1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_XADD 0xc0 /* exclusive add */
25 #define BPF_MOV 0xb0 /* mov reg to reg */
26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28 /* change endianness of a register */
29 #define BPF_END 0xd0 /* flags for endianness conversion: */
30 #define BPF_TO_LE 0x00 /* convert to little-endian */
31 #define BPF_TO_BE 0x08 /* convert to big-endian */
32 #define BPF_FROM_LE BPF_TO_LE
33 #define BPF_FROM_BE BPF_TO_BE
36 #define BPF_JNE 0x50 /* jump != */
37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
43 #define BPF_CALL 0x80 /* function call */
44 #define BPF_EXIT 0x90 /* function return */
46 /* Register numbers */
62 /* BPF has 10 general purpose 64-bit registers and stack frame. */
63 #define MAX_BPF_REG __MAX_BPF_REG
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74 struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data[0]; /* Arbitrary size */
79 struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
84 union bpf_iter_link_info {
90 /* BPF syscall commands, see bpf(2) man-page for details. */
103 BPF_PROG_GET_NEXT_ID,
105 BPF_PROG_GET_FD_BY_ID,
106 BPF_MAP_GET_FD_BY_ID,
107 BPF_OBJ_GET_INFO_BY_FD,
109 BPF_RAW_TRACEPOINT_OPEN,
111 BPF_BTF_GET_FD_BY_ID,
113 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
116 BPF_MAP_LOOKUP_BATCH,
117 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
118 BPF_MAP_UPDATE_BATCH,
119 BPF_MAP_DELETE_BATCH,
122 BPF_LINK_GET_FD_BY_ID,
123 BPF_LINK_GET_NEXT_ID,
133 BPF_MAP_TYPE_PROG_ARRAY,
134 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
135 BPF_MAP_TYPE_PERCPU_HASH,
136 BPF_MAP_TYPE_PERCPU_ARRAY,
137 BPF_MAP_TYPE_STACK_TRACE,
138 BPF_MAP_TYPE_CGROUP_ARRAY,
139 BPF_MAP_TYPE_LRU_HASH,
140 BPF_MAP_TYPE_LRU_PERCPU_HASH,
141 BPF_MAP_TYPE_LPM_TRIE,
142 BPF_MAP_TYPE_ARRAY_OF_MAPS,
143 BPF_MAP_TYPE_HASH_OF_MAPS,
145 BPF_MAP_TYPE_SOCKMAP,
148 BPF_MAP_TYPE_SOCKHASH,
149 BPF_MAP_TYPE_CGROUP_STORAGE,
150 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
151 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
154 BPF_MAP_TYPE_SK_STORAGE,
155 BPF_MAP_TYPE_DEVMAP_HASH,
156 BPF_MAP_TYPE_STRUCT_OPS,
157 BPF_MAP_TYPE_RINGBUF,
158 BPF_MAP_TYPE_INODE_STORAGE,
161 /* Note that tracing related programs such as
162 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
163 * are not subject to a stable API since kernel internal data
164 * structures can change from release to release and may
165 * therefore break existing tracing BPF programs. Tracing BPF
166 * programs correspond to /a/ specific kernel which is to be
167 * analyzed, and not /a/ specific kernel /and/ all future ones.
170 BPF_PROG_TYPE_UNSPEC,
171 BPF_PROG_TYPE_SOCKET_FILTER,
172 BPF_PROG_TYPE_KPROBE,
173 BPF_PROG_TYPE_SCHED_CLS,
174 BPF_PROG_TYPE_SCHED_ACT,
175 BPF_PROG_TYPE_TRACEPOINT,
177 BPF_PROG_TYPE_PERF_EVENT,
178 BPF_PROG_TYPE_CGROUP_SKB,
179 BPF_PROG_TYPE_CGROUP_SOCK,
180 BPF_PROG_TYPE_LWT_IN,
181 BPF_PROG_TYPE_LWT_OUT,
182 BPF_PROG_TYPE_LWT_XMIT,
183 BPF_PROG_TYPE_SOCK_OPS,
184 BPF_PROG_TYPE_SK_SKB,
185 BPF_PROG_TYPE_CGROUP_DEVICE,
186 BPF_PROG_TYPE_SK_MSG,
187 BPF_PROG_TYPE_RAW_TRACEPOINT,
188 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
189 BPF_PROG_TYPE_LWT_SEG6LOCAL,
190 BPF_PROG_TYPE_LIRC_MODE2,
191 BPF_PROG_TYPE_SK_REUSEPORT,
192 BPF_PROG_TYPE_FLOW_DISSECTOR,
193 BPF_PROG_TYPE_CGROUP_SYSCTL,
194 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
195 BPF_PROG_TYPE_CGROUP_SOCKOPT,
196 BPF_PROG_TYPE_TRACING,
197 BPF_PROG_TYPE_STRUCT_OPS,
200 BPF_PROG_TYPE_SK_LOOKUP,
203 enum bpf_attach_type {
204 BPF_CGROUP_INET_INGRESS,
205 BPF_CGROUP_INET_EGRESS,
206 BPF_CGROUP_INET_SOCK_CREATE,
208 BPF_SK_SKB_STREAM_PARSER,
209 BPF_SK_SKB_STREAM_VERDICT,
212 BPF_CGROUP_INET4_BIND,
213 BPF_CGROUP_INET6_BIND,
214 BPF_CGROUP_INET4_CONNECT,
215 BPF_CGROUP_INET6_CONNECT,
216 BPF_CGROUP_INET4_POST_BIND,
217 BPF_CGROUP_INET6_POST_BIND,
218 BPF_CGROUP_UDP4_SENDMSG,
219 BPF_CGROUP_UDP6_SENDMSG,
223 BPF_CGROUP_UDP4_RECVMSG,
224 BPF_CGROUP_UDP6_RECVMSG,
225 BPF_CGROUP_GETSOCKOPT,
226 BPF_CGROUP_SETSOCKOPT,
233 BPF_CGROUP_INET4_GETPEERNAME,
234 BPF_CGROUP_INET6_GETPEERNAME,
235 BPF_CGROUP_INET4_GETSOCKNAME,
236 BPF_CGROUP_INET6_GETSOCKNAME,
238 BPF_CGROUP_INET_SOCK_RELEASE,
242 __MAX_BPF_ATTACH_TYPE
245 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
248 BPF_LINK_TYPE_UNSPEC = 0,
249 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
250 BPF_LINK_TYPE_TRACING = 2,
251 BPF_LINK_TYPE_CGROUP = 3,
252 BPF_LINK_TYPE_ITER = 4,
253 BPF_LINK_TYPE_NETNS = 5,
254 BPF_LINK_TYPE_XDP = 6,
259 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
261 * NONE(default): No further bpf programs allowed in the subtree.
263 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
264 * the program in this cgroup yields to sub-cgroup program.
266 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
267 * that cgroup program gets run in addition to the program in this cgroup.
269 * Only one program is allowed to be attached to a cgroup with
270 * NONE or BPF_F_ALLOW_OVERRIDE flag.
271 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
272 * release old program and attach the new one. Attach flags has to match.
274 * Multiple programs are allowed to be attached to a cgroup with
275 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
276 * (those that were attached first, run first)
277 * The programs of sub-cgroup are executed first, then programs of
278 * this cgroup and then programs of parent cgroup.
279 * When children program makes decision (like picking TCP CA or sock bind)
280 * parent program has a chance to override it.
282 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
283 * programs for a cgroup. Though it's possible to replace an old program at
284 * any position by also specifying BPF_F_REPLACE flag and position itself in
285 * replace_bpf_fd attribute. Old program at this position will be released.
287 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
288 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
290 * cgrp1 (MULTI progs A, B) ->
291 * cgrp2 (OVERRIDE prog C) ->
292 * cgrp3 (MULTI prog D) ->
293 * cgrp4 (OVERRIDE prog E) ->
294 * cgrp5 (NONE prog F)
295 * the event in cgrp5 triggers execution of F,D,A,B in that order.
296 * if prog F is detached, the execution is E,D,A,B
297 * if prog F and D are detached, the execution is E,A,B
298 * if prog F, E and D are detached, the execution is C,A,B
300 * All eligible programs are executed regardless of return code from
303 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
304 #define BPF_F_ALLOW_MULTI (1U << 1)
305 #define BPF_F_REPLACE (1U << 2)
307 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
308 * verifier will perform strict alignment checking as if the kernel
309 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
310 * and NET_IP_ALIGN defined to 2.
312 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
314 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
315 * verifier will allow any alignment whatsoever. On platforms
316 * with strict alignment requirements for loads ands stores (such
317 * as sparc and mips) the verifier validates that all loads and
318 * stores provably follow this requirement. This flag turns that
319 * checking and enforcement off.
321 * It is mostly used for testing when we want to validate the
322 * context and memory access aspects of the verifier, but because
323 * of an unaligned access the alignment check would trigger before
324 * the one we are interested in.
326 #define BPF_F_ANY_ALIGNMENT (1U << 1)
328 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
329 * Verifier does sub-register def/use analysis and identifies instructions whose
330 * def only matters for low 32-bit, high 32-bit is never referenced later
331 * through implicit zero extension. Therefore verifier notifies JIT back-ends
332 * that it is safe to ignore clearing high 32-bit for these instructions. This
333 * saves some back-ends a lot of code-gen. However such optimization is not
334 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
335 * hence hasn't used verifier's analysis result. But, we really want to have a
336 * way to be able to verify the correctness of the described optimization on
337 * x86_64 on which testsuites are frequently exercised.
339 * So, this flag is introduced. Once it is set, verifier will randomize high
340 * 32-bit for those instructions who has been identified as safe to ignore them.
341 * Then, if verifier is not doing correct analysis, such randomization will
342 * regress tests to expose bugs.
344 #define BPF_F_TEST_RND_HI32 (1U << 2)
346 /* The verifier internal test flag. Behavior is undefined */
347 #define BPF_F_TEST_STATE_FREQ (1U << 3)
349 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
350 * restrict map and helper usage for such programs. Sleepable BPF programs can
351 * only be attached to hooks where kernel execution context allows sleeping.
352 * Such programs are allowed to use helpers that may sleep like
353 * bpf_copy_from_user().
355 #define BPF_F_SLEEPABLE (1U << 4)
357 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
360 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
361 * insn[0].imm: map fd map fd
362 * insn[1].imm: 0 offset into value
365 * ldimm64 rewrite: address of map address of map[0]+offset
366 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
368 #define BPF_PSEUDO_MAP_FD 1
369 #define BPF_PSEUDO_MAP_VALUE 2
371 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
372 * offset to another bpf function
374 #define BPF_PSEUDO_CALL 1
376 /* flags for BPF_MAP_UPDATE_ELEM command */
378 BPF_ANY = 0, /* create new element or update existing */
379 BPF_NOEXIST = 1, /* create new element if it didn't exist */
380 BPF_EXIST = 2, /* update existing element */
381 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
384 /* flags for BPF_MAP_CREATE command */
386 BPF_F_NO_PREALLOC = (1U << 0),
387 /* Instead of having one common LRU list in the
388 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
389 * which can scale and perform better.
390 * Note, the LRU nodes (including free nodes) cannot be moved
391 * across different LRU lists.
393 BPF_F_NO_COMMON_LRU = (1U << 1),
394 /* Specify numa node during map creation */
395 BPF_F_NUMA_NODE = (1U << 2),
397 /* Flags for accessing BPF object from syscall side. */
398 BPF_F_RDONLY = (1U << 3),
399 BPF_F_WRONLY = (1U << 4),
401 /* Flag for stack_map, store build_id+offset instead of pointer */
402 BPF_F_STACK_BUILD_ID = (1U << 5),
404 /* Zero-initialize hash function seed. This should only be used for testing. */
405 BPF_F_ZERO_SEED = (1U << 6),
407 /* Flags for accessing BPF object from program side. */
408 BPF_F_RDONLY_PROG = (1U << 7),
409 BPF_F_WRONLY_PROG = (1U << 8),
411 /* Clone map from listener for newly accepted socket */
412 BPF_F_CLONE = (1U << 9),
414 /* Enable memory-mapping BPF map */
415 BPF_F_MMAPABLE = (1U << 10),
418 /* Flags for BPF_PROG_QUERY. */
420 /* Query effective (directly attached + inherited from ancestor cgroups)
421 * programs that will be executed for events within a cgroup.
422 * attach_flags with this flag are returned only for directly attached programs.
424 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
426 /* type for BPF_ENABLE_STATS */
427 enum bpf_stats_type {
428 /* enabled run_time_ns and run_cnt */
429 BPF_STATS_RUN_TIME = 0,
432 enum bpf_stack_build_id_status {
433 /* user space need an empty entry to identify end of a trace */
434 BPF_STACK_BUILD_ID_EMPTY = 0,
435 /* with valid build_id and offset */
436 BPF_STACK_BUILD_ID_VALID = 1,
437 /* couldn't get build_id, fallback to ip */
438 BPF_STACK_BUILD_ID_IP = 2,
441 #define BPF_BUILD_ID_SIZE 20
442 struct bpf_stack_build_id {
444 unsigned char build_id[BPF_BUILD_ID_SIZE];
451 #define BPF_OBJ_NAME_LEN 16U
454 struct { /* anonymous struct used by BPF_MAP_CREATE command */
455 __u32 map_type; /* one of enum bpf_map_type */
456 __u32 key_size; /* size of key in bytes */
457 __u32 value_size; /* size of value in bytes */
458 __u32 max_entries; /* max number of entries in a map */
459 __u32 map_flags; /* BPF_MAP_CREATE related
460 * flags defined above.
462 __u32 inner_map_fd; /* fd pointing to the inner map */
463 __u32 numa_node; /* numa node (effective only if
464 * BPF_F_NUMA_NODE is set).
466 char map_name[BPF_OBJ_NAME_LEN];
467 __u32 map_ifindex; /* ifindex of netdev to create on */
468 __u32 btf_fd; /* fd pointing to a BTF type data */
469 __u32 btf_key_type_id; /* BTF type_id of the key */
470 __u32 btf_value_type_id; /* BTF type_id of the value */
471 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
472 * struct stored as the
477 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
482 __aligned_u64 next_key;
487 struct { /* struct used by BPF_MAP_*_BATCH commands */
488 __aligned_u64 in_batch; /* start batch,
489 * NULL to start from beginning
491 __aligned_u64 out_batch; /* output: next start batch */
493 __aligned_u64 values;
494 __u32 count; /* input/output:
495 * input: # of key/value
497 * output: # of filled elements
504 struct { /* anonymous struct used by BPF_PROG_LOAD command */
505 __u32 prog_type; /* one of enum bpf_prog_type */
508 __aligned_u64 license;
509 __u32 log_level; /* verbosity level of verifier */
510 __u32 log_size; /* size of user buffer */
511 __aligned_u64 log_buf; /* user supplied buffer */
512 __u32 kern_version; /* not used */
514 char prog_name[BPF_OBJ_NAME_LEN];
515 __u32 prog_ifindex; /* ifindex of netdev to prep for */
516 /* For some prog types expected attach type must be known at
517 * load time to verify attach type specific parts of prog
518 * (context accesses, allowed helpers, etc).
520 __u32 expected_attach_type;
521 __u32 prog_btf_fd; /* fd pointing to BTF type data */
522 __u32 func_info_rec_size; /* userspace bpf_func_info size */
523 __aligned_u64 func_info; /* func info */
524 __u32 func_info_cnt; /* number of bpf_func_info records */
525 __u32 line_info_rec_size; /* userspace bpf_line_info size */
526 __aligned_u64 line_info; /* line info */
527 __u32 line_info_cnt; /* number of bpf_line_info records */
528 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
529 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
532 struct { /* anonymous struct used by BPF_OBJ_* commands */
533 __aligned_u64 pathname;
538 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
539 __u32 target_fd; /* container object to attach to */
540 __u32 attach_bpf_fd; /* eBPF program to attach */
543 __u32 replace_bpf_fd; /* previously attached eBPF
544 * program to replace if
545 * BPF_F_REPLACE is used
549 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
552 __u32 data_size_in; /* input: len of data_in */
553 __u32 data_size_out; /* input/output: len of data_out
554 * returns ENOSPC if data_out
557 __aligned_u64 data_in;
558 __aligned_u64 data_out;
561 __u32 ctx_size_in; /* input: len of ctx_in */
562 __u32 ctx_size_out; /* input/output: len of ctx_out
563 * returns ENOSPC if ctx_out
566 __aligned_u64 ctx_in;
567 __aligned_u64 ctx_out;
570 struct { /* anonymous struct used by BPF_*_GET_*_ID */
582 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
588 struct { /* anonymous struct used by BPF_PROG_QUERY command */
589 __u32 target_fd; /* container object to query */
593 __aligned_u64 prog_ids;
597 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
602 struct { /* anonymous struct for BPF_BTF_LOAD */
604 __aligned_u64 btf_log_buf;
611 __u32 pid; /* input: pid */
612 __u32 fd; /* input: fd */
613 __u32 flags; /* input: flags */
614 __u32 buf_len; /* input/output: buf len */
615 __aligned_u64 buf; /* input/output:
616 * tp_name for tracepoint
618 * filename for uprobe
620 __u32 prog_id; /* output: prod_id */
621 __u32 fd_type; /* output: BPF_FD_TYPE_* */
622 __u64 probe_offset; /* output: probe_offset */
623 __u64 probe_addr; /* output: probe_addr */
626 struct { /* struct used by BPF_LINK_CREATE command */
627 __u32 prog_fd; /* eBPF program to attach */
629 __u32 target_fd; /* object to attach to */
630 __u32 target_ifindex; /* target ifindex */
632 __u32 attach_type; /* attach type */
633 __u32 flags; /* extra flags */
634 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
635 __u32 iter_info_len; /* iter_info length */
638 struct { /* struct used by BPF_LINK_UPDATE command */
639 __u32 link_fd; /* link fd */
640 /* new program fd to update link with */
642 __u32 flags; /* extra flags */
643 /* expected link's program fd; is specified only if
644 * BPF_F_REPLACE flag is set in flags */
652 struct { /* struct used by BPF_ENABLE_STATS command */
656 struct { /* struct used by BPF_ITER_CREATE command */
661 } __attribute__((aligned(8)));
663 /* The description below is an attempt at providing documentation to eBPF
664 * developers about the multiple available eBPF helper functions. It can be
665 * parsed and used to produce a manual page. The workflow is the following,
666 * and requires the rst2man utility:
668 * $ ./scripts/bpf_helpers_doc.py \
669 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
670 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
671 * $ man /tmp/bpf-helpers.7
673 * Note that in order to produce this external documentation, some RST
674 * formatting is used in the descriptions to get "bold" and "italics" in
675 * manual pages. Also note that the few trailing white spaces are
676 * intentional, removing them would break paragraphs for rst2man.
678 * Start of BPF helper function descriptions:
680 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
682 * Perform a lookup in *map* for an entry associated to *key*.
684 * Map value associated to *key*, or **NULL** if no entry was
687 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
689 * Add or update the value of the entry associated to *key* in
690 * *map* with *value*. *flags* is one of:
693 * The entry for *key* must not exist in the map.
695 * The entry for *key* must already exist in the map.
697 * No condition on the existence of the entry for *key*.
699 * Flag value **BPF_NOEXIST** cannot be used for maps of types
700 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
701 * elements always exist), the helper would return an error.
703 * 0 on success, or a negative error in case of failure.
705 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
707 * Delete entry with *key* from *map*.
709 * 0 on success, or a negative error in case of failure.
711 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
713 * For tracing programs, safely attempt to read *size* bytes from
714 * kernel space address *unsafe_ptr* and store the data in *dst*.
716 * Generally, use **bpf_probe_read_user**\ () or
717 * **bpf_probe_read_kernel**\ () instead.
719 * 0 on success, or a negative error in case of failure.
721 * u64 bpf_ktime_get_ns(void)
723 * Return the time elapsed since system boot, in nanoseconds.
724 * Does not include time the system was suspended.
725 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
729 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
731 * This helper is a "printk()-like" facility for debugging. It
732 * prints a message defined by format *fmt* (of size *fmt_size*)
733 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
734 * available. It can take up to three additional **u64**
735 * arguments (as an eBPF helpers, the total number of arguments is
738 * Each time the helper is called, it appends a line to the trace.
739 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
740 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
741 * The format of the trace is customizable, and the exact output
742 * one will get depends on the options set in
743 * *\/sys/kernel/debug/tracing/trace_options* (see also the
744 * *README* file under the same directory). However, it usually
745 * defaults to something like:
749 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
753 * * ``telnet`` is the name of the current task.
754 * * ``470`` is the PID of the current task.
755 * * ``001`` is the CPU number on which the task is
757 * * In ``.N..``, each character refers to a set of
758 * options (whether irqs are enabled, scheduling
759 * options, whether hard/softirqs are running, level of
760 * preempt_disabled respectively). **N** means that
761 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
763 * * ``419421.045894`` is a timestamp.
764 * * ``0x00000001`` is a fake value used by BPF for the
765 * instruction pointer register.
766 * * ``<formatted msg>`` is the message formatted with
769 * The conversion specifiers supported by *fmt* are similar, but
770 * more limited than for printk(). They are **%d**, **%i**,
771 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
772 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
773 * of field, padding with zeroes, etc.) is available, and the
774 * helper will return **-EINVAL** (but print nothing) if it
775 * encounters an unknown specifier.
777 * Also, note that **bpf_trace_printk**\ () is slow, and should
778 * only be used for debugging purposes. For this reason, a notice
779 * block (spanning several lines) is printed to kernel logs and
780 * states that the helper should not be used "for production use"
781 * the first time this helper is used (or more precisely, when
782 * **trace_printk**\ () buffers are allocated). For passing values
783 * to user space, perf events should be preferred.
785 * The number of bytes written to the buffer, or a negative error
786 * in case of failure.
788 * u32 bpf_get_prandom_u32(void)
790 * Get a pseudo-random number.
792 * From a security point of view, this helper uses its own
793 * pseudo-random internal state, and cannot be used to infer the
794 * seed of other random functions in the kernel. However, it is
795 * essential to note that the generator used by the helper is not
796 * cryptographically secure.
798 * A random 32-bit unsigned value.
800 * u32 bpf_get_smp_processor_id(void)
802 * Get the SMP (symmetric multiprocessing) processor id. Note that
803 * all programs run with preemption disabled, which means that the
804 * SMP processor id is stable during all the execution of the
807 * The SMP id of the processor running the program.
809 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
811 * Store *len* bytes from address *from* into the packet
812 * associated to *skb*, at *offset*. *flags* are a combination of
813 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
814 * checksum for the packet after storing the bytes) and
815 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
816 * **->swhash** and *skb*\ **->l4hash** to 0).
818 * A call to this helper is susceptible to change the underlying
819 * packet buffer. Therefore, at load time, all checks on pointers
820 * previously done by the verifier are invalidated and must be
821 * performed again, if the helper is used in combination with
822 * direct packet access.
824 * 0 on success, or a negative error in case of failure.
826 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
828 * Recompute the layer 3 (e.g. IP) checksum for the packet
829 * associated to *skb*. Computation is incremental, so the helper
830 * must know the former value of the header field that was
831 * modified (*from*), the new value of this field (*to*), and the
832 * number of bytes (2 or 4) for this field, stored in *size*.
833 * Alternatively, it is possible to store the difference between
834 * the previous and the new values of the header field in *to*, by
835 * setting *from* and *size* to 0. For both methods, *offset*
836 * indicates the location of the IP checksum within the packet.
838 * This helper works in combination with **bpf_csum_diff**\ (),
839 * which does not update the checksum in-place, but offers more
840 * flexibility and can handle sizes larger than 2 or 4 for the
841 * checksum to update.
843 * A call to this helper is susceptible to change the underlying
844 * packet buffer. Therefore, at load time, all checks on pointers
845 * previously done by the verifier are invalidated and must be
846 * performed again, if the helper is used in combination with
847 * direct packet access.
849 * 0 on success, or a negative error in case of failure.
851 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
853 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
854 * packet associated to *skb*. Computation is incremental, so the
855 * helper must know the former value of the header field that was
856 * modified (*from*), the new value of this field (*to*), and the
857 * number of bytes (2 or 4) for this field, stored on the lowest
858 * four bits of *flags*. Alternatively, it is possible to store
859 * the difference between the previous and the new values of the
860 * header field in *to*, by setting *from* and the four lowest
861 * bits of *flags* to 0. For both methods, *offset* indicates the
862 * location of the IP checksum within the packet. In addition to
863 * the size of the field, *flags* can be added (bitwise OR) actual
864 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
865 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
866 * for updates resulting in a null checksum the value is set to
867 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
868 * the checksum is to be computed against a pseudo-header.
870 * This helper works in combination with **bpf_csum_diff**\ (),
871 * which does not update the checksum in-place, but offers more
872 * flexibility and can handle sizes larger than 2 or 4 for the
873 * checksum to update.
875 * A call to this helper is susceptible to change the underlying
876 * packet buffer. Therefore, at load time, all checks on pointers
877 * previously done by the verifier are invalidated and must be
878 * performed again, if the helper is used in combination with
879 * direct packet access.
881 * 0 on success, or a negative error in case of failure.
883 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
885 * This special helper is used to trigger a "tail call", or in
886 * other words, to jump into another eBPF program. The same stack
887 * frame is used (but values on stack and in registers for the
888 * caller are not accessible to the callee). This mechanism allows
889 * for program chaining, either for raising the maximum number of
890 * available eBPF instructions, or to execute given programs in
891 * conditional blocks. For security reasons, there is an upper
892 * limit to the number of successive tail calls that can be
895 * Upon call of this helper, the program attempts to jump into a
896 * program referenced at index *index* in *prog_array_map*, a
897 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
898 * *ctx*, a pointer to the context.
900 * If the call succeeds, the kernel immediately runs the first
901 * instruction of the new program. This is not a function call,
902 * and it never returns to the previous program. If the call
903 * fails, then the helper has no effect, and the caller continues
904 * to run its subsequent instructions. A call can fail if the
905 * destination program for the jump does not exist (i.e. *index*
906 * is superior to the number of entries in *prog_array_map*), or
907 * if the maximum number of tail calls has been reached for this
908 * chain of programs. This limit is defined in the kernel by the
909 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
910 * which is currently set to 32.
912 * 0 on success, or a negative error in case of failure.
914 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
916 * Clone and redirect the packet associated to *skb* to another
917 * net device of index *ifindex*. Both ingress and egress
918 * interfaces can be used for redirection. The **BPF_F_INGRESS**
919 * value in *flags* is used to make the distinction (ingress path
920 * is selected if the flag is present, egress path otherwise).
921 * This is the only flag supported for now.
923 * In comparison with **bpf_redirect**\ () helper,
924 * **bpf_clone_redirect**\ () has the associated cost of
925 * duplicating the packet buffer, but this can be executed out of
926 * the eBPF program. Conversely, **bpf_redirect**\ () is more
927 * efficient, but it is handled through an action code where the
928 * redirection happens only after the eBPF program has returned.
930 * A call to this helper is susceptible to change the underlying
931 * packet buffer. Therefore, at load time, all checks on pointers
932 * previously done by the verifier are invalidated and must be
933 * performed again, if the helper is used in combination with
934 * direct packet access.
936 * 0 on success, or a negative error in case of failure.
938 * u64 bpf_get_current_pid_tgid(void)
940 * A 64-bit integer containing the current tgid and pid, and
942 * *current_task*\ **->tgid << 32 \|**
943 * *current_task*\ **->pid**.
945 * u64 bpf_get_current_uid_gid(void)
947 * A 64-bit integer containing the current GID and UID, and
948 * created as such: *current_gid* **<< 32 \|** *current_uid*.
950 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
952 * Copy the **comm** attribute of the current task into *buf* of
953 * *size_of_buf*. The **comm** attribute contains the name of
954 * the executable (excluding the path) for the current task. The
955 * *size_of_buf* must be strictly positive. On success, the
956 * helper makes sure that the *buf* is NUL-terminated. On failure,
957 * it is filled with zeroes.
959 * 0 on success, or a negative error in case of failure.
961 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
963 * Retrieve the classid for the current task, i.e. for the net_cls
964 * cgroup to which *skb* belongs.
966 * This helper can be used on TC egress path, but not on ingress.
968 * The net_cls cgroup provides an interface to tag network packets
969 * based on a user-provided identifier for all traffic coming from
970 * the tasks belonging to the related cgroup. See also the related
971 * kernel documentation, available from the Linux sources in file
972 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
974 * The Linux kernel has two versions for cgroups: there are
975 * cgroups v1 and cgroups v2. Both are available to users, who can
976 * use a mixture of them, but note that the net_cls cgroup is for
977 * cgroup v1 only. This makes it incompatible with BPF programs
978 * run on cgroups, which is a cgroup-v2-only feature (a socket can
979 * only hold data for one version of cgroups at a time).
981 * This helper is only available is the kernel was compiled with
982 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
983 * "**y**" or to "**m**".
985 * The classid, or 0 for the default unconfigured classid.
987 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
989 * Push a *vlan_tci* (VLAN tag control information) of protocol
990 * *vlan_proto* to the packet associated to *skb*, then update
991 * the checksum. Note that if *vlan_proto* is different from
992 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
993 * be **ETH_P_8021Q**.
995 * A call to this helper is susceptible to change the underlying
996 * packet buffer. Therefore, at load time, all checks on pointers
997 * previously done by the verifier are invalidated and must be
998 * performed again, if the helper is used in combination with
999 * direct packet access.
1001 * 0 on success, or a negative error in case of failure.
1003 * long bpf_skb_vlan_pop(struct sk_buff *skb)
1005 * Pop a VLAN header from the packet associated to *skb*.
1007 * A call to this helper is susceptible to change the underlying
1008 * packet buffer. Therefore, at load time, all checks on pointers
1009 * previously done by the verifier are invalidated and must be
1010 * performed again, if the helper is used in combination with
1011 * direct packet access.
1013 * 0 on success, or a negative error in case of failure.
1015 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1017 * Get tunnel metadata. This helper takes a pointer *key* to an
1018 * empty **struct bpf_tunnel_key** of **size**, that will be
1019 * filled with tunnel metadata for the packet associated to *skb*.
1020 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1021 * indicates that the tunnel is based on IPv6 protocol instead of
1024 * The **struct bpf_tunnel_key** is an object that generalizes the
1025 * principal parameters used by various tunneling protocols into a
1026 * single struct. This way, it can be used to easily make a
1027 * decision based on the contents of the encapsulation header,
1028 * "summarized" in this struct. In particular, it holds the IP
1029 * address of the remote end (IPv4 or IPv6, depending on the case)
1030 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1031 * this struct exposes the *key*\ **->tunnel_id**, which is
1032 * generally mapped to a VNI (Virtual Network Identifier), making
1033 * it programmable together with the **bpf_skb_set_tunnel_key**\
1036 * Let's imagine that the following code is part of a program
1037 * attached to the TC ingress interface, on one end of a GRE
1038 * tunnel, and is supposed to filter out all messages coming from
1039 * remote ends with IPv4 address other than 10.0.0.1:
1044 * struct bpf_tunnel_key key = {};
1046 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1048 * return TC_ACT_SHOT; // drop packet
1050 * if (key.remote_ipv4 != 0x0a000001)
1051 * return TC_ACT_SHOT; // drop packet
1053 * return TC_ACT_OK; // accept packet
1055 * This interface can also be used with all encapsulation devices
1056 * that can operate in "collect metadata" mode: instead of having
1057 * one network device per specific configuration, the "collect
1058 * metadata" mode only requires a single device where the
1059 * configuration can be extracted from this helper.
1061 * This can be used together with various tunnels such as VXLan,
1062 * Geneve, GRE or IP in IP (IPIP).
1064 * 0 on success, or a negative error in case of failure.
1066 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1068 * Populate tunnel metadata for packet associated to *skb.* The
1069 * tunnel metadata is set to the contents of *key*, of *size*. The
1070 * *flags* can be set to a combination of the following values:
1072 * **BPF_F_TUNINFO_IPV6**
1073 * Indicate that the tunnel is based on IPv6 protocol
1075 * **BPF_F_ZERO_CSUM_TX**
1076 * For IPv4 packets, add a flag to tunnel metadata
1077 * indicating that checksum computation should be skipped
1078 * and checksum set to zeroes.
1079 * **BPF_F_DONT_FRAGMENT**
1080 * Add a flag to tunnel metadata indicating that the
1081 * packet should not be fragmented.
1082 * **BPF_F_SEQ_NUMBER**
1083 * Add a flag to tunnel metadata indicating that a
1084 * sequence number should be added to tunnel header before
1085 * sending the packet. This flag was added for GRE
1086 * encapsulation, but might be used with other protocols
1087 * as well in the future.
1089 * Here is a typical usage on the transmit path:
1093 * struct bpf_tunnel_key key;
1095 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1096 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1098 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1099 * helper for additional information.
1101 * 0 on success, or a negative error in case of failure.
1103 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1105 * Read the value of a perf event counter. This helper relies on a
1106 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1107 * the perf event counter is selected when *map* is updated with
1108 * perf event file descriptors. The *map* is an array whose size
1109 * is the number of available CPUs, and each cell contains a value
1110 * relative to one CPU. The value to retrieve is indicated by
1111 * *flags*, that contains the index of the CPU to look up, masked
1112 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1113 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1114 * current CPU should be retrieved.
1116 * Note that before Linux 4.13, only hardware perf event can be
1119 * Also, be aware that the newer helper
1120 * **bpf_perf_event_read_value**\ () is recommended over
1121 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1122 * quirks where error and counter value are used as a return code
1123 * (which is wrong to do since ranges may overlap). This issue is
1124 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1125 * time provides more features over the **bpf_perf_event_read**\
1126 * () interface. Please refer to the description of
1127 * **bpf_perf_event_read_value**\ () for details.
1129 * The value of the perf event counter read from the map, or a
1130 * negative error code in case of failure.
1132 * long bpf_redirect(u32 ifindex, u64 flags)
1134 * Redirect the packet to another net device of index *ifindex*.
1135 * This helper is somewhat similar to **bpf_clone_redirect**\
1136 * (), except that the packet is not cloned, which provides
1137 * increased performance.
1139 * Except for XDP, both ingress and egress interfaces can be used
1140 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1141 * to make the distinction (ingress path is selected if the flag
1142 * is present, egress path otherwise). Currently, XDP only
1143 * supports redirection to the egress interface, and accepts no
1146 * The same effect can also be attained with the more generic
1147 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1148 * redirect target instead of providing it directly to the helper.
1150 * For XDP, the helper returns **XDP_REDIRECT** on success or
1151 * **XDP_ABORTED** on error. For other program types, the values
1152 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1155 * u32 bpf_get_route_realm(struct sk_buff *skb)
1157 * Retrieve the realm or the route, that is to say the
1158 * **tclassid** field of the destination for the *skb*. The
1159 * identifier retrieved is a user-provided tag, similar to the
1160 * one used with the net_cls cgroup (see description for
1161 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1162 * held by a route (a destination entry), not by a task.
1164 * Retrieving this identifier works with the clsact TC egress hook
1165 * (see also **tc-bpf(8)**), or alternatively on conventional
1166 * classful egress qdiscs, but not on TC ingress path. In case of
1167 * clsact TC egress hook, this has the advantage that, internally,
1168 * the destination entry has not been dropped yet in the transmit
1169 * path. Therefore, the destination entry does not need to be
1170 * artificially held via **netif_keep_dst**\ () for a classful
1171 * qdisc until the *skb* is freed.
1173 * This helper is available only if the kernel was compiled with
1174 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1176 * The realm of the route for the packet associated to *skb*, or 0
1177 * if none was found.
1179 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1181 * Write raw *data* blob into a special BPF perf event held by
1182 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1183 * event must have the following attributes: **PERF_SAMPLE_RAW**
1184 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1185 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1187 * The *flags* are used to indicate the index in *map* for which
1188 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1189 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1190 * to indicate that the index of the current CPU core should be
1193 * The value to write, of *size*, is passed through eBPF stack and
1194 * pointed by *data*.
1196 * The context of the program *ctx* needs also be passed to the
1199 * On user space, a program willing to read the values needs to
1200 * call **perf_event_open**\ () on the perf event (either for
1201 * one or for all CPUs) and to store the file descriptor into the
1202 * *map*. This must be done before the eBPF program can send data
1203 * into it. An example is available in file
1204 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1205 * tree (the eBPF program counterpart is in
1206 * *samples/bpf/trace_output_kern.c*).
1208 * **bpf_perf_event_output**\ () achieves better performance
1209 * than **bpf_trace_printk**\ () for sharing data with user
1210 * space, and is much better suitable for streaming data from eBPF
1213 * Note that this helper is not restricted to tracing use cases
1214 * and can be used with programs attached to TC or XDP as well,
1215 * where it allows for passing data to user space listeners. Data
1218 * * Only custom structs,
1219 * * Only the packet payload, or
1220 * * A combination of both.
1222 * 0 on success, or a negative error in case of failure.
1224 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1226 * This helper was provided as an easy way to load data from a
1227 * packet. It can be used to load *len* bytes from *offset* from
1228 * the packet associated to *skb*, into the buffer pointed by
1231 * Since Linux 4.7, usage of this helper has mostly been replaced
1232 * by "direct packet access", enabling packet data to be
1233 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1234 * pointing respectively to the first byte of packet data and to
1235 * the byte after the last byte of packet data. However, it
1236 * remains useful if one wishes to read large quantities of data
1237 * at once from a packet into the eBPF stack.
1239 * 0 on success, or a negative error in case of failure.
1241 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1243 * Walk a user or a kernel stack and return its id. To achieve
1244 * this, the helper needs *ctx*, which is a pointer to the context
1245 * on which the tracing program is executed, and a pointer to a
1246 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1248 * The last argument, *flags*, holds the number of stack frames to
1249 * skip (from 0 to 255), masked with
1250 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1251 * a combination of the following flags:
1253 * **BPF_F_USER_STACK**
1254 * Collect a user space stack instead of a kernel stack.
1255 * **BPF_F_FAST_STACK_CMP**
1256 * Compare stacks by hash only.
1257 * **BPF_F_REUSE_STACKID**
1258 * If two different stacks hash into the same *stackid*,
1259 * discard the old one.
1261 * The stack id retrieved is a 32 bit long integer handle which
1262 * can be further combined with other data (including other stack
1263 * ids) and used as a key into maps. This can be useful for
1264 * generating a variety of graphs (such as flame graphs or off-cpu
1267 * For walking a stack, this helper is an improvement over
1268 * **bpf_probe_read**\ (), which can be used with unrolled loops
1269 * but is not efficient and consumes a lot of eBPF instructions.
1270 * Instead, **bpf_get_stackid**\ () can collect up to
1271 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1272 * this limit can be controlled with the **sysctl** program, and
1273 * that it should be manually increased in order to profile long
1274 * user stacks (such as stacks for Java programs). To do so, use:
1278 * # sysctl kernel.perf_event_max_stack=<new value>
1280 * The positive or null stack id on success, or a negative error
1281 * in case of failure.
1283 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1285 * Compute a checksum difference, from the raw buffer pointed by
1286 * *from*, of length *from_size* (that must be a multiple of 4),
1287 * towards the raw buffer pointed by *to*, of size *to_size*
1288 * (same remark). An optional *seed* can be added to the value
1289 * (this can be cascaded, the seed may come from a previous call
1292 * This is flexible enough to be used in several ways:
1294 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1295 * checksum, it can be used when pushing new data.
1296 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1297 * checksum, it can be used when removing data from a packet.
1298 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1299 * can be used to compute a diff. Note that *from_size* and
1300 * *to_size* do not need to be equal.
1302 * This helper can be used in combination with
1303 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1304 * which one can feed in the difference computed with
1305 * **bpf_csum_diff**\ ().
1307 * The checksum result, or a negative error code in case of
1310 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1312 * Retrieve tunnel options metadata for the packet associated to
1313 * *skb*, and store the raw tunnel option data to the buffer *opt*
1316 * This helper can be used with encapsulation devices that can
1317 * operate in "collect metadata" mode (please refer to the related
1318 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1319 * more details). A particular example where this can be used is
1320 * in combination with the Geneve encapsulation protocol, where it
1321 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1322 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1323 * the eBPF program. This allows for full customization of these
1326 * The size of the option data retrieved.
1328 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1330 * Set tunnel options metadata for the packet associated to *skb*
1331 * to the option data contained in the raw buffer *opt* of *size*.
1333 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1334 * helper for additional information.
1336 * 0 on success, or a negative error in case of failure.
1338 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1340 * Change the protocol of the *skb* to *proto*. Currently
1341 * supported are transition from IPv4 to IPv6, and from IPv6 to
1342 * IPv4. The helper takes care of the groundwork for the
1343 * transition, including resizing the socket buffer. The eBPF
1344 * program is expected to fill the new headers, if any, via
1345 * **skb_store_bytes**\ () and to recompute the checksums with
1346 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1347 * (). The main case for this helper is to perform NAT64
1348 * operations out of an eBPF program.
1350 * Internally, the GSO type is marked as dodgy so that headers are
1351 * checked and segments are recalculated by the GSO/GRO engine.
1352 * The size for GSO target is adapted as well.
1354 * All values for *flags* are reserved for future usage, and must
1357 * A call to this helper is susceptible to change the underlying
1358 * packet buffer. Therefore, at load time, all checks on pointers
1359 * previously done by the verifier are invalidated and must be
1360 * performed again, if the helper is used in combination with
1361 * direct packet access.
1363 * 0 on success, or a negative error in case of failure.
1365 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
1367 * Change the packet type for the packet associated to *skb*. This
1368 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1369 * the eBPF program does not have a write access to *skb*\
1370 * **->pkt_type** beside this helper. Using a helper here allows
1371 * for graceful handling of errors.
1373 * The major use case is to change incoming *skb*s to
1374 * **PACKET_HOST** in a programmatic way instead of having to
1375 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1378 * Note that *type* only allows certain values. At this time, they
1383 * **PACKET_BROADCAST**
1384 * Send packet to all.
1385 * **PACKET_MULTICAST**
1386 * Send packet to group.
1387 * **PACKET_OTHERHOST**
1388 * Send packet to someone else.
1390 * 0 on success, or a negative error in case of failure.
1392 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1394 * Check whether *skb* is a descendant of the cgroup2 held by
1395 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1397 * The return value depends on the result of the test, and can be:
1399 * * 0, if the *skb* failed the cgroup2 descendant test.
1400 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1401 * * A negative error code, if an error occurred.
1403 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1405 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1406 * not set, in particular if the hash was cleared due to mangling,
1407 * recompute this hash. Later accesses to the hash can be done
1408 * directly with *skb*\ **->hash**.
1410 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1411 * prototype with **bpf_skb_change_proto**\ (), or calling
1412 * **bpf_skb_store_bytes**\ () with the
1413 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1414 * the hash and to trigger a new computation for the next call to
1415 * **bpf_get_hash_recalc**\ ().
1419 * u64 bpf_get_current_task(void)
1421 * A pointer to the current task struct.
1423 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
1425 * Attempt in a safe way to write *len* bytes from the buffer
1426 * *src* to *dst* in memory. It only works for threads that are in
1427 * user context, and *dst* must be a valid user space address.
1429 * This helper should not be used to implement any kind of
1430 * security mechanism because of TOC-TOU attacks, but rather to
1431 * debug, divert, and manipulate execution of semi-cooperative
1434 * Keep in mind that this feature is meant for experiments, and it
1435 * has a risk of crashing the system and running programs.
1436 * Therefore, when an eBPF program using this helper is attached,
1437 * a warning including PID and process name is printed to kernel
1440 * 0 on success, or a negative error in case of failure.
1442 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1444 * Check whether the probe is being run is the context of a given
1445 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1446 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1448 * The return value depends on the result of the test, and can be:
1450 * * 0, if the *skb* task belongs to the cgroup2.
1451 * * 1, if the *skb* task does not belong to the cgroup2.
1452 * * A negative error code, if an error occurred.
1454 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1456 * Resize (trim or grow) the packet associated to *skb* to the
1457 * new *len*. The *flags* are reserved for future usage, and must
1460 * The basic idea is that the helper performs the needed work to
1461 * change the size of the packet, then the eBPF program rewrites
1462 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1463 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1464 * and others. This helper is a slow path utility intended for
1465 * replies with control messages. And because it is targeted for
1466 * slow path, the helper itself can afford to be slow: it
1467 * implicitly linearizes, unclones and drops offloads from the
1470 * A call to this helper is susceptible to change the underlying
1471 * packet buffer. Therefore, at load time, all checks on pointers
1472 * previously done by the verifier are invalidated and must be
1473 * performed again, if the helper is used in combination with
1474 * direct packet access.
1476 * 0 on success, or a negative error in case of failure.
1478 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1480 * Pull in non-linear data in case the *skb* is non-linear and not
1481 * all of *len* are part of the linear section. Make *len* bytes
1482 * from *skb* readable and writable. If a zero value is passed for
1483 * *len*, then the whole length of the *skb* is pulled.
1485 * This helper is only needed for reading and writing with direct
1488 * For direct packet access, testing that offsets to access
1489 * are within packet boundaries (test on *skb*\ **->data_end**) is
1490 * susceptible to fail if offsets are invalid, or if the requested
1491 * data is in non-linear parts of the *skb*. On failure the
1492 * program can just bail out, or in the case of a non-linear
1493 * buffer, use a helper to make the data available. The
1494 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1495 * the data. Another one consists in using **bpf_skb_pull_data**
1496 * to pull in once the non-linear parts, then retesting and
1497 * eventually access the data.
1499 * At the same time, this also makes sure the *skb* is uncloned,
1500 * which is a necessary condition for direct write. As this needs
1501 * to be an invariant for the write part only, the verifier
1502 * detects writes and adds a prologue that is calling
1503 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1504 * the very beginning in case it is indeed cloned.
1506 * A call to this helper is susceptible to change the underlying
1507 * packet buffer. Therefore, at load time, all checks on pointers
1508 * previously done by the verifier are invalidated and must be
1509 * performed again, if the helper is used in combination with
1510 * direct packet access.
1512 * 0 on success, or a negative error in case of failure.
1514 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1516 * Add the checksum *csum* into *skb*\ **->csum** in case the
1517 * driver has supplied a checksum for the entire packet into that
1518 * field. Return an error otherwise. This helper is intended to be
1519 * used in combination with **bpf_csum_diff**\ (), in particular
1520 * when the checksum needs to be updated after data has been
1521 * written into the packet through direct packet access.
1523 * The checksum on success, or a negative error code in case of
1526 * void bpf_set_hash_invalid(struct sk_buff *skb)
1528 * Invalidate the current *skb*\ **->hash**. It can be used after
1529 * mangling on headers through direct packet access, in order to
1530 * indicate that the hash is outdated and to trigger a
1531 * recalculation the next time the kernel tries to access this
1532 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1534 * long bpf_get_numa_node_id(void)
1536 * Return the id of the current NUMA node. The primary use case
1537 * for this helper is the selection of sockets for the local NUMA
1538 * node, when the program is attached to sockets using the
1539 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1540 * but the helper is also available to other eBPF program types,
1541 * similarly to **bpf_get_smp_processor_id**\ ().
1543 * The id of current NUMA node.
1545 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1547 * Grows headroom of packet associated to *skb* and adjusts the
1548 * offset of the MAC header accordingly, adding *len* bytes of
1549 * space. It automatically extends and reallocates memory as
1552 * This helper can be used on a layer 3 *skb* to push a MAC header
1553 * for redirection into a layer 2 device.
1555 * All values for *flags* are reserved for future usage, and must
1558 * A call to this helper is susceptible to change the underlying
1559 * packet buffer. Therefore, at load time, all checks on pointers
1560 * previously done by the verifier are invalidated and must be
1561 * performed again, if the helper is used in combination with
1562 * direct packet access.
1564 * 0 on success, or a negative error in case of failure.
1566 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1568 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1569 * it is possible to use a negative value for *delta*. This helper
1570 * can be used to prepare the packet for pushing or popping
1573 * A call to this helper is susceptible to change the underlying
1574 * packet buffer. Therefore, at load time, all checks on pointers
1575 * previously done by the verifier are invalidated and must be
1576 * performed again, if the helper is used in combination with
1577 * direct packet access.
1579 * 0 on success, or a negative error in case of failure.
1581 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1583 * Copy a NUL terminated string from an unsafe kernel address
1584 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
1587 * Generally, use **bpf_probe_read_user_str**\ () or
1588 * **bpf_probe_read_kernel_str**\ () instead.
1590 * On success, the strictly positive length of the string,
1591 * including the trailing NUL character. On error, a negative
1594 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1596 * If the **struct sk_buff** pointed by *skb* has a known socket,
1597 * retrieve the cookie (generated by the kernel) of this socket.
1598 * If no cookie has been set yet, generate a new cookie. Once
1599 * generated, the socket cookie remains stable for the life of the
1600 * socket. This helper can be useful for monitoring per socket
1601 * networking traffic statistics as it provides a global socket
1602 * identifier that can be assumed unique.
1604 * A 8-byte long non-decreasing number on success, or 0 if the
1605 * socket field is missing inside *skb*.
1607 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1609 * Equivalent to bpf_get_socket_cookie() helper that accepts
1610 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1612 * A 8-byte long non-decreasing number.
1614 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1616 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
1617 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1619 * A 8-byte long non-decreasing number.
1621 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1623 * The owner UID of the socket associated to *skb*. If the socket
1624 * is **NULL**, or if it is not a full socket (i.e. if it is a
1625 * time-wait or a request socket instead), **overflowuid** value
1626 * is returned (note that **overflowuid** might also be the actual
1627 * UID value for the socket).
1629 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
1631 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1636 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1638 * Emulate a call to **setsockopt()** on the socket associated to
1639 * *bpf_socket*, which must be a full socket. The *level* at
1640 * which the option resides and the name *optname* of the option
1641 * must be specified, see **setsockopt(2)** for more information.
1642 * The option value of length *optlen* is pointed by *optval*.
1644 * *bpf_socket* should be one of the following:
1646 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1647 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1648 * and **BPF_CGROUP_INET6_CONNECT**.
1650 * This helper actually implements a subset of **setsockopt()**.
1651 * It supports the following *level*\ s:
1653 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1654 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1655 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
1656 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
1657 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1658 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1659 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
1660 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
1661 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**.
1662 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1663 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1665 * 0 on success, or a negative error in case of failure.
1667 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1669 * Grow or shrink the room for data in the packet associated to
1670 * *skb* by *len_diff*, and according to the selected *mode*.
1672 * By default, the helper will reset any offloaded checksum
1673 * indicator of the skb to CHECKSUM_NONE. This can be avoided
1674 * by the following flag:
1676 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
1677 * checksum data of the skb to CHECKSUM_NONE.
1679 * There are two supported modes at this time:
1681 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1682 * (room space is added or removed below the layer 2 header).
1684 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1685 * (room space is added or removed below the layer 3 header).
1687 * The following flags are supported at this time:
1689 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1690 * Adjusting mss in this way is not allowed for datagrams.
1692 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1693 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1694 * Any new space is reserved to hold a tunnel header.
1695 * Configure skb offsets and other fields accordingly.
1697 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1698 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1699 * Use with ENCAP_L3 flags to further specify the tunnel type.
1701 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1702 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1703 * type; *len* is the length of the inner MAC header.
1705 * A call to this helper is susceptible to change the underlying
1706 * packet buffer. Therefore, at load time, all checks on pointers
1707 * previously done by the verifier are invalidated and must be
1708 * performed again, if the helper is used in combination with
1709 * direct packet access.
1711 * 0 on success, or a negative error in case of failure.
1713 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1715 * Redirect the packet to the endpoint referenced by *map* at
1716 * index *key*. Depending on its type, this *map* can contain
1717 * references to net devices (for forwarding packets through other
1718 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1719 * but this is only implemented for native XDP (with driver
1720 * support) as of this writing).
1722 * The lower two bits of *flags* are used as the return code if
1723 * the map lookup fails. This is so that the return value can be
1724 * one of the XDP program return codes up to **XDP_TX**, as chosen
1725 * by the caller. Any higher bits in the *flags* argument must be
1728 * See also **bpf_redirect**\ (), which only supports redirecting
1729 * to an ifindex, but doesn't require a map to do so.
1731 * **XDP_REDIRECT** on success, or the value of the two lower bits
1732 * of the *flags* argument on error.
1734 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1736 * Redirect the packet to the socket referenced by *map* (of type
1737 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1738 * egress interfaces can be used for redirection. The
1739 * **BPF_F_INGRESS** value in *flags* is used to make the
1740 * distinction (ingress path is selected if the flag is present,
1741 * egress path otherwise). This is the only flag supported for now.
1743 * **SK_PASS** on success, or **SK_DROP** on error.
1745 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1747 * Add an entry to, or update a *map* referencing sockets. The
1748 * *skops* is used as a new value for the entry associated to
1749 * *key*. *flags* is one of:
1752 * The entry for *key* must not exist in the map.
1754 * The entry for *key* must already exist in the map.
1756 * No condition on the existence of the entry for *key*.
1758 * If the *map* has eBPF programs (parser and verdict), those will
1759 * be inherited by the socket being added. If the socket is
1760 * already attached to eBPF programs, this results in an error.
1762 * 0 on success, or a negative error in case of failure.
1764 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1766 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1767 * *delta* (which can be positive or negative). Note that this
1768 * operation modifies the address stored in *xdp_md*\ **->data**,
1769 * so the latter must be loaded only after the helper has been
1772 * The use of *xdp_md*\ **->data_meta** is optional and programs
1773 * are not required to use it. The rationale is that when the
1774 * packet is processed with XDP (e.g. as DoS filter), it is
1775 * possible to push further meta data along with it before passing
1776 * to the stack, and to give the guarantee that an ingress eBPF
1777 * program attached as a TC classifier on the same device can pick
1778 * this up for further post-processing. Since TC works with socket
1779 * buffers, it remains possible to set from XDP the **mark** or
1780 * **priority** pointers, or other pointers for the socket buffer.
1781 * Having this scratch space generic and programmable allows for
1782 * more flexibility as the user is free to store whatever meta
1785 * A call to this helper is susceptible to change the underlying
1786 * packet buffer. Therefore, at load time, all checks on pointers
1787 * previously done by the verifier are invalidated and must be
1788 * performed again, if the helper is used in combination with
1789 * direct packet access.
1791 * 0 on success, or a negative error in case of failure.
1793 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1795 * Read the value of a perf event counter, and store it into *buf*
1796 * of size *buf_size*. This helper relies on a *map* of type
1797 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1798 * counter is selected when *map* is updated with perf event file
1799 * descriptors. The *map* is an array whose size is the number of
1800 * available CPUs, and each cell contains a value relative to one
1801 * CPU. The value to retrieve is indicated by *flags*, that
1802 * contains the index of the CPU to look up, masked with
1803 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1804 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1805 * current CPU should be retrieved.
1807 * This helper behaves in a way close to
1808 * **bpf_perf_event_read**\ () helper, save that instead of
1809 * just returning the value observed, it fills the *buf*
1810 * structure. This allows for additional data to be retrieved: in
1811 * particular, the enabled and running times (in *buf*\
1812 * **->enabled** and *buf*\ **->running**, respectively) are
1813 * copied. In general, **bpf_perf_event_read_value**\ () is
1814 * recommended over **bpf_perf_event_read**\ (), which has some
1815 * ABI issues and provides fewer functionalities.
1817 * These values are interesting, because hardware PMU (Performance
1818 * Monitoring Unit) counters are limited resources. When there are
1819 * more PMU based perf events opened than available counters,
1820 * kernel will multiplex these events so each event gets certain
1821 * percentage (but not all) of the PMU time. In case that
1822 * multiplexing happens, the number of samples or counter value
1823 * will not reflect the case compared to when no multiplexing
1824 * occurs. This makes comparison between different runs difficult.
1825 * Typically, the counter value should be normalized before
1826 * comparing to other experiments. The usual normalization is done
1831 * normalized_counter = counter * t_enabled / t_running
1833 * Where t_enabled is the time enabled for event and t_running is
1834 * the time running for event since last normalization. The
1835 * enabled and running times are accumulated since the perf event
1836 * open. To achieve scaling factor between two invocations of an
1837 * eBPF program, users can use CPU id as the key (which is
1838 * typical for perf array usage model) to remember the previous
1839 * value and do the calculation inside the eBPF program.
1841 * 0 on success, or a negative error in case of failure.
1843 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1845 * For en eBPF program attached to a perf event, retrieve the
1846 * value of the event counter associated to *ctx* and store it in
1847 * the structure pointed by *buf* and of size *buf_size*. Enabled
1848 * and running times are also stored in the structure (see
1849 * description of helper **bpf_perf_event_read_value**\ () for
1852 * 0 on success, or a negative error in case of failure.
1854 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1856 * Emulate a call to **getsockopt()** on the socket associated to
1857 * *bpf_socket*, which must be a full socket. The *level* at
1858 * which the option resides and the name *optname* of the option
1859 * must be specified, see **getsockopt(2)** for more information.
1860 * The retrieved value is stored in the structure pointed by
1861 * *opval* and of length *optlen*.
1863 * *bpf_socket* should be one of the following:
1865 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1866 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1867 * and **BPF_CGROUP_INET6_CONNECT**.
1869 * This helper actually implements a subset of **getsockopt()**.
1870 * It supports the following *level*\ s:
1872 * * **IPPROTO_TCP**, which supports *optname*
1873 * **TCP_CONGESTION**.
1874 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1875 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1877 * 0 on success, or a negative error in case of failure.
1879 * long bpf_override_return(struct pt_regs *regs, u64 rc)
1881 * Used for error injection, this helper uses kprobes to override
1882 * the return value of the probed function, and to set it to *rc*.
1883 * The first argument is the context *regs* on which the kprobe
1886 * This helper works by setting the PC (program counter)
1887 * to an override function which is run in place of the original
1888 * probed function. This means the probed function is not run at
1889 * all. The replacement function just returns with the required
1892 * This helper has security implications, and thus is subject to
1893 * restrictions. It is only available if the kernel was compiled
1894 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1895 * option, and in this case it only works on functions tagged with
1896 * **ALLOW_ERROR_INJECTION** in the kernel code.
1898 * Also, the helper is only available for the architectures having
1899 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1900 * x86 architecture is the only one to support this feature.
1904 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1906 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1907 * for the full TCP socket associated to *bpf_sock_ops* to
1910 * The primary use of this field is to determine if there should
1911 * be calls to eBPF programs of type
1912 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1913 * code. A program of the same type can change its value, per
1914 * connection and as necessary, when the connection is
1915 * established. This field is directly accessible for reading, but
1916 * this helper must be used for updates in order to return an
1917 * error if an eBPF program tries to set a callback that is not
1918 * supported in the current kernel.
1920 * *argval* is a flag array which can combine these flags:
1922 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1923 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1924 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1925 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1927 * Therefore, this function can be used to clear a callback flag by
1928 * setting the appropriate bit to zero. e.g. to disable the RTO
1931 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1932 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1934 * Here are some examples of where one could call such eBPF
1938 * * When a packet is retransmitted.
1939 * * When the connection terminates.
1940 * * When a packet is sent.
1941 * * When a packet is received.
1943 * Code **-EINVAL** if the socket is not a full TCP socket;
1944 * otherwise, a positive number containing the bits that could not
1945 * be set is returned (which comes down to 0 if all bits were set
1948 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1950 * This helper is used in programs implementing policies at the
1951 * socket level. If the message *msg* is allowed to pass (i.e. if
1952 * the verdict eBPF program returns **SK_PASS**), redirect it to
1953 * the socket referenced by *map* (of type
1954 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1955 * egress interfaces can be used for redirection. The
1956 * **BPF_F_INGRESS** value in *flags* is used to make the
1957 * distinction (ingress path is selected if the flag is present,
1958 * egress path otherwise). This is the only flag supported for now.
1960 * **SK_PASS** on success, or **SK_DROP** on error.
1962 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1964 * For socket policies, apply the verdict of the eBPF program to
1965 * the next *bytes* (number of bytes) of message *msg*.
1967 * For example, this helper can be used in the following cases:
1969 * * A single **sendmsg**\ () or **sendfile**\ () system call
1970 * contains multiple logical messages that the eBPF program is
1971 * supposed to read and for which it should apply a verdict.
1972 * * An eBPF program only cares to read the first *bytes* of a
1973 * *msg*. If the message has a large payload, then setting up
1974 * and calling the eBPF program repeatedly for all bytes, even
1975 * though the verdict is already known, would create unnecessary
1978 * When called from within an eBPF program, the helper sets a
1979 * counter internal to the BPF infrastructure, that is used to
1980 * apply the last verdict to the next *bytes*. If *bytes* is
1981 * smaller than the current data being processed from a
1982 * **sendmsg**\ () or **sendfile**\ () system call, the first
1983 * *bytes* will be sent and the eBPF program will be re-run with
1984 * the pointer for start of data pointing to byte number *bytes*
1985 * **+ 1**. If *bytes* is larger than the current data being
1986 * processed, then the eBPF verdict will be applied to multiple
1987 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1990 * Note that if a socket closes with the internal counter holding
1991 * a non-zero value, this is not a problem because data is not
1992 * being buffered for *bytes* and is sent as it is received.
1996 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1998 * For socket policies, prevent the execution of the verdict eBPF
1999 * program for message *msg* until *bytes* (byte number) have been
2002 * This can be used when one needs a specific number of bytes
2003 * before a verdict can be assigned, even if the data spans
2004 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2005 * case would be a user calling **sendmsg**\ () repeatedly with
2006 * 1-byte long message segments. Obviously, this is bad for
2007 * performance, but it is still valid. If the eBPF program needs
2008 * *bytes* bytes to validate a header, this helper can be used to
2009 * prevent the eBPF program to be called again until *bytes* have
2014 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2016 * For socket policies, pull in non-linear data from user space
2017 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2018 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2021 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2022 * *msg* it can only parse data that the (**data**, **data_end**)
2023 * pointers have already consumed. For **sendmsg**\ () hooks this
2024 * is likely the first scatterlist element. But for calls relying
2025 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2026 * be the range (**0**, **0**) because the data is shared with
2027 * user space and by default the objective is to avoid allowing
2028 * user space to modify data while (or after) eBPF verdict is
2029 * being decided. This helper can be used to pull in data and to
2030 * set the start and end pointer to given values. Data will be
2031 * copied if necessary (i.e. if data was not linear and if start
2032 * and end pointers do not point to the same chunk).
2034 * A call to this helper is susceptible to change the underlying
2035 * packet buffer. Therefore, at load time, all checks on pointers
2036 * previously done by the verifier are invalidated and must be
2037 * performed again, if the helper is used in combination with
2038 * direct packet access.
2040 * All values for *flags* are reserved for future usage, and must
2043 * 0 on success, or a negative error in case of failure.
2045 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2047 * Bind the socket associated to *ctx* to the address pointed by
2048 * *addr*, of length *addr_len*. This allows for making outgoing
2049 * connection from the desired IP address, which can be useful for
2050 * example when all processes inside a cgroup should use one
2051 * single IP address on a host that has multiple IP configured.
2053 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2054 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2055 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2056 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2057 * behavior and lets the kernel efficiently pick up an unused
2058 * port as long as 4-tuple is unique. Passing non-zero port might
2059 * lead to degraded performance.
2061 * 0 on success, or a negative error in case of failure.
2063 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2065 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2066 * possible to both shrink and grow the packet tail.
2067 * Shrink done via *delta* being a negative integer.
2069 * A call to this helper is susceptible to change the underlying
2070 * packet buffer. Therefore, at load time, all checks on pointers
2071 * previously done by the verifier are invalidated and must be
2072 * performed again, if the helper is used in combination with
2073 * direct packet access.
2075 * 0 on success, or a negative error in case of failure.
2077 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2079 * Retrieve the XFRM state (IP transform framework, see also
2080 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2082 * The retrieved value is stored in the **struct bpf_xfrm_state**
2083 * pointed by *xfrm_state* and of length *size*.
2085 * All values for *flags* are reserved for future usage, and must
2088 * This helper is available only if the kernel was compiled with
2089 * **CONFIG_XFRM** configuration option.
2091 * 0 on success, or a negative error in case of failure.
2093 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2095 * Return a user or a kernel stack in bpf program provided buffer.
2096 * To achieve this, the helper needs *ctx*, which is a pointer
2097 * to the context on which the tracing program is executed.
2098 * To store the stacktrace, the bpf program provides *buf* with
2099 * a nonnegative *size*.
2101 * The last argument, *flags*, holds the number of stack frames to
2102 * skip (from 0 to 255), masked with
2103 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2104 * the following flags:
2106 * **BPF_F_USER_STACK**
2107 * Collect a user space stack instead of a kernel stack.
2108 * **BPF_F_USER_BUILD_ID**
2109 * Collect buildid+offset instead of ips for user stack,
2110 * only valid if **BPF_F_USER_STACK** is also specified.
2112 * **bpf_get_stack**\ () can collect up to
2113 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2114 * to sufficient large buffer size. Note that
2115 * this limit can be controlled with the **sysctl** program, and
2116 * that it should be manually increased in order to profile long
2117 * user stacks (such as stacks for Java programs). To do so, use:
2121 * # sysctl kernel.perf_event_max_stack=<new value>
2123 * A non-negative value equal to or less than *size* on success,
2124 * or a negative error in case of failure.
2126 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2128 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2129 * it provides an easy way to load *len* bytes from *offset*
2130 * from the packet associated to *skb*, into the buffer pointed
2131 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2132 * a fifth argument *start_header* exists in order to select a
2133 * base offset to start from. *start_header* can be one of:
2135 * **BPF_HDR_START_MAC**
2136 * Base offset to load data from is *skb*'s mac header.
2137 * **BPF_HDR_START_NET**
2138 * Base offset to load data from is *skb*'s network header.
2140 * In general, "direct packet access" is the preferred method to
2141 * access packet data, however, this helper is in particular useful
2142 * in socket filters where *skb*\ **->data** does not always point
2143 * to the start of the mac header and where "direct packet access"
2146 * 0 on success, or a negative error in case of failure.
2148 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2150 * Do FIB lookup in kernel tables using parameters in *params*.
2151 * If lookup is successful and result shows packet is to be
2152 * forwarded, the neighbor tables are searched for the nexthop.
2153 * If successful (ie., FIB lookup shows forwarding and nexthop
2154 * is resolved), the nexthop address is returned in ipv4_dst
2155 * or ipv6_dst based on family, smac is set to mac address of
2156 * egress device, dmac is set to nexthop mac address, rt_metric
2157 * is set to metric from route (IPv4/IPv6 only), and ifindex
2158 * is set to the device index of the nexthop from the FIB lookup.
2160 * *plen* argument is the size of the passed in struct.
2161 * *flags* argument can be a combination of one or more of the
2164 * **BPF_FIB_LOOKUP_DIRECT**
2165 * Do a direct table lookup vs full lookup using FIB
2167 * **BPF_FIB_LOOKUP_OUTPUT**
2168 * Perform lookup from an egress perspective (default is
2171 * *ctx* is either **struct xdp_md** for XDP programs or
2172 * **struct sk_buff** tc cls_act programs.
2174 * * < 0 if any input argument is invalid
2175 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2176 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2177 * packet is not forwarded or needs assist from full stack
2179 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2181 * Add an entry to, or update a sockhash *map* referencing sockets.
2182 * The *skops* is used as a new value for the entry associated to
2183 * *key*. *flags* is one of:
2186 * The entry for *key* must not exist in the map.
2188 * The entry for *key* must already exist in the map.
2190 * No condition on the existence of the entry for *key*.
2192 * If the *map* has eBPF programs (parser and verdict), those will
2193 * be inherited by the socket being added. If the socket is
2194 * already attached to eBPF programs, this results in an error.
2196 * 0 on success, or a negative error in case of failure.
2198 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2200 * This helper is used in programs implementing policies at the
2201 * socket level. If the message *msg* is allowed to pass (i.e. if
2202 * the verdict eBPF program returns **SK_PASS**), redirect it to
2203 * the socket referenced by *map* (of type
2204 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2205 * egress interfaces can be used for redirection. The
2206 * **BPF_F_INGRESS** value in *flags* is used to make the
2207 * distinction (ingress path is selected if the flag is present,
2208 * egress path otherwise). This is the only flag supported for now.
2210 * **SK_PASS** on success, or **SK_DROP** on error.
2212 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2214 * This helper is used in programs implementing policies at the
2215 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2216 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2217 * to the socket referenced by *map* (of type
2218 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2219 * egress interfaces can be used for redirection. The
2220 * **BPF_F_INGRESS** value in *flags* is used to make the
2221 * distinction (ingress path is selected if the flag is present,
2222 * egress otherwise). This is the only flag supported for now.
2224 * **SK_PASS** on success, or **SK_DROP** on error.
2226 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2228 * Encapsulate the packet associated to *skb* within a Layer 3
2229 * protocol header. This header is provided in the buffer at
2230 * address *hdr*, with *len* its size in bytes. *type* indicates
2231 * the protocol of the header and can be one of:
2233 * **BPF_LWT_ENCAP_SEG6**
2234 * IPv6 encapsulation with Segment Routing Header
2235 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2236 * the IPv6 header is computed by the kernel.
2237 * **BPF_LWT_ENCAP_SEG6_INLINE**
2238 * Only works if *skb* contains an IPv6 packet. Insert a
2239 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2241 * **BPF_LWT_ENCAP_IP**
2242 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2243 * must be IPv4 or IPv6, followed by zero or more
2244 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2245 * total bytes in all prepended headers. Please note that
2246 * if **skb_is_gso**\ (*skb*) is true, no more than two
2247 * headers can be prepended, and the inner header, if
2248 * present, should be either GRE or UDP/GUE.
2250 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2251 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2252 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2253 * **BPF_PROG_TYPE_LWT_XMIT**.
2255 * A call to this helper is susceptible to change the underlying
2256 * packet buffer. Therefore, at load time, all checks on pointers
2257 * previously done by the verifier are invalidated and must be
2258 * performed again, if the helper is used in combination with
2259 * direct packet access.
2261 * 0 on success, or a negative error in case of failure.
2263 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2265 * Store *len* bytes from address *from* into the packet
2266 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2267 * inside the outermost IPv6 Segment Routing Header can be
2268 * modified through this helper.
2270 * A call to this helper is susceptible to change the underlying
2271 * packet buffer. Therefore, at load time, all checks on pointers
2272 * previously done by the verifier are invalidated and must be
2273 * performed again, if the helper is used in combination with
2274 * direct packet access.
2276 * 0 on success, or a negative error in case of failure.
2278 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2280 * Adjust the size allocated to TLVs in the outermost IPv6
2281 * Segment Routing Header contained in the packet associated to
2282 * *skb*, at position *offset* by *delta* bytes. Only offsets
2283 * after the segments are accepted. *delta* can be as well
2284 * positive (growing) as negative (shrinking).
2286 * A call to this helper is susceptible to change the underlying
2287 * packet buffer. Therefore, at load time, all checks on pointers
2288 * previously done by the verifier are invalidated and must be
2289 * performed again, if the helper is used in combination with
2290 * direct packet access.
2292 * 0 on success, or a negative error in case of failure.
2294 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2296 * Apply an IPv6 Segment Routing action of type *action* to the
2297 * packet associated to *skb*. Each action takes a parameter
2298 * contained at address *param*, and of length *param_len* bytes.
2299 * *action* can be one of:
2301 * **SEG6_LOCAL_ACTION_END_X**
2302 * End.X action: Endpoint with Layer-3 cross-connect.
2303 * Type of *param*: **struct in6_addr**.
2304 * **SEG6_LOCAL_ACTION_END_T**
2305 * End.T action: Endpoint with specific IPv6 table lookup.
2306 * Type of *param*: **int**.
2307 * **SEG6_LOCAL_ACTION_END_B6**
2308 * End.B6 action: Endpoint bound to an SRv6 policy.
2309 * Type of *param*: **struct ipv6_sr_hdr**.
2310 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2311 * End.B6.Encap action: Endpoint bound to an SRv6
2312 * encapsulation policy.
2313 * Type of *param*: **struct ipv6_sr_hdr**.
2315 * A call to this helper is susceptible to change the underlying
2316 * packet buffer. Therefore, at load time, all checks on pointers
2317 * previously done by the verifier are invalidated and must be
2318 * performed again, if the helper is used in combination with
2319 * direct packet access.
2321 * 0 on success, or a negative error in case of failure.
2323 * long bpf_rc_repeat(void *ctx)
2325 * This helper is used in programs implementing IR decoding, to
2326 * report a successfully decoded repeat key message. This delays
2327 * the generation of a key up event for previously generated
2330 * Some IR protocols like NEC have a special IR message for
2331 * repeating last button, for when a button is held down.
2333 * The *ctx* should point to the lirc sample as passed into
2336 * This helper is only available is the kernel was compiled with
2337 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2342 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2344 * This helper is used in programs implementing IR decoding, to
2345 * report a successfully decoded key press with *scancode*,
2346 * *toggle* value in the given *protocol*. The scancode will be
2347 * translated to a keycode using the rc keymap, and reported as
2348 * an input key down event. After a period a key up event is
2349 * generated. This period can be extended by calling either
2350 * **bpf_rc_keydown**\ () again with the same values, or calling
2351 * **bpf_rc_repeat**\ ().
2353 * Some protocols include a toggle bit, in case the button was
2354 * released and pressed again between consecutive scancodes.
2356 * The *ctx* should point to the lirc sample as passed into
2359 * The *protocol* is the decoded protocol number (see
2360 * **enum rc_proto** for some predefined values).
2362 * This helper is only available is the kernel was compiled with
2363 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2368 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2370 * Return the cgroup v2 id of the socket associated with the *skb*.
2371 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2372 * helper for cgroup v1 by providing a tag resp. identifier that
2373 * can be matched on or used for map lookups e.g. to implement
2374 * policy. The cgroup v2 id of a given path in the hierarchy is
2375 * exposed in user space through the f_handle API in order to get
2376 * to the same 64-bit id.
2378 * This helper can be used on TC egress path, but not on ingress,
2379 * and is available only if the kernel was compiled with the
2380 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2382 * The id is returned or 0 in case the id could not be retrieved.
2384 * u64 bpf_get_current_cgroup_id(void)
2386 * A 64-bit integer containing the current cgroup id based
2387 * on the cgroup within which the current task is running.
2389 * void *bpf_get_local_storage(void *map, u64 flags)
2391 * Get the pointer to the local storage area.
2392 * The type and the size of the local storage is defined
2393 * by the *map* argument.
2394 * The *flags* meaning is specific for each map type,
2395 * and has to be 0 for cgroup local storage.
2397 * Depending on the BPF program type, a local storage area
2398 * can be shared between multiple instances of the BPF program,
2399 * running simultaneously.
2401 * A user should care about the synchronization by himself.
2402 * For example, by using the **BPF_STX_XADD** instruction to alter
2405 * A pointer to the local storage area.
2407 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2409 * Select a **SO_REUSEPORT** socket from a
2410 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2411 * It checks the selected socket is matching the incoming
2412 * request in the socket buffer.
2414 * 0 on success, or a negative error in case of failure.
2416 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2418 * Return id of cgroup v2 that is ancestor of cgroup associated
2419 * with the *skb* at the *ancestor_level*. The root cgroup is at
2420 * *ancestor_level* zero and each step down the hierarchy
2421 * increments the level. If *ancestor_level* == level of cgroup
2422 * associated with *skb*, then return value will be same as that
2423 * of **bpf_skb_cgroup_id**\ ().
2425 * The helper is useful to implement policies based on cgroups
2426 * that are upper in hierarchy than immediate cgroup associated
2429 * The format of returned id and helper limitations are same as in
2430 * **bpf_skb_cgroup_id**\ ().
2432 * The id is returned or 0 in case the id could not be retrieved.
2434 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2436 * Look for TCP socket matching *tuple*, optionally in a child
2437 * network namespace *netns*. The return value must be checked,
2438 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2440 * The *ctx* should point to the context of the program, such as
2441 * the skb or socket (depending on the hook in use). This is used
2442 * to determine the base network namespace for the lookup.
2444 * *tuple_size* must be one of:
2446 * **sizeof**\ (*tuple*\ **->ipv4**)
2447 * Look for an IPv4 socket.
2448 * **sizeof**\ (*tuple*\ **->ipv6**)
2449 * Look for an IPv6 socket.
2451 * If the *netns* is a negative signed 32-bit integer, then the
2452 * socket lookup table in the netns associated with the *ctx*
2453 * will be used. For the TC hooks, this is the netns of the device
2454 * in the skb. For socket hooks, this is the netns of the socket.
2455 * If *netns* is any other signed 32-bit value greater than or
2456 * equal to zero then it specifies the ID of the netns relative to
2457 * the netns associated with the *ctx*. *netns* values beyond the
2458 * range of 32-bit integers are reserved for future use.
2460 * All values for *flags* are reserved for future usage, and must
2463 * This helper is available only if the kernel was compiled with
2464 * **CONFIG_NET** configuration option.
2466 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2467 * For sockets with reuseport option, the **struct bpf_sock**
2468 * result is from *reuse*\ **->socks**\ [] using the hash of the
2471 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2473 * Look for UDP socket matching *tuple*, optionally in a child
2474 * network namespace *netns*. The return value must be checked,
2475 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2477 * The *ctx* should point to the context of the program, such as
2478 * the skb or socket (depending on the hook in use). This is used
2479 * to determine the base network namespace for the lookup.
2481 * *tuple_size* must be one of:
2483 * **sizeof**\ (*tuple*\ **->ipv4**)
2484 * Look for an IPv4 socket.
2485 * **sizeof**\ (*tuple*\ **->ipv6**)
2486 * Look for an IPv6 socket.
2488 * If the *netns* is a negative signed 32-bit integer, then the
2489 * socket lookup table in the netns associated with the *ctx*
2490 * will be used. For the TC hooks, this is the netns of the device
2491 * in the skb. For socket hooks, this is the netns of the socket.
2492 * If *netns* is any other signed 32-bit value greater than or
2493 * equal to zero then it specifies the ID of the netns relative to
2494 * the netns associated with the *ctx*. *netns* values beyond the
2495 * range of 32-bit integers are reserved for future use.
2497 * All values for *flags* are reserved for future usage, and must
2500 * This helper is available only if the kernel was compiled with
2501 * **CONFIG_NET** configuration option.
2503 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2504 * For sockets with reuseport option, the **struct bpf_sock**
2505 * result is from *reuse*\ **->socks**\ [] using the hash of the
2508 * long bpf_sk_release(struct bpf_sock *sock)
2510 * Release the reference held by *sock*. *sock* must be a
2511 * non-**NULL** pointer that was returned from
2512 * **bpf_sk_lookup_xxx**\ ().
2514 * 0 on success, or a negative error in case of failure.
2516 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2518 * Push an element *value* in *map*. *flags* is one of:
2521 * If the queue/stack is full, the oldest element is
2522 * removed to make room for this.
2524 * 0 on success, or a negative error in case of failure.
2526 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
2528 * Pop an element from *map*.
2530 * 0 on success, or a negative error in case of failure.
2532 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
2534 * Get an element from *map* without removing it.
2536 * 0 on success, or a negative error in case of failure.
2538 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2540 * For socket policies, insert *len* bytes into *msg* at offset
2543 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2544 * *msg* it may want to insert metadata or options into the *msg*.
2545 * This can later be read and used by any of the lower layer BPF
2548 * This helper may fail if under memory pressure (a malloc
2549 * fails) in these cases BPF programs will get an appropriate
2550 * error and BPF programs will need to handle them.
2552 * 0 on success, or a negative error in case of failure.
2554 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2556 * Will remove *len* bytes from a *msg* starting at byte *start*.
2557 * This may result in **ENOMEM** errors under certain situations if
2558 * an allocation and copy are required due to a full ring buffer.
2559 * However, the helper will try to avoid doing the allocation
2560 * if possible. Other errors can occur if input parameters are
2561 * invalid either due to *start* byte not being valid part of *msg*
2562 * payload and/or *pop* value being to large.
2564 * 0 on success, or a negative error in case of failure.
2566 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2568 * This helper is used in programs implementing IR decoding, to
2569 * report a successfully decoded pointer movement.
2571 * The *ctx* should point to the lirc sample as passed into
2574 * This helper is only available is the kernel was compiled with
2575 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2580 * long bpf_spin_lock(struct bpf_spin_lock *lock)
2582 * Acquire a spinlock represented by the pointer *lock*, which is
2583 * stored as part of a value of a map. Taking the lock allows to
2584 * safely update the rest of the fields in that value. The
2585 * spinlock can (and must) later be released with a call to
2586 * **bpf_spin_unlock**\ (\ *lock*\ ).
2588 * Spinlocks in BPF programs come with a number of restrictions
2591 * * **bpf_spin_lock** objects are only allowed inside maps of
2592 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2593 * list could be extended in the future).
2594 * * BTF description of the map is mandatory.
2595 * * The BPF program can take ONE lock at a time, since taking two
2596 * or more could cause dead locks.
2597 * * Only one **struct bpf_spin_lock** is allowed per map element.
2598 * * When the lock is taken, calls (either BPF to BPF or helpers)
2600 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2601 * allowed inside a spinlock-ed region.
2602 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2603 * the lock, on all execution paths, before it returns.
2604 * * The BPF program can access **struct bpf_spin_lock** only via
2605 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2606 * helpers. Loading or storing data into the **struct
2607 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2608 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2609 * of the map value must be a struct and have **struct
2610 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2611 * Nested lock inside another struct is not allowed.
2612 * * The **struct bpf_spin_lock** *lock* field in a map value must
2613 * be aligned on a multiple of 4 bytes in that value.
2614 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2615 * the **bpf_spin_lock** field to user space.
2616 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2617 * a BPF program, do not update the **bpf_spin_lock** field.
2618 * * **bpf_spin_lock** cannot be on the stack or inside a
2619 * networking packet (it can only be inside of a map values).
2620 * * **bpf_spin_lock** is available to root only.
2621 * * Tracing programs and socket filter programs cannot use
2622 * **bpf_spin_lock**\ () due to insufficient preemption checks
2623 * (but this may change in the future).
2624 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2628 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
2630 * Release the *lock* previously locked by a call to
2631 * **bpf_spin_lock**\ (\ *lock*\ ).
2635 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2637 * This helper gets a **struct bpf_sock** pointer such
2638 * that all the fields in this **bpf_sock** can be accessed.
2640 * A **struct bpf_sock** pointer on success, or **NULL** in
2643 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2645 * This helper gets a **struct bpf_tcp_sock** pointer from a
2646 * **struct bpf_sock** pointer.
2648 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2651 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
2653 * Set ECN (Explicit Congestion Notification) field of IP header
2654 * to **CE** (Congestion Encountered) if current value is **ECT**
2655 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2658 * 1 if the **CE** flag is set (either by the current helper call
2659 * or because it was already present), 0 if it is not set.
2661 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2663 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2664 * **bpf_sk_release**\ () is unnecessary and not allowed.
2666 * A **struct bpf_sock** pointer on success, or **NULL** in
2669 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2671 * Look for TCP socket matching *tuple*, optionally in a child
2672 * network namespace *netns*. The return value must be checked,
2673 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2675 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2676 * that it also returns timewait or request sockets. Use
2677 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2680 * This helper is available only if the kernel was compiled with
2681 * **CONFIG_NET** configuration option.
2683 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2684 * For sockets with reuseport option, the **struct bpf_sock**
2685 * result is from *reuse*\ **->socks**\ [] using the hash of the
2688 * long bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2690 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2691 * the listening socket in *sk*.
2693 * *iph* points to the start of the IPv4 or IPv6 header, while
2694 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2695 * **sizeof**\ (**struct ip6hdr**).
2697 * *th* points to the start of the TCP header, while *th_len*
2698 * contains **sizeof**\ (**struct tcphdr**).
2700 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2703 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2705 * Get name of sysctl in /proc/sys/ and copy it into provided by
2706 * program buffer *buf* of size *buf_len*.
2708 * The buffer is always NUL terminated, unless it's zero-sized.
2710 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2711 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2712 * only (e.g. "tcp_mem").
2714 * Number of character copied (not including the trailing NUL).
2716 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2717 * truncated name in this case).
2719 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2721 * Get current value of sysctl as it is presented in /proc/sys
2722 * (incl. newline, etc), and copy it as a string into provided
2723 * by program buffer *buf* of size *buf_len*.
2725 * The whole value is copied, no matter what file position user
2726 * space issued e.g. sys_read at.
2728 * The buffer is always NUL terminated, unless it's zero-sized.
2730 * Number of character copied (not including the trailing NUL).
2732 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2733 * truncated name in this case).
2735 * **-EINVAL** if current value was unavailable, e.g. because
2736 * sysctl is uninitialized and read returns -EIO for it.
2738 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2740 * Get new value being written by user space to sysctl (before
2741 * the actual write happens) and copy it as a string into
2742 * provided by program buffer *buf* of size *buf_len*.
2744 * User space may write new value at file position > 0.
2746 * The buffer is always NUL terminated, unless it's zero-sized.
2748 * Number of character copied (not including the trailing NUL).
2750 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2751 * truncated name in this case).
2753 * **-EINVAL** if sysctl is being read.
2755 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2757 * Override new value being written by user space to sysctl with
2758 * value provided by program in buffer *buf* of size *buf_len*.
2760 * *buf* should contain a string in same form as provided by user
2761 * space on sysctl write.
2763 * User space may write new value at file position > 0. To override
2764 * the whole sysctl value file position should be set to zero.
2768 * **-E2BIG** if the *buf_len* is too big.
2770 * **-EINVAL** if sysctl is being read.
2772 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2774 * Convert the initial part of the string from buffer *buf* of
2775 * size *buf_len* to a long integer according to the given base
2776 * and save the result in *res*.
2778 * The string may begin with an arbitrary amount of white space
2779 * (as determined by **isspace**\ (3)) followed by a single
2780 * optional '**-**' sign.
2782 * Five least significant bits of *flags* encode base, other bits
2783 * are currently unused.
2785 * Base must be either 8, 10, 16 or 0 to detect it automatically
2786 * similar to user space **strtol**\ (3).
2788 * Number of characters consumed on success. Must be positive but
2789 * no more than *buf_len*.
2791 * **-EINVAL** if no valid digits were found or unsupported base
2794 * **-ERANGE** if resulting value was out of range.
2796 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2798 * Convert the initial part of the string from buffer *buf* of
2799 * size *buf_len* to an unsigned long integer according to the
2800 * given base and save the result in *res*.
2802 * The string may begin with an arbitrary amount of white space
2803 * (as determined by **isspace**\ (3)).
2805 * Five least significant bits of *flags* encode base, other bits
2806 * are currently unused.
2808 * Base must be either 8, 10, 16 or 0 to detect it automatically
2809 * similar to user space **strtoul**\ (3).
2811 * Number of characters consumed on success. Must be positive but
2812 * no more than *buf_len*.
2814 * **-EINVAL** if no valid digits were found or unsupported base
2817 * **-ERANGE** if resulting value was out of range.
2819 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
2821 * Get a bpf-local-storage from a *sk*.
2823 * Logically, it could be thought of getting the value from
2824 * a *map* with *sk* as the **key**. From this
2825 * perspective, the usage is not much different from
2826 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2827 * helper enforces the key must be a full socket and the map must
2828 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2830 * Underneath, the value is stored locally at *sk* instead of
2831 * the *map*. The *map* is used as the bpf-local-storage
2832 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2833 * searched against all bpf-local-storages residing at *sk*.
2835 * *sk* is a kernel **struct sock** pointer for LSM program.
2836 * *sk* is a **struct bpf_sock** pointer for other program types.
2838 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2839 * used such that a new bpf-local-storage will be
2840 * created if one does not exist. *value* can be used
2841 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2842 * the initial value of a bpf-local-storage. If *value* is
2843 * **NULL**, the new bpf-local-storage will be zero initialized.
2845 * A bpf-local-storage pointer is returned on success.
2847 * **NULL** if not found or there was an error in adding
2848 * a new bpf-local-storage.
2850 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
2852 * Delete a bpf-local-storage from a *sk*.
2856 * **-ENOENT** if the bpf-local-storage cannot be found.
2858 * long bpf_send_signal(u32 sig)
2860 * Send signal *sig* to the process of the current task.
2861 * The signal may be delivered to any of this process's threads.
2863 * 0 on success or successfully queued.
2865 * **-EBUSY** if work queue under nmi is full.
2867 * **-EINVAL** if *sig* is invalid.
2869 * **-EPERM** if no permission to send the *sig*.
2871 * **-EAGAIN** if bpf program can try again.
2873 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2875 * Try to issue a SYN cookie for the packet with corresponding
2876 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2878 * *iph* points to the start of the IPv4 or IPv6 header, while
2879 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2880 * **sizeof**\ (**struct ip6hdr**).
2882 * *th* points to the start of the TCP header, while *th_len*
2883 * contains the length of the TCP header.
2885 * On success, lower 32 bits hold the generated SYN cookie in
2886 * followed by 16 bits which hold the MSS value for that cookie,
2887 * and the top 16 bits are unused.
2889 * On failure, the returned value is one of the following:
2891 * **-EINVAL** SYN cookie cannot be issued due to error
2893 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2895 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2897 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2899 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2901 * Write raw *data* blob into a special BPF perf event held by
2902 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2903 * event must have the following attributes: **PERF_SAMPLE_RAW**
2904 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2905 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2907 * The *flags* are used to indicate the index in *map* for which
2908 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2909 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2910 * to indicate that the index of the current CPU core should be
2913 * The value to write, of *size*, is passed through eBPF stack and
2914 * pointed by *data*.
2916 * *ctx* is a pointer to in-kernel struct sk_buff.
2918 * This helper is similar to **bpf_perf_event_output**\ () but
2919 * restricted to raw_tracepoint bpf programs.
2921 * 0 on success, or a negative error in case of failure.
2923 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2925 * Safely attempt to read *size* bytes from user space address
2926 * *unsafe_ptr* and store the data in *dst*.
2928 * 0 on success, or a negative error in case of failure.
2930 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2932 * Safely attempt to read *size* bytes from kernel space address
2933 * *unsafe_ptr* and store the data in *dst*.
2935 * 0 on success, or a negative error in case of failure.
2937 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2939 * Copy a NUL terminated string from an unsafe user address
2940 * *unsafe_ptr* to *dst*. The *size* should include the
2941 * terminating NUL byte. In case the string length is smaller than
2942 * *size*, the target is not padded with further NUL bytes. If the
2943 * string length is larger than *size*, just *size*-1 bytes are
2944 * copied and the last byte is set to NUL.
2946 * On success, the length of the copied string is returned. This
2947 * makes this helper useful in tracing programs for reading
2948 * strings, and more importantly to get its length at runtime. See
2949 * the following snippet:
2953 * SEC("kprobe/sys_open")
2954 * void bpf_sys_open(struct pt_regs *ctx)
2956 * char buf[PATHLEN]; // PATHLEN is defined to 256
2957 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2960 * // Consume buf, for example push it to
2961 * // userspace via bpf_perf_event_output(); we
2962 * // can use res (the string length) as event
2963 * // size, after checking its boundaries.
2966 * In comparison, using **bpf_probe_read_user**\ () helper here
2967 * instead to read the string would require to estimate the length
2968 * at compile time, and would often result in copying more memory
2971 * Another useful use case is when parsing individual process
2972 * arguments or individual environment variables navigating
2973 * *current*\ **->mm->arg_start** and *current*\
2974 * **->mm->env_start**: using this helper and the return value,
2975 * one can quickly iterate at the right offset of the memory area.
2977 * On success, the strictly positive length of the string,
2978 * including the trailing NUL character. On error, a negative
2981 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2983 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
2984 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
2986 * On success, the strictly positive length of the string, including
2987 * the trailing NUL character. On error, a negative value.
2989 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2991 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
2992 * *rcv_nxt* is the ack_seq to be sent out.
2994 * 0 on success, or a negative error in case of failure.
2996 * long bpf_send_signal_thread(u32 sig)
2998 * Send signal *sig* to the thread corresponding to the current task.
3000 * 0 on success or successfully queued.
3002 * **-EBUSY** if work queue under nmi is full.
3004 * **-EINVAL** if *sig* is invalid.
3006 * **-EPERM** if no permission to send the *sig*.
3008 * **-EAGAIN** if bpf program can try again.
3010 * u64 bpf_jiffies64(void)
3012 * Obtain the 64bit jiffies
3014 * The 64 bit jiffies
3016 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3018 * For an eBPF program attached to a perf event, retrieve the
3019 * branch records (**struct perf_branch_entry**) associated to *ctx*
3020 * and store it in the buffer pointed by *buf* up to size
3023 * On success, number of bytes written to *buf*. On error, a
3026 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3027 * instead return the number of bytes required to store all the
3028 * branch entries. If this flag is set, *buf* may be NULL.
3030 * **-EINVAL** if arguments invalid or **size** not a multiple
3031 * of **sizeof**\ (**struct perf_branch_entry**\ ).
3033 * **-ENOENT** if architecture does not support branch records.
3035 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3037 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3038 * *namespace* will be returned in *nsdata*.
3040 * 0 on success, or one of the following in case of failure:
3042 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3043 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3045 * **-ENOENT** if pidns does not exists for the current task.
3047 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3049 * Write raw *data* blob into a special BPF perf event held by
3050 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3051 * event must have the following attributes: **PERF_SAMPLE_RAW**
3052 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3053 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3055 * The *flags* are used to indicate the index in *map* for which
3056 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3057 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3058 * to indicate that the index of the current CPU core should be
3061 * The value to write, of *size*, is passed through eBPF stack and
3062 * pointed by *data*.
3064 * *ctx* is a pointer to in-kernel struct xdp_buff.
3066 * This helper is similar to **bpf_perf_eventoutput**\ () but
3067 * restricted to raw_tracepoint bpf programs.
3069 * 0 on success, or a negative error in case of failure.
3071 * u64 bpf_get_netns_cookie(void *ctx)
3073 * Retrieve the cookie (generated by the kernel) of the network
3074 * namespace the input *ctx* is associated with. The network
3075 * namespace cookie remains stable for its lifetime and provides
3076 * a global identifier that can be assumed unique. If *ctx* is
3077 * NULL, then the helper returns the cookie for the initial
3078 * network namespace. The cookie itself is very similar to that
3079 * of **bpf_get_socket_cookie**\ () helper, but for network
3080 * namespaces instead of sockets.
3082 * A 8-byte long opaque number.
3084 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3086 * Return id of cgroup v2 that is ancestor of the cgroup associated
3087 * with the current task at the *ancestor_level*. The root cgroup
3088 * is at *ancestor_level* zero and each step down the hierarchy
3089 * increments the level. If *ancestor_level* == level of cgroup
3090 * associated with the current task, then return value will be the
3091 * same as that of **bpf_get_current_cgroup_id**\ ().
3093 * The helper is useful to implement policies based on cgroups
3094 * that are upper in hierarchy than immediate cgroup associated
3095 * with the current task.
3097 * The format of returned id and helper limitations are same as in
3098 * **bpf_get_current_cgroup_id**\ ().
3100 * The id is returned or 0 in case the id could not be retrieved.
3102 * long bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags)
3104 * Helper is overloaded depending on BPF program type. This
3105 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3106 * **BPF_PROG_TYPE_SCHED_ACT** programs.
3108 * Assign the *sk* to the *skb*. When combined with appropriate
3109 * routing configuration to receive the packet towards the socket,
3110 * will cause *skb* to be delivered to the specified socket.
3111 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3112 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3113 * interfere with successful delivery to the socket.
3115 * This operation is only valid from TC ingress path.
3117 * The *flags* argument must be zero.
3119 * 0 on success, or a negative error in case of failure:
3121 * **-EINVAL** if specified *flags* are not supported.
3123 * **-ENOENT** if the socket is unavailable for assignment.
3125 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3127 * **-EOPNOTSUPP** if the operation is not supported, for example
3128 * a call from outside of TC ingress.
3130 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3133 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3135 * Helper is overloaded depending on BPF program type. This
3136 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
3138 * Select the *sk* as a result of a socket lookup.
3140 * For the operation to succeed passed socket must be compatible
3141 * with the packet description provided by the *ctx* object.
3143 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
3144 * be an exact match. While IP family (**AF_INET** or
3145 * **AF_INET6**) must be compatible, that is IPv6 sockets
3146 * that are not v6-only can be selected for IPv4 packets.
3148 * Only TCP listeners and UDP unconnected sockets can be
3149 * selected. *sk* can also be NULL to reset any previous
3152 * *flags* argument can combination of following values:
3154 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
3155 * socket selection, potentially done by a BPF program
3156 * that ran before us.
3158 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
3159 * load-balancing within reuseport group for the socket
3162 * On success *ctx->sk* will point to the selected socket.
3165 * 0 on success, or a negative errno in case of failure.
3167 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
3168 * not compatible with packet family (*ctx->family*).
3170 * * **-EEXIST** if socket has been already selected,
3171 * potentially by another program, and
3172 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
3174 * * **-EINVAL** if unsupported flags were specified.
3176 * * **-EPROTOTYPE** if socket L4 protocol
3177 * (*sk->protocol*) doesn't match packet protocol
3178 * (*ctx->protocol*).
3180 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
3181 * state (TCP listening or UDP unconnected).
3183 * u64 bpf_ktime_get_boot_ns(void)
3185 * Return the time elapsed since system boot, in nanoseconds.
3186 * Does include the time the system was suspended.
3187 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
3191 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3193 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3194 * out the format string.
3195 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3196 * the format string itself. The *data* and *data_len* are format string
3197 * arguments. The *data* are a **u64** array and corresponding format string
3198 * values are stored in the array. For strings and pointers where pointees
3199 * are accessed, only the pointer values are stored in the *data* array.
3200 * The *data_len* is the size of *data* in bytes.
3202 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3203 * Reading kernel memory may fail due to either invalid address or
3204 * valid address but requiring a major memory fault. If reading kernel memory
3205 * fails, the string for **%s** will be an empty string, and the ip
3206 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
3207 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
3209 * 0 on success, or a negative error in case of failure:
3211 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3212 * by returning 1 from bpf program.
3214 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3216 * **-E2BIG** if *fmt* contains too many format specifiers.
3218 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3220 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
3222 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
3223 * The *m* represents the seq_file. The *data* and *len* represent the
3224 * data to write in bytes.
3226 * 0 on success, or a negative error in case of failure:
3228 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3230 * u64 bpf_sk_cgroup_id(struct bpf_sock *sk)
3232 * Return the cgroup v2 id of the socket *sk*.
3234 * *sk* must be a non-**NULL** pointer to a full socket, e.g. one
3235 * returned from **bpf_sk_lookup_xxx**\ (),
3236 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3237 * same as in **bpf_skb_cgroup_id**\ ().
3239 * This helper is available only if the kernel was compiled with
3240 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3242 * The id is returned or 0 in case the id could not be retrieved.
3244 * u64 bpf_sk_ancestor_cgroup_id(struct bpf_sock *sk, int ancestor_level)
3246 * Return id of cgroup v2 that is ancestor of cgroup associated
3247 * with the *sk* at the *ancestor_level*. The root cgroup is at
3248 * *ancestor_level* zero and each step down the hierarchy
3249 * increments the level. If *ancestor_level* == level of cgroup
3250 * associated with *sk*, then return value will be same as that
3251 * of **bpf_sk_cgroup_id**\ ().
3253 * The helper is useful to implement policies based on cgroups
3254 * that are upper in hierarchy than immediate cgroup associated
3257 * The format of returned id and helper limitations are same as in
3258 * **bpf_sk_cgroup_id**\ ().
3260 * The id is returned or 0 in case the id could not be retrieved.
3262 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
3264 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
3265 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3266 * of new data availability is sent.
3267 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3268 * of new data availability is sent unconditionally.
3270 * 0 on success, or a negative error in case of failure.
3272 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3274 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3276 * Valid pointer with *size* bytes of memory available; NULL,
3279 * void bpf_ringbuf_submit(void *data, u64 flags)
3281 * Submit reserved ring buffer sample, pointed to by *data*.
3282 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3283 * of new data availability is sent.
3284 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3285 * of new data availability is sent unconditionally.
3287 * Nothing. Always succeeds.
3289 * void bpf_ringbuf_discard(void *data, u64 flags)
3291 * Discard reserved ring buffer sample, pointed to by *data*.
3292 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3293 * of new data availability is sent.
3294 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3295 * of new data availability is sent unconditionally.
3297 * Nothing. Always succeeds.
3299 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3301 * Query various characteristics of provided ring buffer. What
3302 * exactly is queries is determined by *flags*:
3304 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
3305 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
3306 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
3307 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
3309 * Data returned is just a momentary snapshot of actual values
3310 * and could be inaccurate, so this facility should be used to
3311 * power heuristics and for reporting, not to make 100% correct
3314 * Requested value, or 0, if *flags* are not recognized.
3316 * long bpf_csum_level(struct sk_buff *skb, u64 level)
3318 * Change the skbs checksum level by one layer up or down, or
3319 * reset it entirely to none in order to have the stack perform
3320 * checksum validation. The level is applicable to the following
3321 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
3322 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
3323 * through **bpf_skb_adjust_room**\ () helper with passing in
3324 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
3325 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
3326 * the UDP header is removed. Similarly, an encap of the latter
3327 * into the former could be accompanied by a helper call to
3328 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
3329 * skb is still intended to be processed in higher layers of the
3330 * stack instead of just egressing at tc.
3332 * There are three supported level settings at this time:
3334 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
3335 * with CHECKSUM_UNNECESSARY.
3336 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
3337 * with CHECKSUM_UNNECESSARY.
3338 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
3339 * sets CHECKSUM_NONE to force checksum validation by the stack.
3340 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
3343 * 0 on success, or a negative error in case of failure. In the
3344 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
3345 * is returned or the error code -EACCES in case the skb is not
3346 * subject to CHECKSUM_UNNECESSARY.
3348 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
3350 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
3352 * *sk* if casting is valid, or NULL otherwise.
3354 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
3356 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
3358 * *sk* if casting is valid, or NULL otherwise.
3360 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
3362 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
3364 * *sk* if casting is valid, or NULL otherwise.
3366 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
3368 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
3370 * *sk* if casting is valid, or NULL otherwise.
3372 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
3374 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
3376 * *sk* if casting is valid, or NULL otherwise.
3378 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
3380 * Return a user or a kernel stack in bpf program provided buffer.
3381 * To achieve this, the helper needs *task*, which is a valid
3382 * pointer to struct task_struct. To store the stacktrace, the
3383 * bpf program provides *buf* with a nonnegative *size*.
3385 * The last argument, *flags*, holds the number of stack frames to
3386 * skip (from 0 to 255), masked with
3387 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3388 * the following flags:
3390 * **BPF_F_USER_STACK**
3391 * Collect a user space stack instead of a kernel stack.
3392 * **BPF_F_USER_BUILD_ID**
3393 * Collect buildid+offset instead of ips for user stack,
3394 * only valid if **BPF_F_USER_STACK** is also specified.
3396 * **bpf_get_task_stack**\ () can collect up to
3397 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3398 * to sufficient large buffer size. Note that
3399 * this limit can be controlled with the **sysctl** program, and
3400 * that it should be manually increased in order to profile long
3401 * user stacks (such as stacks for Java programs). To do so, use:
3405 * # sysctl kernel.perf_event_max_stack=<new value>
3407 * A non-negative value equal to or less than *size* on success,
3408 * or a negative error in case of failure.
3410 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
3412 * Load header option. Support reading a particular TCP header
3413 * option for bpf program (BPF_PROG_TYPE_SOCK_OPS).
3415 * If *flags* is 0, it will search the option from the
3416 * sock_ops->skb_data. The comment in "struct bpf_sock_ops"
3417 * has details on what skb_data contains under different
3420 * The first byte of the *searchby_res* specifies the
3421 * kind that it wants to search.
3423 * If the searching kind is an experimental kind
3424 * (i.e. 253 or 254 according to RFC6994). It also
3425 * needs to specify the "magic" which is either
3426 * 2 bytes or 4 bytes. It then also needs to
3427 * specify the size of the magic by using
3428 * the 2nd byte which is "kind-length" of a TCP
3429 * header option and the "kind-length" also
3430 * includes the first 2 bytes "kind" and "kind-length"
3431 * itself as a normal TCP header option also does.
3433 * For example, to search experimental kind 254 with
3434 * 2 byte magic 0xeB9F, the searchby_res should be
3435 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
3437 * To search for the standard window scale option (3),
3438 * the searchby_res should be [ 3, 0, 0, .... 0 ].
3439 * Note, kind-length must be 0 for regular option.
3441 * Searching for No-Op (0) and End-of-Option-List (1) are
3444 * *len* must be at least 2 bytes which is the minimal size
3445 * of a header option.
3448 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
3449 * saved_syn packet or the just-received syn packet.
3452 * >0 when found, the header option is copied to *searchby_res*.
3453 * The return value is the total length copied.
3455 * **-EINVAL** If param is invalid
3457 * **-ENOMSG** The option is not found
3459 * **-ENOENT** No syn packet available when
3460 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used
3462 * **-ENOSPC** Not enough space. Only *len* number of
3465 * **-EFAULT** Cannot parse the header options in the packet
3467 * **-EPERM** This helper cannot be used under the
3468 * current sock_ops->op.
3470 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
3472 * Store header option. The data will be copied
3473 * from buffer *from* with length *len* to the TCP header.
3475 * The buffer *from* should have the whole option that
3476 * includes the kind, kind-length, and the actual
3477 * option data. The *len* must be at least kind-length
3478 * long. The kind-length does not have to be 4 byte
3479 * aligned. The kernel will take care of the padding
3480 * and setting the 4 bytes aligned value to th->doff.
3482 * This helper will check for duplicated option
3483 * by searching the same option in the outgoing skb.
3485 * This helper can only be called during
3486 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
3489 * 0 on success, or negative error in case of failure:
3491 * **-EINVAL** If param is invalid
3493 * **-ENOSPC** Not enough space in the header.
3494 * Nothing has been written
3496 * **-EEXIST** The option has already existed
3498 * **-EFAULT** Cannot parse the existing header options
3500 * **-EPERM** This helper cannot be used under the
3501 * current sock_ops->op.
3503 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
3505 * Reserve *len* bytes for the bpf header option. The
3506 * space will be used by bpf_store_hdr_opt() later in
3507 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
3509 * If bpf_reserve_hdr_opt() is called multiple times,
3510 * the total number of bytes will be reserved.
3512 * This helper can only be called during
3513 * BPF_SOCK_OPS_HDR_OPT_LEN_CB.
3516 * 0 on success, or negative error in case of failure:
3518 * **-EINVAL** if param is invalid
3520 * **-ENOSPC** Not enough space in the header.
3522 * **-EPERM** This helper cannot be used under the
3523 * current sock_ops->op.
3525 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
3527 * Get a bpf_local_storage from an *inode*.
3529 * Logically, it could be thought of as getting the value from
3530 * a *map* with *inode* as the **key**. From this
3531 * perspective, the usage is not much different from
3532 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
3533 * helper enforces the key must be an inode and the map must also
3534 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
3536 * Underneath, the value is stored locally at *inode* instead of
3537 * the *map*. The *map* is used as the bpf-local-storage
3538 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3539 * searched against all bpf_local_storage residing at *inode*.
3541 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
3542 * used such that a new bpf_local_storage will be
3543 * created if one does not exist. *value* can be used
3544 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
3545 * the initial value of a bpf_local_storage. If *value* is
3546 * **NULL**, the new bpf_local_storage will be zero initialized.
3548 * A bpf_local_storage pointer is returned on success.
3550 * **NULL** if not found or there was an error in adding
3551 * a new bpf_local_storage.
3553 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
3555 * Delete a bpf_local_storage from an *inode*.
3559 * **-ENOENT** if the bpf_local_storage cannot be found.
3561 * long bpf_d_path(struct path *path, char *buf, u32 sz)
3563 * Return full path for given 'struct path' object, which
3564 * needs to be the kernel BTF 'path' object. The path is
3565 * returned in the provided buffer 'buf' of size 'sz' and
3566 * is zero terminated.
3569 * On success, the strictly positive length of the string,
3570 * including the trailing NUL character. On error, a negative
3573 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
3575 * Read *size* bytes from user space address *user_ptr* and store
3576 * the data in *dst*. This is a wrapper of copy_from_user().
3578 * 0 on success, or a negative error in case of failure.
3580 #define __BPF_FUNC_MAPPER(FN) \
3582 FN(map_lookup_elem), \
3583 FN(map_update_elem), \
3584 FN(map_delete_elem), \
3588 FN(get_prandom_u32), \
3589 FN(get_smp_processor_id), \
3590 FN(skb_store_bytes), \
3591 FN(l3_csum_replace), \
3592 FN(l4_csum_replace), \
3594 FN(clone_redirect), \
3595 FN(get_current_pid_tgid), \
3596 FN(get_current_uid_gid), \
3597 FN(get_current_comm), \
3598 FN(get_cgroup_classid), \
3599 FN(skb_vlan_push), \
3601 FN(skb_get_tunnel_key), \
3602 FN(skb_set_tunnel_key), \
3603 FN(perf_event_read), \
3605 FN(get_route_realm), \
3606 FN(perf_event_output), \
3607 FN(skb_load_bytes), \
3610 FN(skb_get_tunnel_opt), \
3611 FN(skb_set_tunnel_opt), \
3612 FN(skb_change_proto), \
3613 FN(skb_change_type), \
3614 FN(skb_under_cgroup), \
3615 FN(get_hash_recalc), \
3616 FN(get_current_task), \
3617 FN(probe_write_user), \
3618 FN(current_task_under_cgroup), \
3619 FN(skb_change_tail), \
3620 FN(skb_pull_data), \
3622 FN(set_hash_invalid), \
3623 FN(get_numa_node_id), \
3624 FN(skb_change_head), \
3625 FN(xdp_adjust_head), \
3626 FN(probe_read_str), \
3627 FN(get_socket_cookie), \
3628 FN(get_socket_uid), \
3631 FN(skb_adjust_room), \
3633 FN(sk_redirect_map), \
3634 FN(sock_map_update), \
3635 FN(xdp_adjust_meta), \
3636 FN(perf_event_read_value), \
3637 FN(perf_prog_read_value), \
3639 FN(override_return), \
3640 FN(sock_ops_cb_flags_set), \
3641 FN(msg_redirect_map), \
3642 FN(msg_apply_bytes), \
3643 FN(msg_cork_bytes), \
3644 FN(msg_pull_data), \
3646 FN(xdp_adjust_tail), \
3647 FN(skb_get_xfrm_state), \
3649 FN(skb_load_bytes_relative), \
3651 FN(sock_hash_update), \
3652 FN(msg_redirect_hash), \
3653 FN(sk_redirect_hash), \
3654 FN(lwt_push_encap), \
3655 FN(lwt_seg6_store_bytes), \
3656 FN(lwt_seg6_adjust_srh), \
3657 FN(lwt_seg6_action), \
3660 FN(skb_cgroup_id), \
3661 FN(get_current_cgroup_id), \
3662 FN(get_local_storage), \
3663 FN(sk_select_reuseport), \
3664 FN(skb_ancestor_cgroup_id), \
3665 FN(sk_lookup_tcp), \
3666 FN(sk_lookup_udp), \
3668 FN(map_push_elem), \
3670 FN(map_peek_elem), \
3671 FN(msg_push_data), \
3673 FN(rc_pointer_rel), \
3678 FN(skb_ecn_set_ce), \
3679 FN(get_listener_sock), \
3680 FN(skc_lookup_tcp), \
3681 FN(tcp_check_syncookie), \
3682 FN(sysctl_get_name), \
3683 FN(sysctl_get_current_value), \
3684 FN(sysctl_get_new_value), \
3685 FN(sysctl_set_new_value), \
3688 FN(sk_storage_get), \
3689 FN(sk_storage_delete), \
3691 FN(tcp_gen_syncookie), \
3693 FN(probe_read_user), \
3694 FN(probe_read_kernel), \
3695 FN(probe_read_user_str), \
3696 FN(probe_read_kernel_str), \
3698 FN(send_signal_thread), \
3700 FN(read_branch_records), \
3701 FN(get_ns_current_pid_tgid), \
3703 FN(get_netns_cookie), \
3704 FN(get_current_ancestor_cgroup_id), \
3706 FN(ktime_get_boot_ns), \
3710 FN(sk_ancestor_cgroup_id), \
3711 FN(ringbuf_output), \
3712 FN(ringbuf_reserve), \
3713 FN(ringbuf_submit), \
3714 FN(ringbuf_discard), \
3715 FN(ringbuf_query), \
3717 FN(skc_to_tcp6_sock), \
3718 FN(skc_to_tcp_sock), \
3719 FN(skc_to_tcp_timewait_sock), \
3720 FN(skc_to_tcp_request_sock), \
3721 FN(skc_to_udp6_sock), \
3722 FN(get_task_stack), \
3724 FN(store_hdr_opt), \
3725 FN(reserve_hdr_opt), \
3726 FN(inode_storage_get), \
3727 FN(inode_storage_delete), \
3729 FN(copy_from_user), \
3732 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
3733 * function eBPF program intends to call
3735 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
3737 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
3740 #undef __BPF_ENUM_FN
3742 /* All flags used by eBPF helper functions, placed here. */
3744 /* BPF_FUNC_skb_store_bytes flags. */
3746 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3747 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3750 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3751 * First 4 bits are for passing the header field size.
3754 BPF_F_HDR_FIELD_MASK = 0xfULL,
3757 /* BPF_FUNC_l4_csum_replace flags. */
3759 BPF_F_PSEUDO_HDR = (1ULL << 4),
3760 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3761 BPF_F_MARK_ENFORCE = (1ULL << 6),
3764 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
3766 BPF_F_INGRESS = (1ULL << 0),
3769 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
3771 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3774 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
3776 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3777 BPF_F_USER_STACK = (1ULL << 8),
3778 /* flags used by BPF_FUNC_get_stackid only. */
3779 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3780 BPF_F_REUSE_STACKID = (1ULL << 10),
3781 /* flags used by BPF_FUNC_get_stack only. */
3782 BPF_F_USER_BUILD_ID = (1ULL << 11),
3785 /* BPF_FUNC_skb_set_tunnel_key flags. */
3787 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3788 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3789 BPF_F_SEQ_NUMBER = (1ULL << 3),
3792 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3793 * BPF_FUNC_perf_event_read_value flags.
3796 BPF_F_INDEX_MASK = 0xffffffffULL,
3797 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
3798 /* BPF_FUNC_perf_event_output for sk_buff input context. */
3799 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3802 /* Current network namespace */
3804 BPF_F_CURRENT_NETNS = (-1L),
3807 /* BPF_FUNC_csum_level level values. */
3809 BPF_CSUM_LEVEL_QUERY,
3812 BPF_CSUM_LEVEL_RESET,
3815 /* BPF_FUNC_skb_adjust_room flags. */
3817 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3818 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3819 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3820 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3821 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
3822 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
3826 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3827 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3830 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
3831 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3832 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
3834 /* BPF_FUNC_sysctl_get_name flags. */
3836 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3839 /* BPF_FUNC_<kernel_obj>_storage_get flags */
3841 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
3842 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
3843 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
3845 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
3848 /* BPF_FUNC_read_branch_records flags. */
3850 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3853 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
3854 * BPF_FUNC_bpf_ringbuf_output flags.
3857 BPF_RB_NO_WAKEUP = (1ULL << 0),
3858 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
3861 /* BPF_FUNC_bpf_ringbuf_query flags */
3863 BPF_RB_AVAIL_DATA = 0,
3864 BPF_RB_RING_SIZE = 1,
3865 BPF_RB_CONS_POS = 2,
3866 BPF_RB_PROD_POS = 3,
3869 /* BPF ring buffer constants */
3871 BPF_RINGBUF_BUSY_BIT = (1U << 31),
3872 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
3873 BPF_RINGBUF_HDR_SZ = 8,
3876 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
3878 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
3879 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
3882 /* Mode for BPF_FUNC_skb_adjust_room helper. */
3883 enum bpf_adj_room_mode {
3888 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3889 enum bpf_hdr_start_off {
3894 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3895 enum bpf_lwt_encap_mode {
3897 BPF_LWT_ENCAP_SEG6_INLINE,
3901 #define __bpf_md_ptr(type, name) \
3905 } __attribute__((aligned(8)))
3907 /* user accessible mirror of in-kernel sk_buff.
3908 * new fields can only be added to the end of this structure
3914 __u32 queue_mapping;
3920 __u32 ingress_ifindex;
3930 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
3932 __u32 remote_ip4; /* Stored in network byte order */
3933 __u32 local_ip4; /* Stored in network byte order */
3934 __u32 remote_ip6[4]; /* Stored in network byte order */
3935 __u32 local_ip6[4]; /* Stored in network byte order */
3936 __u32 remote_port; /* Stored in network byte order */
3937 __u32 local_port; /* stored in host byte order */
3941 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
3945 __bpf_md_ptr(struct bpf_sock *, sk);
3949 struct bpf_tunnel_key {
3953 __u32 remote_ipv6[4];
3957 __u16 tunnel_ext; /* Padding, future use. */
3961 /* user accessible mirror of in-kernel xfrm_state.
3962 * new fields can only be added to the end of this structure
3964 struct bpf_xfrm_state {
3966 __u32 spi; /* Stored in network byte order */
3968 __u16 ext; /* Padding, future use. */
3970 __u32 remote_ipv4; /* Stored in network byte order */
3971 __u32 remote_ipv6[4]; /* Stored in network byte order */
3975 /* Generic BPF return codes which all BPF program types may support.
3976 * The values are binary compatible with their TC_ACT_* counter-part to
3977 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3980 * XDP is handled seprately, see XDP_*.
3988 /* >127 are reserved for prog type specific return codes.
3990 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3991 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3992 * changed and should be routed based on its new L3 header.
3993 * (This is an L3 redirect, as opposed to L2 redirect
3994 * represented by BPF_REDIRECT above).
3996 BPF_LWT_REROUTE = 128,
4006 /* IP address also allows 1 and 2 bytes access */
4009 __u32 src_port; /* host byte order */
4010 __u32 dst_port; /* network byte order */
4014 __s32 rx_queue_mapping;
4017 struct bpf_tcp_sock {
4018 __u32 snd_cwnd; /* Sending congestion window */
4019 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
4021 __u32 snd_ssthresh; /* Slow start size threshold */
4022 __u32 rcv_nxt; /* What we want to receive next */
4023 __u32 snd_nxt; /* Next sequence we send */
4024 __u32 snd_una; /* First byte we want an ack for */
4025 __u32 mss_cache; /* Cached effective mss, not including SACKS */
4026 __u32 ecn_flags; /* ECN status bits. */
4027 __u32 rate_delivered; /* saved rate sample: packets delivered */
4028 __u32 rate_interval_us; /* saved rate sample: time elapsed */
4029 __u32 packets_out; /* Packets which are "in flight" */
4030 __u32 retrans_out; /* Retransmitted packets out */
4031 __u32 total_retrans; /* Total retransmits for entire connection */
4032 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
4033 * total number of segments in.
4035 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
4036 * total number of data segments in.
4038 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
4039 * The total number of segments sent.
4041 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
4042 * total number of data segments sent.
4044 __u32 lost_out; /* Lost packets */
4045 __u32 sacked_out; /* SACK'd packets */
4046 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
4047 * sum(delta(rcv_nxt)), or how many bytes
4050 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
4051 * sum(delta(snd_una)), or how many bytes
4054 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
4055 * total number of DSACK blocks received
4057 __u32 delivered; /* Total data packets delivered incl. rexmits */
4058 __u32 delivered_ce; /* Like the above but only ECE marked packets */
4059 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
4062 struct bpf_sock_tuple {
4079 struct bpf_xdp_sock {
4083 #define XDP_PACKET_HEADROOM 256
4085 /* User return codes for XDP prog type.
4086 * A valid XDP program must return one of these defined values. All other
4087 * return codes are reserved for future use. Unknown return codes will
4088 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
4098 /* user accessible metadata for XDP packet hook
4099 * new fields must be added to the end of this structure
4105 /* Below access go through struct xdp_rxq_info */
4106 __u32 ingress_ifindex; /* rxq->dev->ifindex */
4107 __u32 rx_queue_index; /* rxq->queue_index */
4109 __u32 egress_ifindex; /* txq->dev->ifindex */
4112 /* DEVMAP map-value layout
4114 * The struct data-layout of map-value is a configuration interface.
4115 * New members can only be added to the end of this structure.
4117 struct bpf_devmap_val {
4118 __u32 ifindex; /* device index */
4120 int fd; /* prog fd on map write */
4121 __u32 id; /* prog id on map read */
4125 /* CPUMAP map-value layout
4127 * The struct data-layout of map-value is a configuration interface.
4128 * New members can only be added to the end of this structure.
4130 struct bpf_cpumap_val {
4131 __u32 qsize; /* queue size to remote target CPU */
4133 int fd; /* prog fd on map write */
4134 __u32 id; /* prog id on map read */
4143 /* user accessible metadata for SK_MSG packet hook, new fields must
4144 * be added to the end of this structure
4147 __bpf_md_ptr(void *, data);
4148 __bpf_md_ptr(void *, data_end);
4151 __u32 remote_ip4; /* Stored in network byte order */
4152 __u32 local_ip4; /* Stored in network byte order */
4153 __u32 remote_ip6[4]; /* Stored in network byte order */
4154 __u32 local_ip6[4]; /* Stored in network byte order */
4155 __u32 remote_port; /* Stored in network byte order */
4156 __u32 local_port; /* stored in host byte order */
4157 __u32 size; /* Total size of sk_msg */
4159 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
4162 struct sk_reuseport_md {
4164 * Start of directly accessible data. It begins from
4165 * the tcp/udp header.
4167 __bpf_md_ptr(void *, data);
4168 /* End of directly accessible data */
4169 __bpf_md_ptr(void *, data_end);
4171 * Total length of packet (starting from the tcp/udp header).
4172 * Note that the directly accessible bytes (data_end - data)
4173 * could be less than this "len". Those bytes could be
4174 * indirectly read by a helper "bpf_skb_load_bytes()".
4178 * Eth protocol in the mac header (network byte order). e.g.
4179 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
4182 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
4183 __u32 bind_inany; /* Is sock bound to an INANY address? */
4184 __u32 hash; /* A hash of the packet 4 tuples */
4187 #define BPF_TAG_SIZE 8
4189 struct bpf_prog_info {
4192 __u8 tag[BPF_TAG_SIZE];
4193 __u32 jited_prog_len;
4194 __u32 xlated_prog_len;
4195 __aligned_u64 jited_prog_insns;
4196 __aligned_u64 xlated_prog_insns;
4197 __u64 load_time; /* ns since boottime */
4198 __u32 created_by_uid;
4200 __aligned_u64 map_ids;
4201 char name[BPF_OBJ_NAME_LEN];
4203 __u32 gpl_compatible:1;
4204 __u32 :31; /* alignment pad */
4207 __u32 nr_jited_ksyms;
4208 __u32 nr_jited_func_lens;
4209 __aligned_u64 jited_ksyms;
4210 __aligned_u64 jited_func_lens;
4212 __u32 func_info_rec_size;
4213 __aligned_u64 func_info;
4216 __aligned_u64 line_info;
4217 __aligned_u64 jited_line_info;
4218 __u32 nr_jited_line_info;
4219 __u32 line_info_rec_size;
4220 __u32 jited_line_info_rec_size;
4222 __aligned_u64 prog_tags;
4225 } __attribute__((aligned(8)));
4227 struct bpf_map_info {
4234 char name[BPF_OBJ_NAME_LEN];
4236 __u32 btf_vmlinux_value_type_id;
4240 __u32 btf_key_type_id;
4241 __u32 btf_value_type_id;
4242 } __attribute__((aligned(8)));
4244 struct bpf_btf_info {
4248 } __attribute__((aligned(8)));
4250 struct bpf_link_info {
4256 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
4257 __u32 tp_name_len; /* in/out: tp_name buffer len */
4267 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
4268 __u32 target_name_len; /* in/out: target_name buffer len */
4283 } __attribute__((aligned(8)));
4285 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
4286 * by user and intended to be used by socket (e.g. to bind to, depends on
4289 struct bpf_sock_addr {
4290 __u32 user_family; /* Allows 4-byte read, but no write. */
4291 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4292 * Stored in network byte order.
4294 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4295 * Stored in network byte order.
4297 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
4298 * Stored in network byte order
4300 __u32 family; /* Allows 4-byte read, but no write */
4301 __u32 type; /* Allows 4-byte read, but no write */
4302 __u32 protocol; /* Allows 4-byte read, but no write */
4303 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4304 * Stored in network byte order.
4306 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4307 * Stored in network byte order.
4309 __bpf_md_ptr(struct bpf_sock *, sk);
4312 /* User bpf_sock_ops struct to access socket values and specify request ops
4313 * and their replies.
4314 * Some of this fields are in network (bigendian) byte order and may need
4315 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
4316 * New fields can only be added at the end of this structure
4318 struct bpf_sock_ops {
4321 __u32 args[4]; /* Optionally passed to bpf program */
4322 __u32 reply; /* Returned by bpf program */
4323 __u32 replylong[4]; /* Optionally returned by bpf prog */
4326 __u32 remote_ip4; /* Stored in network byte order */
4327 __u32 local_ip4; /* Stored in network byte order */
4328 __u32 remote_ip6[4]; /* Stored in network byte order */
4329 __u32 local_ip6[4]; /* Stored in network byte order */
4330 __u32 remote_port; /* Stored in network byte order */
4331 __u32 local_port; /* stored in host byte order */
4332 __u32 is_fullsock; /* Some TCP fields are only valid if
4333 * there is a full socket. If not, the
4334 * fields read as zero.
4337 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
4338 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
4347 __u32 rate_delivered;
4348 __u32 rate_interval_us;
4351 __u32 total_retrans;
4355 __u32 data_segs_out;
4359 __u64 bytes_received;
4361 __bpf_md_ptr(struct bpf_sock *, sk);
4362 /* [skb_data, skb_data_end) covers the whole TCP header.
4364 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
4365 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
4366 * header has not been written.
4367 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
4368 * been written so far.
4369 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
4371 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
4374 * bpf_load_hdr_opt() can also be used to read a particular option.
4376 __bpf_md_ptr(void *, skb_data);
4377 __bpf_md_ptr(void *, skb_data_end);
4378 __u32 skb_len; /* The total length of a packet.
4379 * It includes the header, options,
4382 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
4383 * an easy way to check for tcp_flags
4384 * without parsing skb_data.
4386 * In particular, the skb_tcp_flags
4387 * will still be available in
4388 * BPF_SOCK_OPS_HDR_OPT_LEN even though
4389 * the outgoing header has not
4394 /* Definitions for bpf_sock_ops_cb_flags */
4396 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
4397 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
4398 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
4399 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
4400 /* Call bpf for all received TCP headers. The bpf prog will be
4401 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4403 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4404 * for the header option related helpers that will be useful
4405 * to the bpf programs.
4407 * It could be used at the client/active side (i.e. connect() side)
4408 * when the server told it that the server was in syncookie
4409 * mode and required the active side to resend the bpf-written
4410 * options. The active side can keep writing the bpf-options until
4411 * it received a valid packet from the server side to confirm
4412 * the earlier packet (and options) has been received. The later
4413 * example patch is using it like this at the active side when the
4414 * server is in syncookie mode.
4416 * The bpf prog will usually turn this off in the common cases.
4418 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
4419 /* Call bpf when kernel has received a header option that
4420 * the kernel cannot handle. The bpf prog will be called under
4421 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
4423 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4424 * for the header option related helpers that will be useful
4425 * to the bpf programs.
4427 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
4428 /* Call bpf when the kernel is writing header options for the
4429 * outgoing packet. The bpf prog will first be called
4430 * to reserve space in a skb under
4431 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
4432 * the bpf prog will be called to write the header option(s)
4433 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4435 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
4436 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
4437 * related helpers that will be useful to the bpf programs.
4439 * The kernel gets its chance to reserve space and write
4440 * options first before the BPF program does.
4442 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
4443 /* Mask of all currently supported cb flags */
4444 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
4447 /* List of known BPF sock_ops operators.
4448 * New entries can only be added at the end
4452 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
4453 * -1 if default value should be used
4455 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
4456 * window (in packets) or -1 if default
4457 * value should be used
4459 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
4460 * active connection is initialized
4462 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
4463 * active connection is
4466 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
4467 * passive connection is
4470 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
4473 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
4474 * based on the path and may be
4475 * dependent on the congestion control
4476 * algorithm. In general it indicates
4477 * a congestion threshold. RTTs above
4478 * this indicate congestion
4480 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
4481 * Arg1: value of icsk_retransmits
4482 * Arg2: value of icsk_rto
4483 * Arg3: whether RTO has expired
4485 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
4486 * Arg1: sequence number of 1st byte
4488 * Arg3: return value of
4489 * tcp_transmit_skb (0 => success)
4491 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
4495 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
4496 * socket transition to LISTEN state.
4498 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
4500 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
4501 * It will be called to handle
4502 * the packets received at
4503 * an already established
4506 * sock_ops->skb_data:
4507 * Referring to the received skb.
4508 * It covers the TCP header only.
4510 * bpf_load_hdr_opt() can also
4511 * be used to search for a
4512 * particular option.
4514 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
4515 * header option later in
4516 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4517 * Arg1: bool want_cookie. (in
4518 * writing SYNACK only)
4520 * sock_ops->skb_data:
4521 * Not available because no header has
4524 * sock_ops->skb_tcp_flags:
4525 * The tcp_flags of the
4526 * outgoing skb. (e.g. SYN, ACK, FIN).
4528 * bpf_reserve_hdr_opt() should
4529 * be used to reserve space.
4531 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
4532 * Arg1: bool want_cookie. (in
4533 * writing SYNACK only)
4535 * sock_ops->skb_data:
4536 * Referring to the outgoing skb.
4537 * It covers the TCP header
4538 * that has already been written
4539 * by the kernel and the
4540 * earlier bpf-progs.
4542 * sock_ops->skb_tcp_flags:
4543 * The tcp_flags of the outgoing
4544 * skb. (e.g. SYN, ACK, FIN).
4546 * bpf_store_hdr_opt() should
4547 * be used to write the
4550 * bpf_load_hdr_opt() can also
4551 * be used to search for a
4552 * particular option that
4553 * has already been written
4554 * by the kernel or the
4555 * earlier bpf-progs.
4559 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
4560 * changes between the TCP and BPF versions. Ideally this should never happen.
4561 * If it does, we need to add code to convert them before calling
4562 * the BPF sock_ops function.
4565 BPF_TCP_ESTABLISHED = 1,
4575 BPF_TCP_CLOSING, /* Now a valid state */
4576 BPF_TCP_NEW_SYN_RECV,
4578 BPF_TCP_MAX_STATES /* Leave at the end! */
4582 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4583 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
4584 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
4585 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
4586 /* Copy the SYN pkt to optval
4588 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
4589 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
4590 * to only getting from the saved_syn. It can either get the
4593 * 1. the just-received SYN packet (only available when writing the
4594 * SYNACK). It will be useful when it is not necessary to
4595 * save the SYN packet for latter use. It is also the only way
4596 * to get the SYN during syncookie mode because the syn
4597 * packet cannot be saved during syncookie.
4601 * 2. the earlier saved syn which was done by
4602 * bpf_setsockopt(TCP_SAVE_SYN).
4604 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
4605 * SYN packet is obtained.
4607 * If the bpf-prog does not need the IP[46] header, the
4608 * bpf-prog can avoid parsing the IP header by using
4609 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
4610 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
4612 * >0: Total number of bytes copied
4613 * -ENOSPC: Not enough space in optval. Only optlen number of
4615 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
4616 * is not saved by setsockopt(TCP_SAVE_SYN).
4618 TCP_BPF_SYN = 1005, /* Copy the TCP header */
4619 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
4620 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
4624 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
4627 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
4628 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4631 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
4632 * total option spaces
4633 * required for an established
4634 * sk in order to calculate the
4635 * MSS. No skb is actually
4638 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
4639 * when sending a SYN.
4643 struct bpf_perf_event_value {
4650 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
4651 BPF_DEVCG_ACC_READ = (1ULL << 1),
4652 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
4656 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
4657 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
4660 struct bpf_cgroup_dev_ctx {
4661 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
4667 struct bpf_raw_tracepoint_args {
4671 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
4672 * OUTPUT: Do lookup from egress perspective; default is ingress
4675 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
4676 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
4680 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
4681 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
4682 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
4683 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
4684 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
4685 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
4686 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
4687 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
4688 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
4691 struct bpf_fib_lookup {
4692 /* input: network family for lookup (AF_INET, AF_INET6)
4693 * output: network family of egress nexthop
4697 /* set if lookup is to consider L4 data - e.g., FIB rules */
4702 /* total length of packet from network header - used for MTU check */
4705 /* input: L3 device index for lookup
4706 * output: device index from FIB lookup
4711 /* inputs to lookup */
4712 __u8 tos; /* AF_INET */
4713 __be32 flowinfo; /* AF_INET6, flow_label + priority */
4715 /* output: metric of fib result (IPv4/IPv6 only) */
4721 __u32 ipv6_src[4]; /* in6_addr; network order */
4724 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
4725 * network header. output: bpf_fib_lookup sets to gateway address
4726 * if FIB lookup returns gateway route
4730 __u32 ipv6_dst[4]; /* in6_addr; network order */
4734 __be16 h_vlan_proto;
4736 __u8 smac[6]; /* ETH_ALEN */
4737 __u8 dmac[6]; /* ETH_ALEN */
4740 enum bpf_task_fd_type {
4741 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
4742 BPF_FD_TYPE_TRACEPOINT, /* tp name */
4743 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
4744 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
4745 BPF_FD_TYPE_UPROBE, /* filename + offset */
4746 BPF_FD_TYPE_URETPROBE, /* filename + offset */
4750 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
4751 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
4752 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
4755 struct bpf_flow_keys {
4758 __u16 addr_proto; /* ETH_P_* of valid addrs */
4772 __u32 ipv6_src[4]; /* in6_addr; network order */
4773 __u32 ipv6_dst[4]; /* in6_addr; network order */
4780 struct bpf_func_info {
4785 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
4786 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
4788 struct bpf_line_info {
4790 __u32 file_name_off;
4795 struct bpf_spin_lock {
4800 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
4801 * Allows 1,2,4-byte read, but no write.
4803 __u32 file_pos; /* Sysctl file position to read from, write to.
4804 * Allows 1,2,4-byte read an 4-byte write.
4808 struct bpf_sockopt {
4809 __bpf_md_ptr(struct bpf_sock *, sk);
4810 __bpf_md_ptr(void *, optval);
4811 __bpf_md_ptr(void *, optval_end);
4819 struct bpf_pidns_info {
4824 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
4825 struct bpf_sk_lookup {
4826 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
4828 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
4829 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
4830 __u32 remote_ip4; /* Network byte order */
4831 __u32 remote_ip6[4]; /* Network byte order */
4832 __u32 remote_port; /* Network byte order */
4833 __u32 local_ip4; /* Network byte order */
4834 __u32 local_ip6[4]; /* Network byte order */
4835 __u32 local_port; /* Host byte order */
4838 #endif /* _UAPI__LINUX_BPF_H__ */