEN */
3984 __u8 dmac[6]; /* ETH_ALEN */
3987 enum bpf_task_fd_type {
3988 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3989 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3990 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3991 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3992 BPF_FD_TYPE_UPROBE, /* filename + offset */
3993 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3997 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
3998 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
3999 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
4002 struct bpf_flow_keys {
4005 __u16 addr_proto; /* ETH_P_* of valid addrs */
4019 __u32 ipv6_src[4]; /* in6_addr; network order */
4020 __u32 ipv6_dst[4]; /* in6_addr; network order */
4027 struct bpf_func_info {
4032 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
4033 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
4035 struct bpf_line_info {
4037 __u32 file_name_off;
4042 struct bpf_spin_lock {
4047 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
4048 * Allows 1,2,4-byte read, but no write.
4050 __u32 file_pos; /* Sysctl file position to read from, write to.
4051 * Allows 1,2,4-byte read an 4-byte write.
4055 struct bpf_sockopt {
4056 __bpf_md_ptr(struct bpf_sock *, sk);
4057 __bpf_md_ptr(void *, optval);
4058 __bpf_md_ptr(void *, optval_end);
4066 struct bpf_pidns_info {
4070 #endif /* _UAPI__LINUX_BPF_H__ */