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 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
352 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
353 * insn[0].imm: map fd map fd
354 * insn[1].imm: 0 offset into value
357 * ldimm64 rewrite: address of map address of map[0]+offset
358 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
360 #define BPF_PSEUDO_MAP_FD 1
361 #define BPF_PSEUDO_MAP_VALUE 2
363 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
364 * offset to another bpf function
366 #define BPF_PSEUDO_CALL 1
368 /* flags for BPF_MAP_UPDATE_ELEM command */
370 BPF_ANY = 0, /* create new element or update existing */
371 BPF_NOEXIST = 1, /* create new element if it didn't exist */
372 BPF_EXIST = 2, /* update existing element */
373 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
376 /* flags for BPF_MAP_CREATE command */
378 BPF_F_NO_PREALLOC = (1U << 0),
379 /* Instead of having one common LRU list in the
380 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
381 * which can scale and perform better.
382 * Note, the LRU nodes (including free nodes) cannot be moved
383 * across different LRU lists.
385 BPF_F_NO_COMMON_LRU = (1U << 1),
386 /* Specify numa node during map creation */
387 BPF_F_NUMA_NODE = (1U << 2),
389 /* Flags for accessing BPF object from syscall side. */
390 BPF_F_RDONLY = (1U << 3),
391 BPF_F_WRONLY = (1U << 4),
393 /* Flag for stack_map, store build_id+offset instead of pointer */
394 BPF_F_STACK_BUILD_ID = (1U << 5),
396 /* Zero-initialize hash function seed. This should only be used for testing. */
397 BPF_F_ZERO_SEED = (1U << 6),
399 /* Flags for accessing BPF object from program side. */
400 BPF_F_RDONLY_PROG = (1U << 7),
401 BPF_F_WRONLY_PROG = (1U << 8),
403 /* Clone map from listener for newly accepted socket */
404 BPF_F_CLONE = (1U << 9),
406 /* Enable memory-mapping BPF map */
407 BPF_F_MMAPABLE = (1U << 10),
410 /* Flags for BPF_PROG_QUERY. */
412 /* Query effective (directly attached + inherited from ancestor cgroups)
413 * programs that will be executed for events within a cgroup.
414 * attach_flags with this flag are returned only for directly attached programs.
416 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
418 /* type for BPF_ENABLE_STATS */
419 enum bpf_stats_type {
420 /* enabled run_time_ns and run_cnt */
421 BPF_STATS_RUN_TIME = 0,
424 enum bpf_stack_build_id_status {
425 /* user space need an empty entry to identify end of a trace */
426 BPF_STACK_BUILD_ID_EMPTY = 0,
427 /* with valid build_id and offset */
428 BPF_STACK_BUILD_ID_VALID = 1,
429 /* couldn't get build_id, fallback to ip */
430 BPF_STACK_BUILD_ID_IP = 2,
433 #define BPF_BUILD_ID_SIZE 20
434 struct bpf_stack_build_id {
436 unsigned char build_id[BPF_BUILD_ID_SIZE];
443 #define BPF_OBJ_NAME_LEN 16U
446 struct { /* anonymous struct used by BPF_MAP_CREATE command */
447 __u32 map_type; /* one of enum bpf_map_type */
448 __u32 key_size; /* size of key in bytes */
449 __u32 value_size; /* size of value in bytes */
450 __u32 max_entries; /* max number of entries in a map */
451 __u32 map_flags; /* BPF_MAP_CREATE related
452 * flags defined above.
454 __u32 inner_map_fd; /* fd pointing to the inner map */
455 __u32 numa_node; /* numa node (effective only if
456 * BPF_F_NUMA_NODE is set).
458 char map_name[BPF_OBJ_NAME_LEN];
459 __u32 map_ifindex; /* ifindex of netdev to create on */
460 __u32 btf_fd; /* fd pointing to a BTF type data */
461 __u32 btf_key_type_id; /* BTF type_id of the key */
462 __u32 btf_value_type_id; /* BTF type_id of the value */
463 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
464 * struct stored as the
469 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
474 __aligned_u64 next_key;
479 struct { /* struct used by BPF_MAP_*_BATCH commands */
480 __aligned_u64 in_batch; /* start batch,
481 * NULL to start from beginning
483 __aligned_u64 out_batch; /* output: next start batch */
485 __aligned_u64 values;
486 __u32 count; /* input/output:
487 * input: # of key/value
489 * output: # of filled elements
496 struct { /* anonymous struct used by BPF_PROG_LOAD command */
497 __u32 prog_type; /* one of enum bpf_prog_type */
500 __aligned_u64 license;
501 __u32 log_level; /* verbosity level of verifier */
502 __u32 log_size; /* size of user buffer */
503 __aligned_u64 log_buf; /* user supplied buffer */
504 __u32 kern_version; /* not used */
506 char prog_name[BPF_OBJ_NAME_LEN];
507 __u32 prog_ifindex; /* ifindex of netdev to prep for */
508 /* For some prog types expected attach type must be known at
509 * load time to verify attach type specific parts of prog
510 * (context accesses, allowed helpers, etc).
512 __u32 expected_attach_type;
513 __u32 prog_btf_fd; /* fd pointing to BTF type data */
514 __u32 func_info_rec_size; /* userspace bpf_func_info size */
515 __aligned_u64 func_info; /* func info */
516 __u32 func_info_cnt; /* number of bpf_func_info records */
517 __u32 line_info_rec_size; /* userspace bpf_line_info size */
518 __aligned_u64 line_info; /* line info */
519 __u32 line_info_cnt; /* number of bpf_line_info records */
520 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
521 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
524 struct { /* anonymous struct used by BPF_OBJ_* commands */
525 __aligned_u64 pathname;
530 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
531 __u32 target_fd; /* container object to attach to */
532 __u32 attach_bpf_fd; /* eBPF program to attach */
535 __u32 replace_bpf_fd; /* previously attached eBPF
536 * program to replace if
537 * BPF_F_REPLACE is used
541 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
544 __u32 data_size_in; /* input: len of data_in */
545 __u32 data_size_out; /* input/output: len of data_out
546 * returns ENOSPC if data_out
549 __aligned_u64 data_in;
550 __aligned_u64 data_out;
553 __u32 ctx_size_in; /* input: len of ctx_in */
554 __u32 ctx_size_out; /* input/output: len of ctx_out
555 * returns ENOSPC if ctx_out
558 __aligned_u64 ctx_in;
559 __aligned_u64 ctx_out;
562 struct { /* anonymous struct used by BPF_*_GET_*_ID */
574 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
580 struct { /* anonymous struct used by BPF_PROG_QUERY command */
581 __u32 target_fd; /* container object to query */
585 __aligned_u64 prog_ids;
589 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
594 struct { /* anonymous struct for BPF_BTF_LOAD */
596 __aligned_u64 btf_log_buf;
603 __u32 pid; /* input: pid */
604 __u32 fd; /* input: fd */
605 __u32 flags; /* input: flags */
606 __u32 buf_len; /* input/output: buf len */
607 __aligned_u64 buf; /* input/output:
608 * tp_name for tracepoint
610 * filename for uprobe
612 __u32 prog_id; /* output: prod_id */
613 __u32 fd_type; /* output: BPF_FD_TYPE_* */
614 __u64 probe_offset; /* output: probe_offset */
615 __u64 probe_addr; /* output: probe_addr */
618 struct { /* struct used by BPF_LINK_CREATE command */
619 __u32 prog_fd; /* eBPF program to attach */
621 __u32 target_fd; /* object to attach to */
622 __u32 target_ifindex; /* target ifindex */
624 __u32 attach_type; /* attach type */
625 __u32 flags; /* extra flags */
626 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
627 __u32 iter_info_len; /* iter_info length */
630 struct { /* struct used by BPF_LINK_UPDATE command */
631 __u32 link_fd; /* link fd */
632 /* new program fd to update link with */
634 __u32 flags; /* extra flags */
635 /* expected link's program fd; is specified only if
636 * BPF_F_REPLACE flag is set in flags */
644 struct { /* struct used by BPF_ENABLE_STATS command */
648 struct { /* struct used by BPF_ITER_CREATE command */
653 } __attribute__((aligned(8)));
655 /* The description below is an attempt at providing documentation to eBPF
656 * developers about the multiple available eBPF helper functions. It can be
657 * parsed and used to produce a manual page. The workflow is the following,
658 * and requires the rst2man utility:
660 * $ ./scripts/bpf_helpers_doc.py \
661 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
662 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
663 * $ man /tmp/bpf-helpers.7
665 * Note that in order to produce this external documentation, some RST
666 * formatting is used in the descriptions to get "bold" and "italics" in
667 * manual pages. Also note that the few trailing white spaces are
668 * intentional, removing them would break paragraphs for rst2man.
670 * Start of BPF helper function descriptions:
672 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
674 * Perform a lookup in *map* for an entry associated to *key*.
676 * Map value associated to *key*, or **NULL** if no entry was
679 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
681 * Add or update the value of the entry associated to *key* in
682 * *map* with *value*. *flags* is one of:
685 * The entry for *key* must not exist in the map.
687 * The entry for *key* must already exist in the map.
689 * No condition on the existence of the entry for *key*.
691 * Flag value **BPF_NOEXIST** cannot be used for maps of types
692 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
693 * elements always exist), the helper would return an error.
695 * 0 on success, or a negative error in case of failure.
697 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
699 * Delete entry with *key* from *map*.
701 * 0 on success, or a negative error in case of failure.
703 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
705 * For tracing programs, safely attempt to read *size* bytes from
706 * kernel space address *unsafe_ptr* and store the data in *dst*.
708 * Generally, use **bpf_probe_read_user**\ () or
709 * **bpf_probe_read_kernel**\ () instead.
711 * 0 on success, or a negative error in case of failure.
713 * u64 bpf_ktime_get_ns(void)
715 * Return the time elapsed since system boot, in nanoseconds.
716 * Does not include time the system was suspended.
717 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
721 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
723 * This helper is a "printk()-like" facility for debugging. It
724 * prints a message defined by format *fmt* (of size *fmt_size*)
725 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
726 * available. It can take up to three additional **u64**
727 * arguments (as an eBPF helpers, the total number of arguments is
730 * Each time the helper is called, it appends a line to the trace.
731 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
732 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
733 * The format of the trace is customizable, and the exact output
734 * one will get depends on the options set in
735 * *\/sys/kernel/debug/tracing/trace_options* (see also the
736 * *README* file under the same directory). However, it usually
737 * defaults to something like:
741 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
745 * * ``telnet`` is the name of the current task.
746 * * ``470`` is the PID of the current task.
747 * * ``001`` is the CPU number on which the task is
749 * * In ``.N..``, each character refers to a set of
750 * options (whether irqs are enabled, scheduling
751 * options, whether hard/softirqs are running, level of
752 * preempt_disabled respectively). **N** means that
753 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
755 * * ``419421.045894`` is a timestamp.
756 * * ``0x00000001`` is a fake value used by BPF for the
757 * instruction pointer register.
758 * * ``<formatted msg>`` is the message formatted with
761 * The conversion specifiers supported by *fmt* are similar, but
762 * more limited than for printk(). They are **%d**, **%i**,
763 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
764 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
765 * of field, padding with zeroes, etc.) is available, and the
766 * helper will return **-EINVAL** (but print nothing) if it
767 * encounters an unknown specifier.
769 * Also, note that **bpf_trace_printk**\ () is slow, and should
770 * only be used for debugging purposes. For this reason, a notice
771 * bloc (spanning several lines) is printed to kernel logs and
772 * states that the helper should not be used "for production use"
773 * the first time this helper is used (or more precisely, when
774 * **trace_printk**\ () buffers are allocated). For passing values
775 * to user space, perf events should be preferred.
777 * The number of bytes written to the buffer, or a negative error
778 * in case of failure.
780 * u32 bpf_get_prandom_u32(void)
782 * Get a pseudo-random number.
784 * From a security point of view, this helper uses its own
785 * pseudo-random internal state, and cannot be used to infer the
786 * seed of other random functions in the kernel. However, it is
787 * essential to note that the generator used by the helper is not
788 * cryptographically secure.
790 * A random 32-bit unsigned value.
792 * u32 bpf_get_smp_processor_id(void)
794 * Get the SMP (symmetric multiprocessing) processor id. Note that
795 * all programs run with preemption disabled, which means that the
796 * SMP processor id is stable during all the execution of the
799 * The SMP id of the processor running the program.
801 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
803 * Store *len* bytes from address *from* into the packet
804 * associated to *skb*, at *offset*. *flags* are a combination of
805 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
806 * checksum for the packet after storing the bytes) and
807 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
808 * **->swhash** and *skb*\ **->l4hash** to 0).
810 * A call to this helper is susceptible to change the underlying
811 * packet buffer. Therefore, at load time, all checks on pointers
812 * previously done by the verifier are invalidated and must be
813 * performed again, if the helper is used in combination with
814 * direct packet access.
816 * 0 on success, or a negative error in case of failure.
818 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
820 * Recompute the layer 3 (e.g. IP) checksum for the packet
821 * associated to *skb*. Computation is incremental, so the helper
822 * must know the former value of the header field that was
823 * modified (*from*), the new value of this field (*to*), and the
824 * number of bytes (2 or 4) for this field, stored in *size*.
825 * Alternatively, it is possible to store the difference between
826 * the previous and the new values of the header field in *to*, by
827 * setting *from* and *size* to 0. For both methods, *offset*
828 * indicates the location of the IP checksum within the packet.
830 * This helper works in combination with **bpf_csum_diff**\ (),
831 * which does not update the checksum in-place, but offers more
832 * flexibility and can handle sizes larger than 2 or 4 for the
833 * checksum to update.
835 * A call to this helper is susceptible to change the underlying
836 * packet buffer. Therefore, at load time, all checks on pointers
837 * previously done by the verifier are invalidated and must be
838 * performed again, if the helper is used in combination with
839 * direct packet access.
841 * 0 on success, or a negative error in case of failure.
843 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
845 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
846 * packet associated to *skb*. Computation is incremental, so the
847 * helper must know the former value of the header field that was
848 * modified (*from*), the new value of this field (*to*), and the
849 * number of bytes (2 or 4) for this field, stored on the lowest
850 * four bits of *flags*. Alternatively, it is possible to store
851 * the difference between the previous and the new values of the
852 * header field in *to*, by setting *from* and the four lowest
853 * bits of *flags* to 0. For both methods, *offset* indicates the
854 * location of the IP checksum within the packet. In addition to
855 * the size of the field, *flags* can be added (bitwise OR) actual
856 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
857 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
858 * for updates resulting in a null checksum the value is set to
859 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
860 * the checksum is to be computed against a pseudo-header.
862 * This helper works in combination with **bpf_csum_diff**\ (),
863 * which does not update the checksum in-place, but offers more
864 * flexibility and can handle sizes larger than 2 or 4 for the
865 * checksum to update.
867 * A call to this helper is susceptible to change the underlying
868 * packet buffer. Therefore, at load time, all checks on pointers
869 * previously done by the verifier are invalidated and must be
870 * performed again, if the helper is used in combination with
871 * direct packet access.
873 * 0 on success, or a negative error in case of failure.
875 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
877 * This special helper is used to trigger a "tail call", or in
878 * other words, to jump into another eBPF program. The same stack
879 * frame is used (but values on stack and in registers for the
880 * caller are not accessible to the callee). This mechanism allows
881 * for program chaining, either for raising the maximum number of
882 * available eBPF instructions, or to execute given programs in
883 * conditional blocks. For security reasons, there is an upper
884 * limit to the number of successive tail calls that can be
887 * Upon call of this helper, the program attempts to jump into a
888 * program referenced at index *index* in *prog_array_map*, a
889 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
890 * *ctx*, a pointer to the context.
892 * If the call succeeds, the kernel immediately runs the first
893 * instruction of the new program. This is not a function call,
894 * and it never returns to the previous program. If the call
895 * fails, then the helper has no effect, and the caller continues
896 * to run its subsequent instructions. A call can fail if the
897 * destination program for the jump does not exist (i.e. *index*
898 * is superior to the number of entries in *prog_array_map*), or
899 * if the maximum number of tail calls has been reached for this
900 * chain of programs. This limit is defined in the kernel by the
901 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
902 * which is currently set to 32.
904 * 0 on success, or a negative error in case of failure.
906 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
908 * Clone and redirect the packet associated to *skb* to another
909 * net device of index *ifindex*. Both ingress and egress
910 * interfaces can be used for redirection. The **BPF_F_INGRESS**
911 * value in *flags* is used to make the distinction (ingress path
912 * is selected if the flag is present, egress path otherwise).
913 * This is the only flag supported for now.
915 * In comparison with **bpf_redirect**\ () helper,
916 * **bpf_clone_redirect**\ () has the associated cost of
917 * duplicating the packet buffer, but this can be executed out of
918 * the eBPF program. Conversely, **bpf_redirect**\ () is more
919 * efficient, but it is handled through an action code where the
920 * redirection happens only after the eBPF program has returned.
922 * A call to this helper is susceptible to change the underlying
923 * packet buffer. Therefore, at load time, all checks on pointers
924 * previously done by the verifier are invalidated and must be
925 * performed again, if the helper is used in combination with
926 * direct packet access.
928 * 0 on success, or a negative error in case of failure.
930 * u64 bpf_get_current_pid_tgid(void)
932 * A 64-bit integer containing the current tgid and pid, and
934 * *current_task*\ **->tgid << 32 \|**
935 * *current_task*\ **->pid**.
937 * u64 bpf_get_current_uid_gid(void)
939 * A 64-bit integer containing the current GID and UID, and
940 * created as such: *current_gid* **<< 32 \|** *current_uid*.
942 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
944 * Copy the **comm** attribute of the current task into *buf* of
945 * *size_of_buf*. The **comm** attribute contains the name of
946 * the executable (excluding the path) for the current task. The
947 * *size_of_buf* must be strictly positive. On success, the
948 * helper makes sure that the *buf* is NUL-terminated. On failure,
949 * it is filled with zeroes.
951 * 0 on success, or a negative error in case of failure.
953 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
955 * Retrieve the classid for the current task, i.e. for the net_cls
956 * cgroup to which *skb* belongs.
958 * This helper can be used on TC egress path, but not on ingress.
960 * The net_cls cgroup provides an interface to tag network packets
961 * based on a user-provided identifier for all traffic coming from
962 * the tasks belonging to the related cgroup. See also the related
963 * kernel documentation, available from the Linux sources in file
964 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
966 * The Linux kernel has two versions for cgroups: there are
967 * cgroups v1 and cgroups v2. Both are available to users, who can
968 * use a mixture of them, but note that the net_cls cgroup is for
969 * cgroup v1 only. This makes it incompatible with BPF programs
970 * run on cgroups, which is a cgroup-v2-only feature (a socket can
971 * only hold data for one version of cgroups at a time).
973 * This helper is only available is the kernel was compiled with
974 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
975 * "**y**" or to "**m**".
977 * The classid, or 0 for the default unconfigured classid.
979 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
981 * Push a *vlan_tci* (VLAN tag control information) of protocol
982 * *vlan_proto* to the packet associated to *skb*, then update
983 * the checksum. Note that if *vlan_proto* is different from
984 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
985 * be **ETH_P_8021Q**.
987 * A call to this helper is susceptible to change the underlying
988 * packet buffer. Therefore, at load time, all checks on pointers
989 * previously done by the verifier are invalidated and must be
990 * performed again, if the helper is used in combination with
991 * direct packet access.
993 * 0 on success, or a negative error in case of failure.
995 * long bpf_skb_vlan_pop(struct sk_buff *skb)
997 * Pop a VLAN header from the packet associated to *skb*.
999 * A call to this helper is susceptible to change the underlying
1000 * packet buffer. Therefore, at load time, all checks on pointers
1001 * previously done by the verifier are invalidated and must be
1002 * performed again, if the helper is used in combination with
1003 * direct packet access.
1005 * 0 on success, or a negative error in case of failure.
1007 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1009 * Get tunnel metadata. This helper takes a pointer *key* to an
1010 * empty **struct bpf_tunnel_key** of **size**, that will be
1011 * filled with tunnel metadata for the packet associated to *skb*.
1012 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1013 * indicates that the tunnel is based on IPv6 protocol instead of
1016 * The **struct bpf_tunnel_key** is an object that generalizes the
1017 * principal parameters used by various tunneling protocols into a
1018 * single struct. This way, it can be used to easily make a
1019 * decision based on the contents of the encapsulation header,
1020 * "summarized" in this struct. In particular, it holds the IP
1021 * address of the remote end (IPv4 or IPv6, depending on the case)
1022 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1023 * this struct exposes the *key*\ **->tunnel_id**, which is
1024 * generally mapped to a VNI (Virtual Network Identifier), making
1025 * it programmable together with the **bpf_skb_set_tunnel_key**\
1028 * Let's imagine that the following code is part of a program
1029 * attached to the TC ingress interface, on one end of a GRE
1030 * tunnel, and is supposed to filter out all messages coming from
1031 * remote ends with IPv4 address other than 10.0.0.1:
1036 * struct bpf_tunnel_key key = {};
1038 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1040 * return TC_ACT_SHOT; // drop packet
1042 * if (key.remote_ipv4 != 0x0a000001)
1043 * return TC_ACT_SHOT; // drop packet
1045 * return TC_ACT_OK; // accept packet
1047 * This interface can also be used with all encapsulation devices
1048 * that can operate in "collect metadata" mode: instead of having
1049 * one network device per specific configuration, the "collect
1050 * metadata" mode only requires a single device where the
1051 * configuration can be extracted from this helper.
1053 * This can be used together with various tunnels such as VXLan,
1054 * Geneve, GRE or IP in IP (IPIP).
1056 * 0 on success, or a negative error in case of failure.
1058 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1060 * Populate tunnel metadata for packet associated to *skb.* The
1061 * tunnel metadata is set to the contents of *key*, of *size*. The
1062 * *flags* can be set to a combination of the following values:
1064 * **BPF_F_TUNINFO_IPV6**
1065 * Indicate that the tunnel is based on IPv6 protocol
1067 * **BPF_F_ZERO_CSUM_TX**
1068 * For IPv4 packets, add a flag to tunnel metadata
1069 * indicating that checksum computation should be skipped
1070 * and checksum set to zeroes.
1071 * **BPF_F_DONT_FRAGMENT**
1072 * Add a flag to tunnel metadata indicating that the
1073 * packet should not be fragmented.
1074 * **BPF_F_SEQ_NUMBER**
1075 * Add a flag to tunnel metadata indicating that a
1076 * sequence number should be added to tunnel header before
1077 * sending the packet. This flag was added for GRE
1078 * encapsulation, but might be used with other protocols
1079 * as well in the future.
1081 * Here is a typical usage on the transmit path:
1085 * struct bpf_tunnel_key key;
1087 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1088 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1090 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1091 * helper for additional information.
1093 * 0 on success, or a negative error in case of failure.
1095 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1097 * Read the value of a perf event counter. This helper relies on a
1098 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1099 * the perf event counter is selected when *map* is updated with
1100 * perf event file descriptors. The *map* is an array whose size
1101 * is the number of available CPUs, and each cell contains a value
1102 * relative to one CPU. The value to retrieve is indicated by
1103 * *flags*, that contains the index of the CPU to look up, masked
1104 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1105 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1106 * current CPU should be retrieved.
1108 * Note that before Linux 4.13, only hardware perf event can be
1111 * Also, be aware that the newer helper
1112 * **bpf_perf_event_read_value**\ () is recommended over
1113 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1114 * quirks where error and counter value are used as a return code
1115 * (which is wrong to do since ranges may overlap). This issue is
1116 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1117 * time provides more features over the **bpf_perf_event_read**\
1118 * () interface. Please refer to the description of
1119 * **bpf_perf_event_read_value**\ () for details.
1121 * The value of the perf event counter read from the map, or a
1122 * negative error code in case of failure.
1124 * long bpf_redirect(u32 ifindex, u64 flags)
1126 * Redirect the packet to another net device of index *ifindex*.
1127 * This helper is somewhat similar to **bpf_clone_redirect**\
1128 * (), except that the packet is not cloned, which provides
1129 * increased performance.
1131 * Except for XDP, both ingress and egress interfaces can be used
1132 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1133 * to make the distinction (ingress path is selected if the flag
1134 * is present, egress path otherwise). Currently, XDP only
1135 * supports redirection to the egress interface, and accepts no
1138 * The same effect can also be attained with the more generic
1139 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1140 * redirect target instead of providing it directly to the helper.
1142 * For XDP, the helper returns **XDP_REDIRECT** on success or
1143 * **XDP_ABORTED** on error. For other program types, the values
1144 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1147 * u32 bpf_get_route_realm(struct sk_buff *skb)
1149 * Retrieve the realm or the route, that is to say the
1150 * **tclassid** field of the destination for the *skb*. The
1151 * indentifier retrieved is a user-provided tag, similar to the
1152 * one used with the net_cls cgroup (see description for
1153 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1154 * held by a route (a destination entry), not by a task.
1156 * Retrieving this identifier works with the clsact TC egress hook
1157 * (see also **tc-bpf(8)**), or alternatively on conventional
1158 * classful egress qdiscs, but not on TC ingress path. In case of
1159 * clsact TC egress hook, this has the advantage that, internally,
1160 * the destination entry has not been dropped yet in the transmit
1161 * path. Therefore, the destination entry does not need to be
1162 * artificially held via **netif_keep_dst**\ () for a classful
1163 * qdisc until the *skb* is freed.
1165 * This helper is available only if the kernel was compiled with
1166 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1168 * The realm of the route for the packet associated to *skb*, or 0
1169 * if none was found.
1171 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1173 * Write raw *data* blob into a special BPF perf event held by
1174 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1175 * event must have the following attributes: **PERF_SAMPLE_RAW**
1176 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1177 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1179 * The *flags* are used to indicate the index in *map* for which
1180 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1181 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1182 * to indicate that the index of the current CPU core should be
1185 * The value to write, of *size*, is passed through eBPF stack and
1186 * pointed by *data*.
1188 * The context of the program *ctx* needs also be passed to the
1191 * On user space, a program willing to read the values needs to
1192 * call **perf_event_open**\ () on the perf event (either for
1193 * one or for all CPUs) and to store the file descriptor into the
1194 * *map*. This must be done before the eBPF program can send data
1195 * into it. An example is available in file
1196 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1197 * tree (the eBPF program counterpart is in
1198 * *samples/bpf/trace_output_kern.c*).
1200 * **bpf_perf_event_output**\ () achieves better performance
1201 * than **bpf_trace_printk**\ () for sharing data with user
1202 * space, and is much better suitable for streaming data from eBPF
1205 * Note that this helper is not restricted to tracing use cases
1206 * and can be used with programs attached to TC or XDP as well,
1207 * where it allows for passing data to user space listeners. Data
1210 * * Only custom structs,
1211 * * Only the packet payload, or
1212 * * A combination of both.
1214 * 0 on success, or a negative error in case of failure.
1216 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1218 * This helper was provided as an easy way to load data from a
1219 * packet. It can be used to load *len* bytes from *offset* from
1220 * the packet associated to *skb*, into the buffer pointed by
1223 * Since Linux 4.7, usage of this helper has mostly been replaced
1224 * by "direct packet access", enabling packet data to be
1225 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1226 * pointing respectively to the first byte of packet data and to
1227 * the byte after the last byte of packet data. However, it
1228 * remains useful if one wishes to read large quantities of data
1229 * at once from a packet into the eBPF stack.
1231 * 0 on success, or a negative error in case of failure.
1233 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1235 * Walk a user or a kernel stack and return its id. To achieve
1236 * this, the helper needs *ctx*, which is a pointer to the context
1237 * on which the tracing program is executed, and a pointer to a
1238 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1240 * The last argument, *flags*, holds the number of stack frames to
1241 * skip (from 0 to 255), masked with
1242 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1243 * a combination of the following flags:
1245 * **BPF_F_USER_STACK**
1246 * Collect a user space stack instead of a kernel stack.
1247 * **BPF_F_FAST_STACK_CMP**
1248 * Compare stacks by hash only.
1249 * **BPF_F_REUSE_STACKID**
1250 * If two different stacks hash into the same *stackid*,
1251 * discard the old one.
1253 * The stack id retrieved is a 32 bit long integer handle which
1254 * can be further combined with other data (including other stack
1255 * ids) and used as a key into maps. This can be useful for
1256 * generating a variety of graphs (such as flame graphs or off-cpu
1259 * For walking a stack, this helper is an improvement over
1260 * **bpf_probe_read**\ (), which can be used with unrolled loops
1261 * but is not efficient and consumes a lot of eBPF instructions.
1262 * Instead, **bpf_get_stackid**\ () can collect up to
1263 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1264 * this limit can be controlled with the **sysctl** program, and
1265 * that it should be manually increased in order to profile long
1266 * user stacks (such as stacks for Java programs). To do so, use:
1270 * # sysctl kernel.perf_event_max_stack=<new value>
1272 * The positive or null stack id on success, or a negative error
1273 * in case of failure.
1275 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1277 * Compute a checksum difference, from the raw buffer pointed by
1278 * *from*, of length *from_size* (that must be a multiple of 4),
1279 * towards the raw buffer pointed by *to*, of size *to_size*
1280 * (same remark). An optional *seed* can be added to the value
1281 * (this can be cascaded, the seed may come from a previous call
1284 * This is flexible enough to be used in several ways:
1286 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1287 * checksum, it can be used when pushing new data.
1288 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1289 * checksum, it can be used when removing data from a packet.
1290 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1291 * can be used to compute a diff. Note that *from_size* and
1292 * *to_size* do not need to be equal.
1294 * This helper can be used in combination with
1295 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1296 * which one can feed in the difference computed with
1297 * **bpf_csum_diff**\ ().
1299 * The checksum result, or a negative error code in case of
1302 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1304 * Retrieve tunnel options metadata for the packet associated to
1305 * *skb*, and store the raw tunnel option data to the buffer *opt*
1308 * This helper can be used with encapsulation devices that can
1309 * operate in "collect metadata" mode (please refer to the related
1310 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1311 * more details). A particular example where this can be used is
1312 * in combination with the Geneve encapsulation protocol, where it
1313 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1314 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1315 * the eBPF program. This allows for full customization of these
1318 * The size of the option data retrieved.
1320 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1322 * Set tunnel options metadata for the packet associated to *skb*
1323 * to the option data contained in the raw buffer *opt* of *size*.
1325 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1326 * helper for additional information.
1328 * 0 on success, or a negative error in case of failure.
1330 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1332 * Change the protocol of the *skb* to *proto*. Currently
1333 * supported are transition from IPv4 to IPv6, and from IPv6 to
1334 * IPv4. The helper takes care of the groundwork for the
1335 * transition, including resizing the socket buffer. The eBPF
1336 * program is expected to fill the new headers, if any, via
1337 * **skb_store_bytes**\ () and to recompute the checksums with
1338 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1339 * (). The main case for this helper is to perform NAT64
1340 * operations out of an eBPF program.
1342 * Internally, the GSO type is marked as dodgy so that headers are
1343 * checked and segments are recalculated by the GSO/GRO engine.
1344 * The size for GSO target is adapted as well.
1346 * All values for *flags* are reserved for future usage, and must
1349 * A call to this helper is susceptible to change the underlying
1350 * packet buffer. Therefore, at load time, all checks on pointers
1351 * previously done by the verifier are invalidated and must be
1352 * performed again, if the helper is used in combination with
1353 * direct packet access.
1355 * 0 on success, or a negative error in case of failure.
1357 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
1359 * Change the packet type for the packet associated to *skb*. This
1360 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1361 * the eBPF program does not have a write access to *skb*\
1362 * **->pkt_type** beside this helper. Using a helper here allows
1363 * for graceful handling of errors.
1365 * The major use case is to change incoming *skb*s to
1366 * **PACKET_HOST** in a programmatic way instead of having to
1367 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1370 * Note that *type* only allows certain values. At this time, they
1375 * **PACKET_BROADCAST**
1376 * Send packet to all.
1377 * **PACKET_MULTICAST**
1378 * Send packet to group.
1379 * **PACKET_OTHERHOST**
1380 * Send packet to someone else.
1382 * 0 on success, or a negative error in case of failure.
1384 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1386 * Check whether *skb* is a descendant of the cgroup2 held by
1387 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1389 * The return value depends on the result of the test, and can be:
1391 * * 0, if the *skb* failed the cgroup2 descendant test.
1392 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1393 * * A negative error code, if an error occurred.
1395 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1397 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1398 * not set, in particular if the hash was cleared due to mangling,
1399 * recompute this hash. Later accesses to the hash can be done
1400 * directly with *skb*\ **->hash**.
1402 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1403 * prototype with **bpf_skb_change_proto**\ (), or calling
1404 * **bpf_skb_store_bytes**\ () with the
1405 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1406 * the hash and to trigger a new computation for the next call to
1407 * **bpf_get_hash_recalc**\ ().
1411 * u64 bpf_get_current_task(void)
1413 * A pointer to the current task struct.
1415 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
1417 * Attempt in a safe way to write *len* bytes from the buffer
1418 * *src* to *dst* in memory. It only works for threads that are in
1419 * user context, and *dst* must be a valid user space address.
1421 * This helper should not be used to implement any kind of
1422 * security mechanism because of TOC-TOU attacks, but rather to
1423 * debug, divert, and manipulate execution of semi-cooperative
1426 * Keep in mind that this feature is meant for experiments, and it
1427 * has a risk of crashing the system and running programs.
1428 * Therefore, when an eBPF program using this helper is attached,
1429 * a warning including PID and process name is printed to kernel
1432 * 0 on success, or a negative error in case of failure.
1434 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1436 * Check whether the probe is being run is the context of a given
1437 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1438 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1440 * The return value depends on the result of the test, and can be:
1442 * * 0, if the *skb* task belongs to the cgroup2.
1443 * * 1, if the *skb* task does not belong to the cgroup2.
1444 * * A negative error code, if an error occurred.
1446 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1448 * Resize (trim or grow) the packet associated to *skb* to the
1449 * new *len*. The *flags* are reserved for future usage, and must
1452 * The basic idea is that the helper performs the needed work to
1453 * change the size of the packet, then the eBPF program rewrites
1454 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1455 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1456 * and others. This helper is a slow path utility intended for
1457 * replies with control messages. And because it is targeted for
1458 * slow path, the helper itself can afford to be slow: it
1459 * implicitly linearizes, unclones and drops offloads from the
1462 * A call to this helper is susceptible to change the underlying
1463 * packet buffer. Therefore, at load time, all checks on pointers
1464 * previously done by the verifier are invalidated and must be
1465 * performed again, if the helper is used in combination with
1466 * direct packet access.
1468 * 0 on success, or a negative error in case of failure.
1470 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1472 * Pull in non-linear data in case the *skb* is non-linear and not
1473 * all of *len* are part of the linear section. Make *len* bytes
1474 * from *skb* readable and writable. If a zero value is passed for
1475 * *len*, then the whole length of the *skb* is pulled.
1477 * This helper is only needed for reading and writing with direct
1480 * For direct packet access, testing that offsets to access
1481 * are within packet boundaries (test on *skb*\ **->data_end**) is
1482 * susceptible to fail if offsets are invalid, or if the requested
1483 * data is in non-linear parts of the *skb*. On failure the
1484 * program can just bail out, or in the case of a non-linear
1485 * buffer, use a helper to make the data available. The
1486 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1487 * the data. Another one consists in using **bpf_skb_pull_data**
1488 * to pull in once the non-linear parts, then retesting and
1489 * eventually access the data.
1491 * At the same time, this also makes sure the *skb* is uncloned,
1492 * which is a necessary condition for direct write. As this needs
1493 * to be an invariant for the write part only, the verifier
1494 * detects writes and adds a prologue that is calling
1495 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1496 * the very beginning in case it is indeed cloned.
1498 * A call to this helper is susceptible to change the underlying
1499 * packet buffer. Therefore, at load time, all checks on pointers
1500 * previously done by the verifier are invalidated and must be
1501 * performed again, if the helper is used in combination with
1502 * direct packet access.
1504 * 0 on success, or a negative error in case of failure.
1506 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1508 * Add the checksum *csum* into *skb*\ **->csum** in case the
1509 * driver has supplied a checksum for the entire packet into that
1510 * field. Return an error otherwise. This helper is intended to be
1511 * used in combination with **bpf_csum_diff**\ (), in particular
1512 * when the checksum needs to be updated after data has been
1513 * written into the packet through direct packet access.
1515 * The checksum on success, or a negative error code in case of
1518 * void bpf_set_hash_invalid(struct sk_buff *skb)
1520 * Invalidate the current *skb*\ **->hash**. It can be used after
1521 * mangling on headers through direct packet access, in order to
1522 * indicate that the hash is outdated and to trigger a
1523 * recalculation the next time the kernel tries to access this
1524 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1526 * long bpf_get_numa_node_id(void)
1528 * Return the id of the current NUMA node. The primary use case
1529 * for this helper is the selection of sockets for the local NUMA
1530 * node, when the program is attached to sockets using the
1531 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1532 * but the helper is also available to other eBPF program types,
1533 * similarly to **bpf_get_smp_processor_id**\ ().
1535 * The id of current NUMA node.
1537 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1539 * Grows headroom of packet associated to *skb* and adjusts the
1540 * offset of the MAC header accordingly, adding *len* bytes of
1541 * space. It automatically extends and reallocates memory as
1544 * This helper can be used on a layer 3 *skb* to push a MAC header
1545 * for redirection into a layer 2 device.
1547 * All values for *flags* are reserved for future usage, and must
1550 * A call to this helper is susceptible to change the underlying
1551 * packet buffer. Therefore, at load time, all checks on pointers
1552 * previously done by the verifier are invalidated and must be
1553 * performed again, if the helper is used in combination with
1554 * direct packet access.
1556 * 0 on success, or a negative error in case of failure.
1558 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1560 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1561 * it is possible to use a negative value for *delta*. This helper
1562 * can be used to prepare the packet for pushing or popping
1565 * A call to this helper is susceptible to change the underlying
1566 * packet buffer. Therefore, at load time, all checks on pointers
1567 * previously done by the verifier are invalidated and must be
1568 * performed again, if the helper is used in combination with
1569 * direct packet access.
1571 * 0 on success, or a negative error in case of failure.
1573 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1575 * Copy a NUL terminated string from an unsafe kernel address
1576 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
1579 * Generally, use **bpf_probe_read_user_str**\ () or
1580 * **bpf_probe_read_kernel_str**\ () instead.
1582 * On success, the strictly positive length of the string,
1583 * including the trailing NUL character. On error, a negative
1586 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1588 * If the **struct sk_buff** pointed by *skb* has a known socket,
1589 * retrieve the cookie (generated by the kernel) of this socket.
1590 * If no cookie has been set yet, generate a new cookie. Once
1591 * generated, the socket cookie remains stable for the life of the
1592 * socket. This helper can be useful for monitoring per socket
1593 * networking traffic statistics as it provides a global socket
1594 * identifier that can be assumed unique.
1596 * A 8-byte long non-decreasing number on success, or 0 if the
1597 * socket field is missing inside *skb*.
1599 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1601 * Equivalent to bpf_get_socket_cookie() helper that accepts
1602 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1604 * A 8-byte long non-decreasing number.
1606 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1608 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
1609 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1611 * A 8-byte long non-decreasing number.
1613 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1615 * The owner UID of the socket associated to *skb*. If the socket
1616 * is **NULL**, or if it is not a full socket (i.e. if it is a
1617 * time-wait or a request socket instead), **overflowuid** value
1618 * is returned (note that **overflowuid** might also be the actual
1619 * UID value for the socket).
1621 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
1623 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1628 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1630 * Emulate a call to **setsockopt()** on the socket associated to
1631 * *bpf_socket*, which must be a full socket. The *level* at
1632 * which the option resides and the name *optname* of the option
1633 * must be specified, see **setsockopt(2)** for more information.
1634 * The option value of length *optlen* is pointed by *optval*.
1636 * *bpf_socket* should be one of the following:
1638 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1639 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1640 * and **BPF_CGROUP_INET6_CONNECT**.
1642 * This helper actually implements a subset of **setsockopt()**.
1643 * It supports the following *level*\ s:
1645 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1646 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1647 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
1648 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
1649 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1650 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1651 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
1652 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
1653 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**.
1654 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1655 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1657 * 0 on success, or a negative error in case of failure.
1659 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1661 * Grow or shrink the room for data in the packet associated to
1662 * *skb* by *len_diff*, and according to the selected *mode*.
1664 * By default, the helper will reset any offloaded checksum
1665 * indicator of the skb to CHECKSUM_NONE. This can be avoided
1666 * by the following flag:
1668 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
1669 * checksum data of the skb to CHECKSUM_NONE.
1671 * There are two supported modes at this time:
1673 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1674 * (room space is added or removed below the layer 2 header).
1676 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1677 * (room space is added or removed below the layer 3 header).
1679 * The following flags are supported at this time:
1681 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1682 * Adjusting mss in this way is not allowed for datagrams.
1684 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1685 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1686 * Any new space is reserved to hold a tunnel header.
1687 * Configure skb offsets and other fields accordingly.
1689 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1690 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1691 * Use with ENCAP_L3 flags to further specify the tunnel type.
1693 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1694 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1695 * type; *len* is the length of the inner MAC header.
1697 * A call to this helper is susceptible to change the underlying
1698 * packet buffer. Therefore, at load time, all checks on pointers
1699 * previously done by the verifier are invalidated and must be
1700 * performed again, if the helper is used in combination with
1701 * direct packet access.
1703 * 0 on success, or a negative error in case of failure.
1705 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1707 * Redirect the packet to the endpoint referenced by *map* at
1708 * index *key*. Depending on its type, this *map* can contain
1709 * references to net devices (for forwarding packets through other
1710 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1711 * but this is only implemented for native XDP (with driver
1712 * support) as of this writing).
1714 * The lower two bits of *flags* are used as the return code if
1715 * the map lookup fails. This is so that the return value can be
1716 * one of the XDP program return codes up to **XDP_TX**, as chosen
1717 * by the caller. Any higher bits in the *flags* argument must be
1720 * See also **bpf_redirect**\ (), which only supports redirecting
1721 * to an ifindex, but doesn't require a map to do so.
1723 * **XDP_REDIRECT** on success, or the value of the two lower bits
1724 * of the *flags* argument on error.
1726 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1728 * Redirect the packet to the socket referenced by *map* (of type
1729 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1730 * egress interfaces can be used for redirection. The
1731 * **BPF_F_INGRESS** value in *flags* is used to make the
1732 * distinction (ingress path is selected if the flag is present,
1733 * egress path otherwise). This is the only flag supported for now.
1735 * **SK_PASS** on success, or **SK_DROP** on error.
1737 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1739 * Add an entry to, or update a *map* referencing sockets. The
1740 * *skops* is used as a new value for the entry associated to
1741 * *key*. *flags* is one of:
1744 * The entry for *key* must not exist in the map.
1746 * The entry for *key* must already exist in the map.
1748 * No condition on the existence of the entry for *key*.
1750 * If the *map* has eBPF programs (parser and verdict), those will
1751 * be inherited by the socket being added. If the socket is
1752 * already attached to eBPF programs, this results in an error.
1754 * 0 on success, or a negative error in case of failure.
1756 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1758 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1759 * *delta* (which can be positive or negative). Note that this
1760 * operation modifies the address stored in *xdp_md*\ **->data**,
1761 * so the latter must be loaded only after the helper has been
1764 * The use of *xdp_md*\ **->data_meta** is optional and programs
1765 * are not required to use it. The rationale is that when the
1766 * packet is processed with XDP (e.g. as DoS filter), it is
1767 * possible to push further meta data along with it before passing
1768 * to the stack, and to give the guarantee that an ingress eBPF
1769 * program attached as a TC classifier on the same device can pick
1770 * this up for further post-processing. Since TC works with socket
1771 * buffers, it remains possible to set from XDP the **mark** or
1772 * **priority** pointers, or other pointers for the socket buffer.
1773 * Having this scratch space generic and programmable allows for
1774 * more flexibility as the user is free to store whatever meta
1777 * A call to this helper is susceptible to change the underlying
1778 * packet buffer. Therefore, at load time, all checks on pointers
1779 * previously done by the verifier are invalidated and must be
1780 * performed again, if the helper is used in combination with
1781 * direct packet access.
1783 * 0 on success, or a negative error in case of failure.
1785 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1787 * Read the value of a perf event counter, and store it into *buf*
1788 * of size *buf_size*. This helper relies on a *map* of type
1789 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1790 * counter is selected when *map* is updated with perf event file
1791 * descriptors. The *map* is an array whose size is the number of
1792 * available CPUs, and each cell contains a value relative to one
1793 * CPU. The value to retrieve is indicated by *flags*, that
1794 * contains the index of the CPU to look up, masked with
1795 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1796 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1797 * current CPU should be retrieved.
1799 * This helper behaves in a way close to
1800 * **bpf_perf_event_read**\ () helper, save that instead of
1801 * just returning the value observed, it fills the *buf*
1802 * structure. This allows for additional data to be retrieved: in
1803 * particular, the enabled and running times (in *buf*\
1804 * **->enabled** and *buf*\ **->running**, respectively) are
1805 * copied. In general, **bpf_perf_event_read_value**\ () is
1806 * recommended over **bpf_perf_event_read**\ (), which has some
1807 * ABI issues and provides fewer functionalities.
1809 * These values are interesting, because hardware PMU (Performance
1810 * Monitoring Unit) counters are limited resources. When there are
1811 * more PMU based perf events opened than available counters,
1812 * kernel will multiplex these events so each event gets certain
1813 * percentage (but not all) of the PMU time. In case that
1814 * multiplexing happens, the number of samples or counter value
1815 * will not reflect the case compared to when no multiplexing
1816 * occurs. This makes comparison between different runs difficult.
1817 * Typically, the counter value should be normalized before
1818 * comparing to other experiments. The usual normalization is done
1823 * normalized_counter = counter * t_enabled / t_running
1825 * Where t_enabled is the time enabled for event and t_running is
1826 * the time running for event since last normalization. The
1827 * enabled and running times are accumulated since the perf event
1828 * open. To achieve scaling factor between two invocations of an
1829 * eBPF program, users can use CPU id as the key (which is
1830 * typical for perf array usage model) to remember the previous
1831 * value and do the calculation inside the eBPF program.
1833 * 0 on success, or a negative error in case of failure.
1835 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1837 * For en eBPF program attached to a perf event, retrieve the
1838 * value of the event counter associated to *ctx* and store it in
1839 * the structure pointed by *buf* and of size *buf_size*. Enabled
1840 * and running times are also stored in the structure (see
1841 * description of helper **bpf_perf_event_read_value**\ () for
1844 * 0 on success, or a negative error in case of failure.
1846 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1848 * Emulate a call to **getsockopt()** on the socket associated to
1849 * *bpf_socket*, which must be a full socket. The *level* at
1850 * which the option resides and the name *optname* of the option
1851 * must be specified, see **getsockopt(2)** for more information.
1852 * The retrieved value is stored in the structure pointed by
1853 * *opval* and of length *optlen*.
1855 * *bpf_socket* should be one of the following:
1857 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1858 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1859 * and **BPF_CGROUP_INET6_CONNECT**.
1861 * This helper actually implements a subset of **getsockopt()**.
1862 * It supports the following *level*\ s:
1864 * * **IPPROTO_TCP**, which supports *optname*
1865 * **TCP_CONGESTION**.
1866 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1867 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1869 * 0 on success, or a negative error in case of failure.
1871 * long bpf_override_return(struct pt_regs *regs, u64 rc)
1873 * Used for error injection, this helper uses kprobes to override
1874 * the return value of the probed function, and to set it to *rc*.
1875 * The first argument is the context *regs* on which the kprobe
1878 * This helper works by setting the PC (program counter)
1879 * to an override function which is run in place of the original
1880 * probed function. This means the probed function is not run at
1881 * all. The replacement function just returns with the required
1884 * This helper has security implications, and thus is subject to
1885 * restrictions. It is only available if the kernel was compiled
1886 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1887 * option, and in this case it only works on functions tagged with
1888 * **ALLOW_ERROR_INJECTION** in the kernel code.
1890 * Also, the helper is only available for the architectures having
1891 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1892 * x86 architecture is the only one to support this feature.
1896 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1898 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1899 * for the full TCP socket associated to *bpf_sock_ops* to
1902 * The primary use of this field is to determine if there should
1903 * be calls to eBPF programs of type
1904 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1905 * code. A program of the same type can change its value, per
1906 * connection and as necessary, when the connection is
1907 * established. This field is directly accessible for reading, but
1908 * this helper must be used for updates in order to return an
1909 * error if an eBPF program tries to set a callback that is not
1910 * supported in the current kernel.
1912 * *argval* is a flag array which can combine these flags:
1914 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1915 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1916 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1917 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1919 * Therefore, this function can be used to clear a callback flag by
1920 * setting the appropriate bit to zero. e.g. to disable the RTO
1923 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1924 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1926 * Here are some examples of where one could call such eBPF
1930 * * When a packet is retransmitted.
1931 * * When the connection terminates.
1932 * * When a packet is sent.
1933 * * When a packet is received.
1935 * Code **-EINVAL** if the socket is not a full TCP socket;
1936 * otherwise, a positive number containing the bits that could not
1937 * be set is returned (which comes down to 0 if all bits were set
1940 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1942 * This helper is used in programs implementing policies at the
1943 * socket level. If the message *msg* is allowed to pass (i.e. if
1944 * the verdict eBPF program returns **SK_PASS**), redirect it to
1945 * the socket referenced by *map* (of type
1946 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1947 * egress interfaces can be used for redirection. The
1948 * **BPF_F_INGRESS** value in *flags* is used to make the
1949 * distinction (ingress path is selected if the flag is present,
1950 * egress path otherwise). This is the only flag supported for now.
1952 * **SK_PASS** on success, or **SK_DROP** on error.
1954 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1956 * For socket policies, apply the verdict of the eBPF program to
1957 * the next *bytes* (number of bytes) of message *msg*.
1959 * For example, this helper can be used in the following cases:
1961 * * A single **sendmsg**\ () or **sendfile**\ () system call
1962 * contains multiple logical messages that the eBPF program is
1963 * supposed to read and for which it should apply a verdict.
1964 * * An eBPF program only cares to read the first *bytes* of a
1965 * *msg*. If the message has a large payload, then setting up
1966 * and calling the eBPF program repeatedly for all bytes, even
1967 * though the verdict is already known, would create unnecessary
1970 * When called from within an eBPF program, the helper sets a
1971 * counter internal to the BPF infrastructure, that is used to
1972 * apply the last verdict to the next *bytes*. If *bytes* is
1973 * smaller than the current data being processed from a
1974 * **sendmsg**\ () or **sendfile**\ () system call, the first
1975 * *bytes* will be sent and the eBPF program will be re-run with
1976 * the pointer for start of data pointing to byte number *bytes*
1977 * **+ 1**. If *bytes* is larger than the current data being
1978 * processed, then the eBPF verdict will be applied to multiple
1979 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1982 * Note that if a socket closes with the internal counter holding
1983 * a non-zero value, this is not a problem because data is not
1984 * being buffered for *bytes* and is sent as it is received.
1988 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1990 * For socket policies, prevent the execution of the verdict eBPF
1991 * program for message *msg* until *bytes* (byte number) have been
1994 * This can be used when one needs a specific number of bytes
1995 * before a verdict can be assigned, even if the data spans
1996 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1997 * case would be a user calling **sendmsg**\ () repeatedly with
1998 * 1-byte long message segments. Obviously, this is bad for
1999 * performance, but it is still valid. If the eBPF program needs
2000 * *bytes* bytes to validate a header, this helper can be used to
2001 * prevent the eBPF program to be called again until *bytes* have
2006 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2008 * For socket policies, pull in non-linear data from user space
2009 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2010 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2013 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2014 * *msg* it can only parse data that the (**data**, **data_end**)
2015 * pointers have already consumed. For **sendmsg**\ () hooks this
2016 * is likely the first scatterlist element. But for calls relying
2017 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2018 * be the range (**0**, **0**) because the data is shared with
2019 * user space and by default the objective is to avoid allowing
2020 * user space to modify data while (or after) eBPF verdict is
2021 * being decided. This helper can be used to pull in data and to
2022 * set the start and end pointer to given values. Data will be
2023 * copied if necessary (i.e. if data was not linear and if start
2024 * and end pointers do not point to the same chunk).
2026 * A call to this helper is susceptible to change the underlying
2027 * packet buffer. Therefore, at load time, all checks on pointers
2028 * previously done by the verifier are invalidated and must be
2029 * performed again, if the helper is used in combination with
2030 * direct packet access.
2032 * All values for *flags* are reserved for future usage, and must
2035 * 0 on success, or a negative error in case of failure.
2037 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2039 * Bind the socket associated to *ctx* to the address pointed by
2040 * *addr*, of length *addr_len*. This allows for making outgoing
2041 * connection from the desired IP address, which can be useful for
2042 * example when all processes inside a cgroup should use one
2043 * single IP address on a host that has multiple IP configured.
2045 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2046 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2047 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2048 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2049 * behavior and lets the kernel efficiently pick up an unused
2050 * port as long as 4-tuple is unique. Passing non-zero port might
2051 * lead to degraded performance.
2053 * 0 on success, or a negative error in case of failure.
2055 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2057 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2058 * possible to both shrink and grow the packet tail.
2059 * Shrink done via *delta* being a negative integer.
2061 * A call to this helper is susceptible to change the underlying
2062 * packet buffer. Therefore, at load time, all checks on pointers
2063 * previously done by the verifier are invalidated and must be
2064 * performed again, if the helper is used in combination with
2065 * direct packet access.
2067 * 0 on success, or a negative error in case of failure.
2069 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2071 * Retrieve the XFRM state (IP transform framework, see also
2072 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2074 * The retrieved value is stored in the **struct bpf_xfrm_state**
2075 * pointed by *xfrm_state* and of length *size*.
2077 * All values for *flags* are reserved for future usage, and must
2080 * This helper is available only if the kernel was compiled with
2081 * **CONFIG_XFRM** configuration option.
2083 * 0 on success, or a negative error in case of failure.
2085 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2087 * Return a user or a kernel stack in bpf program provided buffer.
2088 * To achieve this, the helper needs *ctx*, which is a pointer
2089 * to the context on which the tracing program is executed.
2090 * To store the stacktrace, the bpf program provides *buf* with
2091 * a nonnegative *size*.
2093 * The last argument, *flags*, holds the number of stack frames to
2094 * skip (from 0 to 255), masked with
2095 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2096 * the following flags:
2098 * **BPF_F_USER_STACK**
2099 * Collect a user space stack instead of a kernel stack.
2100 * **BPF_F_USER_BUILD_ID**
2101 * Collect buildid+offset instead of ips for user stack,
2102 * only valid if **BPF_F_USER_STACK** is also specified.
2104 * **bpf_get_stack**\ () can collect up to
2105 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2106 * to sufficient large buffer size. Note that
2107 * this limit can be controlled with the **sysctl** program, and
2108 * that it should be manually increased in order to profile long
2109 * user stacks (such as stacks for Java programs). To do so, use:
2113 * # sysctl kernel.perf_event_max_stack=<new value>
2115 * A non-negative value equal to or less than *size* on success,
2116 * or a negative error in case of failure.
2118 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2120 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2121 * it provides an easy way to load *len* bytes from *offset*
2122 * from the packet associated to *skb*, into the buffer pointed
2123 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2124 * a fifth argument *start_header* exists in order to select a
2125 * base offset to start from. *start_header* can be one of:
2127 * **BPF_HDR_START_MAC**
2128 * Base offset to load data from is *skb*'s mac header.
2129 * **BPF_HDR_START_NET**
2130 * Base offset to load data from is *skb*'s network header.
2132 * In general, "direct packet access" is the preferred method to
2133 * access packet data, however, this helper is in particular useful
2134 * in socket filters where *skb*\ **->data** does not always point
2135 * to the start of the mac header and where "direct packet access"
2138 * 0 on success, or a negative error in case of failure.
2140 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2142 * Do FIB lookup in kernel tables using parameters in *params*.
2143 * If lookup is successful and result shows packet is to be
2144 * forwarded, the neighbor tables are searched for the nexthop.
2145 * If successful (ie., FIB lookup shows forwarding and nexthop
2146 * is resolved), the nexthop address is returned in ipv4_dst
2147 * or ipv6_dst based on family, smac is set to mac address of
2148 * egress device, dmac is set to nexthop mac address, rt_metric
2149 * is set to metric from route (IPv4/IPv6 only), and ifindex
2150 * is set to the device index of the nexthop from the FIB lookup.
2152 * *plen* argument is the size of the passed in struct.
2153 * *flags* argument can be a combination of one or more of the
2156 * **BPF_FIB_LOOKUP_DIRECT**
2157 * Do a direct table lookup vs full lookup using FIB
2159 * **BPF_FIB_LOOKUP_OUTPUT**
2160 * Perform lookup from an egress perspective (default is
2163 * *ctx* is either **struct xdp_md** for XDP programs or
2164 * **struct sk_buff** tc cls_act programs.
2166 * * < 0 if any input argument is invalid
2167 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2168 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2169 * packet is not forwarded or needs assist from full stack
2171 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2173 * Add an entry to, or update a sockhash *map* referencing sockets.
2174 * The *skops* is used as a new value for the entry associated to
2175 * *key*. *flags* is one of:
2178 * The entry for *key* must not exist in the map.
2180 * The entry for *key* must already exist in the map.
2182 * No condition on the existence of the entry for *key*.
2184 * If the *map* has eBPF programs (parser and verdict), those will
2185 * be inherited by the socket being added. If the socket is
2186 * already attached to eBPF programs, this results in an error.
2188 * 0 on success, or a negative error in case of failure.
2190 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2192 * This helper is used in programs implementing policies at the
2193 * socket level. If the message *msg* is allowed to pass (i.e. if
2194 * the verdict eBPF program returns **SK_PASS**), redirect it to
2195 * the socket referenced by *map* (of type
2196 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2197 * egress interfaces can be used for redirection. The
2198 * **BPF_F_INGRESS** value in *flags* is used to make the
2199 * distinction (ingress path is selected if the flag is present,
2200 * egress path otherwise). This is the only flag supported for now.
2202 * **SK_PASS** on success, or **SK_DROP** on error.
2204 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2206 * This helper is used in programs implementing policies at the
2207 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2208 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2209 * to the socket referenced by *map* (of type
2210 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2211 * egress interfaces can be used for redirection. The
2212 * **BPF_F_INGRESS** value in *flags* is used to make the
2213 * distinction (ingress path is selected if the flag is present,
2214 * egress otherwise). This is the only flag supported for now.
2216 * **SK_PASS** on success, or **SK_DROP** on error.
2218 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2220 * Encapsulate the packet associated to *skb* within a Layer 3
2221 * protocol header. This header is provided in the buffer at
2222 * address *hdr*, with *len* its size in bytes. *type* indicates
2223 * the protocol of the header and can be one of:
2225 * **BPF_LWT_ENCAP_SEG6**
2226 * IPv6 encapsulation with Segment Routing Header
2227 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2228 * the IPv6 header is computed by the kernel.
2229 * **BPF_LWT_ENCAP_SEG6_INLINE**
2230 * Only works if *skb* contains an IPv6 packet. Insert a
2231 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2233 * **BPF_LWT_ENCAP_IP**
2234 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2235 * must be IPv4 or IPv6, followed by zero or more
2236 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2237 * total bytes in all prepended headers. Please note that
2238 * if **skb_is_gso**\ (*skb*) is true, no more than two
2239 * headers can be prepended, and the inner header, if
2240 * present, should be either GRE or UDP/GUE.
2242 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2243 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2244 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2245 * **BPF_PROG_TYPE_LWT_XMIT**.
2247 * A call to this helper is susceptible to change the underlying
2248 * packet buffer. Therefore, at load time, all checks on pointers
2249 * previously done by the verifier are invalidated and must be
2250 * performed again, if the helper is used in combination with
2251 * direct packet access.
2253 * 0 on success, or a negative error in case of failure.
2255 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2257 * Store *len* bytes from address *from* into the packet
2258 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2259 * inside the outermost IPv6 Segment Routing Header can be
2260 * modified through this helper.
2262 * A call to this helper is susceptible to change the underlying
2263 * packet buffer. Therefore, at load time, all checks on pointers
2264 * previously done by the verifier are invalidated and must be
2265 * performed again, if the helper is used in combination with
2266 * direct packet access.
2268 * 0 on success, or a negative error in case of failure.
2270 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2272 * Adjust the size allocated to TLVs in the outermost IPv6
2273 * Segment Routing Header contained in the packet associated to
2274 * *skb*, at position *offset* by *delta* bytes. Only offsets
2275 * after the segments are accepted. *delta* can be as well
2276 * positive (growing) as negative (shrinking).
2278 * A call to this helper is susceptible to change the underlying
2279 * packet buffer. Therefore, at load time, all checks on pointers
2280 * previously done by the verifier are invalidated and must be
2281 * performed again, if the helper is used in combination with
2282 * direct packet access.
2284 * 0 on success, or a negative error in case of failure.
2286 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2288 * Apply an IPv6 Segment Routing action of type *action* to the
2289 * packet associated to *skb*. Each action takes a parameter
2290 * contained at address *param*, and of length *param_len* bytes.
2291 * *action* can be one of:
2293 * **SEG6_LOCAL_ACTION_END_X**
2294 * End.X action: Endpoint with Layer-3 cross-connect.
2295 * Type of *param*: **struct in6_addr**.
2296 * **SEG6_LOCAL_ACTION_END_T**
2297 * End.T action: Endpoint with specific IPv6 table lookup.
2298 * Type of *param*: **int**.
2299 * **SEG6_LOCAL_ACTION_END_B6**
2300 * End.B6 action: Endpoint bound to an SRv6 policy.
2301 * Type of *param*: **struct ipv6_sr_hdr**.
2302 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2303 * End.B6.Encap action: Endpoint bound to an SRv6
2304 * encapsulation policy.
2305 * Type of *param*: **struct ipv6_sr_hdr**.
2307 * A call to this helper is susceptible to change the underlying
2308 * packet buffer. Therefore, at load time, all checks on pointers
2309 * previously done by the verifier are invalidated and must be
2310 * performed again, if the helper is used in combination with
2311 * direct packet access.
2313 * 0 on success, or a negative error in case of failure.
2315 * long bpf_rc_repeat(void *ctx)
2317 * This helper is used in programs implementing IR decoding, to
2318 * report a successfully decoded repeat key message. This delays
2319 * the generation of a key up event for previously generated
2322 * Some IR protocols like NEC have a special IR message for
2323 * repeating last button, for when a button is held down.
2325 * The *ctx* should point to the lirc sample as passed into
2328 * This helper is only available is the kernel was compiled with
2329 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2334 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2336 * This helper is used in programs implementing IR decoding, to
2337 * report a successfully decoded key press with *scancode*,
2338 * *toggle* value in the given *protocol*. The scancode will be
2339 * translated to a keycode using the rc keymap, and reported as
2340 * an input key down event. After a period a key up event is
2341 * generated. This period can be extended by calling either
2342 * **bpf_rc_keydown**\ () again with the same values, or calling
2343 * **bpf_rc_repeat**\ ().
2345 * Some protocols include a toggle bit, in case the button was
2346 * released and pressed again between consecutive scancodes.
2348 * The *ctx* should point to the lirc sample as passed into
2351 * The *protocol* is the decoded protocol number (see
2352 * **enum rc_proto** for some predefined values).
2354 * This helper is only available is the kernel was compiled with
2355 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2360 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2362 * Return the cgroup v2 id of the socket associated with the *skb*.
2363 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2364 * helper for cgroup v1 by providing a tag resp. identifier that
2365 * can be matched on or used for map lookups e.g. to implement
2366 * policy. The cgroup v2 id of a given path in the hierarchy is
2367 * exposed in user space through the f_handle API in order to get
2368 * to the same 64-bit id.
2370 * This helper can be used on TC egress path, but not on ingress,
2371 * and is available only if the kernel was compiled with the
2372 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2374 * The id is returned or 0 in case the id could not be retrieved.
2376 * u64 bpf_get_current_cgroup_id(void)
2378 * A 64-bit integer containing the current cgroup id based
2379 * on the cgroup within which the current task is running.
2381 * void *bpf_get_local_storage(void *map, u64 flags)
2383 * Get the pointer to the local storage area.
2384 * The type and the size of the local storage is defined
2385 * by the *map* argument.
2386 * The *flags* meaning is specific for each map type,
2387 * and has to be 0 for cgroup local storage.
2389 * Depending on the BPF program type, a local storage area
2390 * can be shared between multiple instances of the BPF program,
2391 * running simultaneously.
2393 * A user should care about the synchronization by himself.
2394 * For example, by using the **BPF_STX_XADD** instruction to alter
2397 * A pointer to the local storage area.
2399 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2401 * Select a **SO_REUSEPORT** socket from a
2402 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2403 * It checks the selected socket is matching the incoming
2404 * request in the socket buffer.
2406 * 0 on success, or a negative error in case of failure.
2408 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2410 * Return id of cgroup v2 that is ancestor of cgroup associated
2411 * with the *skb* at the *ancestor_level*. The root cgroup is at
2412 * *ancestor_level* zero and each step down the hierarchy
2413 * increments the level. If *ancestor_level* == level of cgroup
2414 * associated with *skb*, then return value will be same as that
2415 * of **bpf_skb_cgroup_id**\ ().
2417 * The helper is useful to implement policies based on cgroups
2418 * that are upper in hierarchy than immediate cgroup associated
2421 * The format of returned id and helper limitations are same as in
2422 * **bpf_skb_cgroup_id**\ ().
2424 * The id is returned or 0 in case the id could not be retrieved.
2426 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2428 * Look for TCP socket matching *tuple*, optionally in a child
2429 * network namespace *netns*. The return value must be checked,
2430 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2432 * The *ctx* should point to the context of the program, such as
2433 * the skb or socket (depending on the hook in use). This is used
2434 * to determine the base network namespace for the lookup.
2436 * *tuple_size* must be one of:
2438 * **sizeof**\ (*tuple*\ **->ipv4**)
2439 * Look for an IPv4 socket.
2440 * **sizeof**\ (*tuple*\ **->ipv6**)
2441 * Look for an IPv6 socket.
2443 * If the *netns* is a negative signed 32-bit integer, then the
2444 * socket lookup table in the netns associated with the *ctx*
2445 * will be used. For the TC hooks, this is the netns of the device
2446 * in the skb. For socket hooks, this is the netns of the socket.
2447 * If *netns* is any other signed 32-bit value greater than or
2448 * equal to zero then it specifies the ID of the netns relative to
2449 * the netns associated with the *ctx*. *netns* values beyond the
2450 * range of 32-bit integers are reserved for future use.
2452 * All values for *flags* are reserved for future usage, and must
2455 * This helper is available only if the kernel was compiled with
2456 * **CONFIG_NET** configuration option.
2458 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2459 * For sockets with reuseport option, the **struct bpf_sock**
2460 * result is from *reuse*\ **->socks**\ [] using the hash of the
2463 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2465 * Look for UDP socket matching *tuple*, optionally in a child
2466 * network namespace *netns*. The return value must be checked,
2467 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2469 * The *ctx* should point to the context of the program, such as
2470 * the skb or socket (depending on the hook in use). This is used
2471 * to determine the base network namespace for the lookup.
2473 * *tuple_size* must be one of:
2475 * **sizeof**\ (*tuple*\ **->ipv4**)
2476 * Look for an IPv4 socket.
2477 * **sizeof**\ (*tuple*\ **->ipv6**)
2478 * Look for an IPv6 socket.
2480 * If the *netns* is a negative signed 32-bit integer, then the
2481 * socket lookup table in the netns associated with the *ctx*
2482 * will be used. For the TC hooks, this is the netns of the device
2483 * in the skb. For socket hooks, this is the netns of the socket.
2484 * If *netns* is any other signed 32-bit value greater than or
2485 * equal to zero then it specifies the ID of the netns relative to
2486 * the netns associated with the *ctx*. *netns* values beyond the
2487 * range of 32-bit integers are reserved for future use.
2489 * All values for *flags* are reserved for future usage, and must
2492 * This helper is available only if the kernel was compiled with
2493 * **CONFIG_NET** configuration option.
2495 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2496 * For sockets with reuseport option, the **struct bpf_sock**
2497 * result is from *reuse*\ **->socks**\ [] using the hash of the
2500 * long bpf_sk_release(struct bpf_sock *sock)
2502 * Release the reference held by *sock*. *sock* must be a
2503 * non-**NULL** pointer that was returned from
2504 * **bpf_sk_lookup_xxx**\ ().
2506 * 0 on success, or a negative error in case of failure.
2508 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2510 * Push an element *value* in *map*. *flags* is one of:
2513 * If the queue/stack is full, the oldest element is
2514 * removed to make room for this.
2516 * 0 on success, or a negative error in case of failure.
2518 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
2520 * Pop an element from *map*.
2522 * 0 on success, or a negative error in case of failure.
2524 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
2526 * Get an element from *map* without removing it.
2528 * 0 on success, or a negative error in case of failure.
2530 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2532 * For socket policies, insert *len* bytes into *msg* at offset
2535 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2536 * *msg* it may want to insert metadata or options into the *msg*.
2537 * This can later be read and used by any of the lower layer BPF
2540 * This helper may fail if under memory pressure (a malloc
2541 * fails) in these cases BPF programs will get an appropriate
2542 * error and BPF programs will need to handle them.
2544 * 0 on success, or a negative error in case of failure.
2546 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2548 * Will remove *len* bytes from a *msg* starting at byte *start*.
2549 * This may result in **ENOMEM** errors under certain situations if
2550 * an allocation and copy are required due to a full ring buffer.
2551 * However, the helper will try to avoid doing the allocation
2552 * if possible. Other errors can occur if input parameters are
2553 * invalid either due to *start* byte not being valid part of *msg*
2554 * payload and/or *pop* value being to large.
2556 * 0 on success, or a negative error in case of failure.
2558 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2560 * This helper is used in programs implementing IR decoding, to
2561 * report a successfully decoded pointer movement.
2563 * The *ctx* should point to the lirc sample as passed into
2566 * This helper is only available is the kernel was compiled with
2567 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2572 * long bpf_spin_lock(struct bpf_spin_lock *lock)
2574 * Acquire a spinlock represented by the pointer *lock*, which is
2575 * stored as part of a value of a map. Taking the lock allows to
2576 * safely update the rest of the fields in that value. The
2577 * spinlock can (and must) later be released with a call to
2578 * **bpf_spin_unlock**\ (\ *lock*\ ).
2580 * Spinlocks in BPF programs come with a number of restrictions
2583 * * **bpf_spin_lock** objects are only allowed inside maps of
2584 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2585 * list could be extended in the future).
2586 * * BTF description of the map is mandatory.
2587 * * The BPF program can take ONE lock at a time, since taking two
2588 * or more could cause dead locks.
2589 * * Only one **struct bpf_spin_lock** is allowed per map element.
2590 * * When the lock is taken, calls (either BPF to BPF or helpers)
2592 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2593 * allowed inside a spinlock-ed region.
2594 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2595 * the lock, on all execution paths, before it returns.
2596 * * The BPF program can access **struct bpf_spin_lock** only via
2597 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2598 * helpers. Loading or storing data into the **struct
2599 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2600 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2601 * of the map value must be a struct and have **struct
2602 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2603 * Nested lock inside another struct is not allowed.
2604 * * The **struct bpf_spin_lock** *lock* field in a map value must
2605 * be aligned on a multiple of 4 bytes in that value.
2606 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2607 * the **bpf_spin_lock** field to user space.
2608 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2609 * a BPF program, do not update the **bpf_spin_lock** field.
2610 * * **bpf_spin_lock** cannot be on the stack or inside a
2611 * networking packet (it can only be inside of a map values).
2612 * * **bpf_spin_lock** is available to root only.
2613 * * Tracing programs and socket filter programs cannot use
2614 * **bpf_spin_lock**\ () due to insufficient preemption checks
2615 * (but this may change in the future).
2616 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2620 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
2622 * Release the *lock* previously locked by a call to
2623 * **bpf_spin_lock**\ (\ *lock*\ ).
2627 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2629 * This helper gets a **struct bpf_sock** pointer such
2630 * that all the fields in this **bpf_sock** can be accessed.
2632 * A **struct bpf_sock** pointer on success, or **NULL** in
2635 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2637 * This helper gets a **struct bpf_tcp_sock** pointer from a
2638 * **struct bpf_sock** pointer.
2640 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2643 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
2645 * Set ECN (Explicit Congestion Notification) field of IP header
2646 * to **CE** (Congestion Encountered) if current value is **ECT**
2647 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2650 * 1 if the **CE** flag is set (either by the current helper call
2651 * or because it was already present), 0 if it is not set.
2653 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2655 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2656 * **bpf_sk_release**\ () is unnecessary and not allowed.
2658 * A **struct bpf_sock** pointer on success, or **NULL** in
2661 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2663 * Look for TCP socket matching *tuple*, optionally in a child
2664 * network namespace *netns*. The return value must be checked,
2665 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2667 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2668 * that it also returns timewait or request sockets. Use
2669 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2672 * This helper is available only if the kernel was compiled with
2673 * **CONFIG_NET** configuration option.
2675 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2676 * For sockets with reuseport option, the **struct bpf_sock**
2677 * result is from *reuse*\ **->socks**\ [] using the hash of the
2680 * long bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2682 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2683 * the listening socket in *sk*.
2685 * *iph* points to the start of the IPv4 or IPv6 header, while
2686 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2687 * **sizeof**\ (**struct ip6hdr**).
2689 * *th* points to the start of the TCP header, while *th_len*
2690 * contains **sizeof**\ (**struct tcphdr**).
2692 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2695 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2697 * Get name of sysctl in /proc/sys/ and copy it into provided by
2698 * program buffer *buf* of size *buf_len*.
2700 * The buffer is always NUL terminated, unless it's zero-sized.
2702 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2703 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2704 * only (e.g. "tcp_mem").
2706 * Number of character copied (not including the trailing NUL).
2708 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2709 * truncated name in this case).
2711 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2713 * Get current value of sysctl as it is presented in /proc/sys
2714 * (incl. newline, etc), and copy it as a string into provided
2715 * by program buffer *buf* of size *buf_len*.
2717 * The whole value is copied, no matter what file position user
2718 * space issued e.g. sys_read at.
2720 * The buffer is always NUL terminated, unless it's zero-sized.
2722 * Number of character copied (not including the trailing NUL).
2724 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2725 * truncated name in this case).
2727 * **-EINVAL** if current value was unavailable, e.g. because
2728 * sysctl is uninitialized and read returns -EIO for it.
2730 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2732 * Get new value being written by user space to sysctl (before
2733 * the actual write happens) and copy it as a string into
2734 * provided by program buffer *buf* of size *buf_len*.
2736 * User space may write new value at file position > 0.
2738 * The buffer is always NUL terminated, unless it's zero-sized.
2740 * Number of character copied (not including the trailing NUL).
2742 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2743 * truncated name in this case).
2745 * **-EINVAL** if sysctl is being read.
2747 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2749 * Override new value being written by user space to sysctl with
2750 * value provided by program in buffer *buf* of size *buf_len*.
2752 * *buf* should contain a string in same form as provided by user
2753 * space on sysctl write.
2755 * User space may write new value at file position > 0. To override
2756 * the whole sysctl value file position should be set to zero.
2760 * **-E2BIG** if the *buf_len* is too big.
2762 * **-EINVAL** if sysctl is being read.
2764 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2766 * Convert the initial part of the string from buffer *buf* of
2767 * size *buf_len* to a long integer according to the given base
2768 * and save the result in *res*.
2770 * The string may begin with an arbitrary amount of white space
2771 * (as determined by **isspace**\ (3)) followed by a single
2772 * optional '**-**' sign.
2774 * Five least significant bits of *flags* encode base, other bits
2775 * are currently unused.
2777 * Base must be either 8, 10, 16 or 0 to detect it automatically
2778 * similar to user space **strtol**\ (3).
2780 * Number of characters consumed on success. Must be positive but
2781 * no more than *buf_len*.
2783 * **-EINVAL** if no valid digits were found or unsupported base
2786 * **-ERANGE** if resulting value was out of range.
2788 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2790 * Convert the initial part of the string from buffer *buf* of
2791 * size *buf_len* to an unsigned long integer according to the
2792 * given base and save the result in *res*.
2794 * The string may begin with an arbitrary amount of white space
2795 * (as determined by **isspace**\ (3)).
2797 * Five least significant bits of *flags* encode base, other bits
2798 * are currently unused.
2800 * Base must be either 8, 10, 16 or 0 to detect it automatically
2801 * similar to user space **strtoul**\ (3).
2803 * Number of characters consumed on success. Must be positive but
2804 * no more than *buf_len*.
2806 * **-EINVAL** if no valid digits were found or unsupported base
2809 * **-ERANGE** if resulting value was out of range.
2811 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
2813 * Get a bpf-local-storage from a *sk*.
2815 * Logically, it could be thought of getting the value from
2816 * a *map* with *sk* as the **key**. From this
2817 * perspective, the usage is not much different from
2818 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2819 * helper enforces the key must be a full socket and the map must
2820 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2822 * Underneath, the value is stored locally at *sk* instead of
2823 * the *map*. The *map* is used as the bpf-local-storage
2824 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2825 * searched against all bpf-local-storages residing at *sk*.
2827 * *sk* is a kernel **struct sock** pointer for LSM program.
2828 * *sk* is a **struct bpf_sock** pointer for other program types.
2830 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2831 * used such that a new bpf-local-storage will be
2832 * created if one does not exist. *value* can be used
2833 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2834 * the initial value of a bpf-local-storage. If *value* is
2835 * **NULL**, the new bpf-local-storage will be zero initialized.
2837 * A bpf-local-storage pointer is returned on success.
2839 * **NULL** if not found or there was an error in adding
2840 * a new bpf-local-storage.
2842 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
2844 * Delete a bpf-local-storage from a *sk*.
2848 * **-ENOENT** if the bpf-local-storage cannot be found.
2850 * long bpf_send_signal(u32 sig)
2852 * Send signal *sig* to the process of the current task.
2853 * The signal may be delivered to any of this process's threads.
2855 * 0 on success or successfully queued.
2857 * **-EBUSY** if work queue under nmi is full.
2859 * **-EINVAL** if *sig* is invalid.
2861 * **-EPERM** if no permission to send the *sig*.
2863 * **-EAGAIN** if bpf program can try again.
2865 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2867 * Try to issue a SYN cookie for the packet with corresponding
2868 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2870 * *iph* points to the start of the IPv4 or IPv6 header, while
2871 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2872 * **sizeof**\ (**struct ip6hdr**).
2874 * *th* points to the start of the TCP header, while *th_len*
2875 * contains the length of the TCP header.
2877 * On success, lower 32 bits hold the generated SYN cookie in
2878 * followed by 16 bits which hold the MSS value for that cookie,
2879 * and the top 16 bits are unused.
2881 * On failure, the returned value is one of the following:
2883 * **-EINVAL** SYN cookie cannot be issued due to error
2885 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2887 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2889 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2891 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2893 * Write raw *data* blob into a special BPF perf event held by
2894 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2895 * event must have the following attributes: **PERF_SAMPLE_RAW**
2896 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2897 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2899 * The *flags* are used to indicate the index in *map* for which
2900 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2901 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2902 * to indicate that the index of the current CPU core should be
2905 * The value to write, of *size*, is passed through eBPF stack and
2906 * pointed by *data*.
2908 * *ctx* is a pointer to in-kernel struct sk_buff.
2910 * This helper is similar to **bpf_perf_event_output**\ () but
2911 * restricted to raw_tracepoint bpf programs.
2913 * 0 on success, or a negative error in case of failure.
2915 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2917 * Safely attempt to read *size* bytes from user space address
2918 * *unsafe_ptr* and store the data in *dst*.
2920 * 0 on success, or a negative error in case of failure.
2922 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2924 * Safely attempt to read *size* bytes from kernel space address
2925 * *unsafe_ptr* and store the data in *dst*.
2927 * 0 on success, or a negative error in case of failure.
2929 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2931 * Copy a NUL terminated string from an unsafe user address
2932 * *unsafe_ptr* to *dst*. The *size* should include the
2933 * terminating NUL byte. In case the string length is smaller than
2934 * *size*, the target is not padded with further NUL bytes. If the
2935 * string length is larger than *size*, just *size*-1 bytes are
2936 * copied and the last byte is set to NUL.
2938 * On success, the length of the copied string is returned. This
2939 * makes this helper useful in tracing programs for reading
2940 * strings, and more importantly to get its length at runtime. See
2941 * the following snippet:
2945 * SEC("kprobe/sys_open")
2946 * void bpf_sys_open(struct pt_regs *ctx)
2948 * char buf[PATHLEN]; // PATHLEN is defined to 256
2949 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2952 * // Consume buf, for example push it to
2953 * // userspace via bpf_perf_event_output(); we
2954 * // can use res (the string length) as event
2955 * // size, after checking its boundaries.
2958 * In comparison, using **bpf_probe_read_user**\ () helper here
2959 * instead to read the string would require to estimate the length
2960 * at compile time, and would often result in copying more memory
2963 * Another useful use case is when parsing individual process
2964 * arguments or individual environment variables navigating
2965 * *current*\ **->mm->arg_start** and *current*\
2966 * **->mm->env_start**: using this helper and the return value,
2967 * one can quickly iterate at the right offset of the memory area.
2969 * On success, the strictly positive length of the string,
2970 * including the trailing NUL character. On error, a negative
2973 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2975 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
2976 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
2978 * On success, the strictly positive length of the string, including
2979 * the trailing NUL character. On error, a negative value.
2981 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2983 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
2984 * *rcv_nxt* is the ack_seq to be sent out.
2986 * 0 on success, or a negative error in case of failure.
2988 * long bpf_send_signal_thread(u32 sig)
2990 * Send signal *sig* to the thread corresponding to the current task.
2992 * 0 on success or successfully queued.
2994 * **-EBUSY** if work queue under nmi is full.
2996 * **-EINVAL** if *sig* is invalid.
2998 * **-EPERM** if no permission to send the *sig*.
3000 * **-EAGAIN** if bpf program can try again.
3002 * u64 bpf_jiffies64(void)
3004 * Obtain the 64bit jiffies
3006 * The 64 bit jiffies
3008 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3010 * For an eBPF program attached to a perf event, retrieve the
3011 * branch records (**struct perf_branch_entry**) associated to *ctx*
3012 * and store it in the buffer pointed by *buf* up to size
3015 * On success, number of bytes written to *buf*. On error, a
3018 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3019 * instead return the number of bytes required to store all the
3020 * branch entries. If this flag is set, *buf* may be NULL.
3022 * **-EINVAL** if arguments invalid or **size** not a multiple
3023 * of **sizeof**\ (**struct perf_branch_entry**\ ).
3025 * **-ENOENT** if architecture does not support branch records.
3027 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3029 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3030 * *namespace* will be returned in *nsdata*.
3032 * 0 on success, or one of the following in case of failure:
3034 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3035 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3037 * **-ENOENT** if pidns does not exists for the current task.
3039 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3041 * Write raw *data* blob into a special BPF perf event held by
3042 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3043 * event must have the following attributes: **PERF_SAMPLE_RAW**
3044 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3045 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3047 * The *flags* are used to indicate the index in *map* for which
3048 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3049 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3050 * to indicate that the index of the current CPU core should be
3053 * The value to write, of *size*, is passed through eBPF stack and
3054 * pointed by *data*.
3056 * *ctx* is a pointer to in-kernel struct xdp_buff.
3058 * This helper is similar to **bpf_perf_eventoutput**\ () but
3059 * restricted to raw_tracepoint bpf programs.
3061 * 0 on success, or a negative error in case of failure.
3063 * u64 bpf_get_netns_cookie(void *ctx)
3065 * Retrieve the cookie (generated by the kernel) of the network
3066 * namespace the input *ctx* is associated with. The network
3067 * namespace cookie remains stable for its lifetime and provides
3068 * a global identifier that can be assumed unique. If *ctx* is
3069 * NULL, then the helper returns the cookie for the initial
3070 * network namespace. The cookie itself is very similar to that
3071 * of **bpf_get_socket_cookie**\ () helper, but for network
3072 * namespaces instead of sockets.
3074 * A 8-byte long opaque number.
3076 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3078 * Return id of cgroup v2 that is ancestor of the cgroup associated
3079 * with the current task at the *ancestor_level*. The root cgroup
3080 * is at *ancestor_level* zero and each step down the hierarchy
3081 * increments the level. If *ancestor_level* == level of cgroup
3082 * associated with the current task, then return value will be the
3083 * same as that of **bpf_get_current_cgroup_id**\ ().
3085 * The helper is useful to implement policies based on cgroups
3086 * that are upper in hierarchy than immediate cgroup associated
3087 * with the current task.
3089 * The format of returned id and helper limitations are same as in
3090 * **bpf_get_current_cgroup_id**\ ().
3092 * The id is returned or 0 in case the id could not be retrieved.
3094 * long bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags)
3096 * Helper is overloaded depending on BPF program type. This
3097 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3098 * **BPF_PROG_TYPE_SCHED_ACT** programs.
3100 * Assign the *sk* to the *skb*. When combined with appropriate
3101 * routing configuration to receive the packet towards the socket,
3102 * will cause *skb* to be delivered to the specified socket.
3103 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3104 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3105 * interfere with successful delivery to the socket.
3107 * This operation is only valid from TC ingress path.
3109 * The *flags* argument must be zero.
3111 * 0 on success, or a negative error in case of failure:
3113 * **-EINVAL** if specified *flags* are not supported.
3115 * **-ENOENT** if the socket is unavailable for assignment.
3117 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3119 * **-EOPNOTSUPP** if the operation is not supported, for example
3120 * a call from outside of TC ingress.
3122 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3125 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3127 * Helper is overloaded depending on BPF program type. This
3128 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
3130 * Select the *sk* as a result of a socket lookup.
3132 * For the operation to succeed passed socket must be compatible
3133 * with the packet description provided by the *ctx* object.
3135 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
3136 * be an exact match. While IP family (**AF_INET** or
3137 * **AF_INET6**) must be compatible, that is IPv6 sockets
3138 * that are not v6-only can be selected for IPv4 packets.
3140 * Only TCP listeners and UDP unconnected sockets can be
3141 * selected. *sk* can also be NULL to reset any previous
3144 * *flags* argument can combination of following values:
3146 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
3147 * socket selection, potentially done by a BPF program
3148 * that ran before us.
3150 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
3151 * load-balancing within reuseport group for the socket
3154 * On success *ctx->sk* will point to the selected socket.
3157 * 0 on success, or a negative errno in case of failure.
3159 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
3160 * not compatible with packet family (*ctx->family*).
3162 * * **-EEXIST** if socket has been already selected,
3163 * potentially by another program, and
3164 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
3166 * * **-EINVAL** if unsupported flags were specified.
3168 * * **-EPROTOTYPE** if socket L4 protocol
3169 * (*sk->protocol*) doesn't match packet protocol
3170 * (*ctx->protocol*).
3172 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
3173 * state (TCP listening or UDP unconnected).
3175 * u64 bpf_ktime_get_boot_ns(void)
3177 * Return the time elapsed since system boot, in nanoseconds.
3178 * Does include the time the system was suspended.
3179 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
3183 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3185 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3186 * out the format string.
3187 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3188 * the format string itself. The *data* and *data_len* are format string
3189 * arguments. The *data* are a **u64** array and corresponding format string
3190 * values are stored in the array. For strings and pointers where pointees
3191 * are accessed, only the pointer values are stored in the *data* array.
3192 * The *data_len* is the size of *data* in bytes.
3194 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3195 * Reading kernel memory may fail due to either invalid address or
3196 * valid address but requiring a major memory fault. If reading kernel memory
3197 * fails, the string for **%s** will be an empty string, and the ip
3198 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
3199 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
3201 * 0 on success, or a negative error in case of failure:
3203 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3204 * by returning 1 from bpf program.
3206 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3208 * **-E2BIG** if *fmt* contains too many format specifiers.
3210 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3212 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
3214 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
3215 * The *m* represents the seq_file. The *data* and *len* represent the
3216 * data to write in bytes.
3218 * 0 on success, or a negative error in case of failure:
3220 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3222 * u64 bpf_sk_cgroup_id(struct bpf_sock *sk)
3224 * Return the cgroup v2 id of the socket *sk*.
3226 * *sk* must be a non-**NULL** pointer to a full socket, e.g. one
3227 * returned from **bpf_sk_lookup_xxx**\ (),
3228 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3229 * same as in **bpf_skb_cgroup_id**\ ().
3231 * This helper is available only if the kernel was compiled with
3232 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3234 * The id is returned or 0 in case the id could not be retrieved.
3236 * u64 bpf_sk_ancestor_cgroup_id(struct bpf_sock *sk, int ancestor_level)
3238 * Return id of cgroup v2 that is ancestor of cgroup associated
3239 * with the *sk* at the *ancestor_level*. The root cgroup is at
3240 * *ancestor_level* zero and each step down the hierarchy
3241 * increments the level. If *ancestor_level* == level of cgroup
3242 * associated with *sk*, then return value will be same as that
3243 * of **bpf_sk_cgroup_id**\ ().
3245 * The helper is useful to implement policies based on cgroups
3246 * that are upper in hierarchy than immediate cgroup associated
3249 * The format of returned id and helper limitations are same as in
3250 * **bpf_sk_cgroup_id**\ ().
3252 * The id is returned or 0 in case the id could not be retrieved.
3254 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
3256 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
3257 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3258 * of new data availability is sent.
3259 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3260 * of new data availability is sent unconditionally.
3262 * 0 on success, or a negative error in case of failure.
3264 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3266 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3268 * Valid pointer with *size* bytes of memory available; NULL,
3271 * void bpf_ringbuf_submit(void *data, u64 flags)
3273 * Submit reserved ring buffer sample, pointed to by *data*.
3274 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3275 * of new data availability is sent.
3276 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3277 * of new data availability is sent unconditionally.
3279 * Nothing. Always succeeds.
3281 * void bpf_ringbuf_discard(void *data, u64 flags)
3283 * Discard reserved ring buffer sample, pointed to by *data*.
3284 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3285 * of new data availability is sent.
3286 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3287 * of new data availability is sent unconditionally.
3289 * Nothing. Always succeeds.
3291 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3293 * Query various characteristics of provided ring buffer. What
3294 * exactly is queries is determined by *flags*:
3296 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
3297 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
3298 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
3299 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
3301 * Data returned is just a momentary snapshot of actual values
3302 * and could be inaccurate, so this facility should be used to
3303 * power heuristics and for reporting, not to make 100% correct
3306 * Requested value, or 0, if *flags* are not recognized.
3308 * long bpf_csum_level(struct sk_buff *skb, u64 level)
3310 * Change the skbs checksum level by one layer up or down, or
3311 * reset it entirely to none in order to have the stack perform
3312 * checksum validation. The level is applicable to the following
3313 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
3314 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
3315 * through **bpf_skb_adjust_room**\ () helper with passing in
3316 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
3317 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
3318 * the UDP header is removed. Similarly, an encap of the latter
3319 * into the former could be accompanied by a helper call to
3320 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
3321 * skb is still intended to be processed in higher layers of the
3322 * stack instead of just egressing at tc.
3324 * There are three supported level settings at this time:
3326 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
3327 * with CHECKSUM_UNNECESSARY.
3328 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
3329 * with CHECKSUM_UNNECESSARY.
3330 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
3331 * sets CHECKSUM_NONE to force checksum validation by the stack.
3332 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
3335 * 0 on success, or a negative error in case of failure. In the
3336 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
3337 * is returned or the error code -EACCES in case the skb is not
3338 * subject to CHECKSUM_UNNECESSARY.
3340 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
3342 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
3344 * *sk* if casting is valid, or NULL otherwise.
3346 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
3348 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
3350 * *sk* if casting is valid, or NULL otherwise.
3352 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
3354 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
3356 * *sk* if casting is valid, or NULL otherwise.
3358 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
3360 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
3362 * *sk* if casting is valid, or NULL otherwise.
3364 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
3366 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
3368 * *sk* if casting is valid, or NULL otherwise.
3370 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
3372 * Return a user or a kernel stack in bpf program provided buffer.
3373 * To achieve this, the helper needs *task*, which is a valid
3374 * pointer to struct task_struct. To store the stacktrace, the
3375 * bpf program provides *buf* with a nonnegative *size*.
3377 * The last argument, *flags*, holds the number of stack frames to
3378 * skip (from 0 to 255), masked with
3379 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3380 * the following flags:
3382 * **BPF_F_USER_STACK**
3383 * Collect a user space stack instead of a kernel stack.
3384 * **BPF_F_USER_BUILD_ID**
3385 * Collect buildid+offset instead of ips for user stack,
3386 * only valid if **BPF_F_USER_STACK** is also specified.
3388 * **bpf_get_task_stack**\ () can collect up to
3389 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3390 * to sufficient large buffer size. Note that
3391 * this limit can be controlled with the **sysctl** program, and
3392 * that it should be manually increased in order to profile long
3393 * user stacks (such as stacks for Java programs). To do so, use:
3397 * # sysctl kernel.perf_event_max_stack=<new value>
3399 * A non-negative value equal to or less than *size* on success,
3400 * or a negative error in case of failure.
3402 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
3404 * Load header option. Support reading a particular TCP header
3405 * option for bpf program (BPF_PROG_TYPE_SOCK_OPS).
3407 * If *flags* is 0, it will search the option from the
3408 * sock_ops->skb_data. The comment in "struct bpf_sock_ops"
3409 * has details on what skb_data contains under different
3412 * The first byte of the *searchby_res* specifies the
3413 * kind that it wants to search.
3415 * If the searching kind is an experimental kind
3416 * (i.e. 253 or 254 according to RFC6994). It also
3417 * needs to specify the "magic" which is either
3418 * 2 bytes or 4 bytes. It then also needs to
3419 * specify the size of the magic by using
3420 * the 2nd byte which is "kind-length" of a TCP
3421 * header option and the "kind-length" also
3422 * includes the first 2 bytes "kind" and "kind-length"
3423 * itself as a normal TCP header option also does.
3425 * For example, to search experimental kind 254 with
3426 * 2 byte magic 0xeB9F, the searchby_res should be
3427 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
3429 * To search for the standard window scale option (3),
3430 * the searchby_res should be [ 3, 0, 0, .... 0 ].
3431 * Note, kind-length must be 0 for regular option.
3433 * Searching for No-Op (0) and End-of-Option-List (1) are
3436 * *len* must be at least 2 bytes which is the minimal size
3437 * of a header option.
3440 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
3441 * saved_syn packet or the just-received syn packet.
3444 * >0 when found, the header option is copied to *searchby_res*.
3445 * The return value is the total length copied.
3447 * **-EINVAL** If param is invalid
3449 * **-ENOMSG** The option is not found
3451 * **-ENOENT** No syn packet available when
3452 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used
3454 * **-ENOSPC** Not enough space. Only *len* number of
3457 * **-EFAULT** Cannot parse the header options in the packet
3459 * **-EPERM** This helper cannot be used under the
3460 * current sock_ops->op.
3462 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
3464 * Store header option. The data will be copied
3465 * from buffer *from* with length *len* to the TCP header.
3467 * The buffer *from* should have the whole option that
3468 * includes the kind, kind-length, and the actual
3469 * option data. The *len* must be at least kind-length
3470 * long. The kind-length does not have to be 4 byte
3471 * aligned. The kernel will take care of the padding
3472 * and setting the 4 bytes aligned value to th->doff.
3474 * This helper will check for duplicated option
3475 * by searching the same option in the outgoing skb.
3477 * This helper can only be called during
3478 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
3481 * 0 on success, or negative error in case of failure:
3483 * **-EINVAL** If param is invalid
3485 * **-ENOSPC** Not enough space in the header.
3486 * Nothing has been written
3488 * **-EEXIST** The option has already existed
3490 * **-EFAULT** Cannot parse the existing header options
3492 * **-EPERM** This helper cannot be used under the
3493 * current sock_ops->op.
3495 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
3497 * Reserve *len* bytes for the bpf header option. The
3498 * space will be used by bpf_store_hdr_opt() later in
3499 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
3501 * If bpf_reserve_hdr_opt() is called multiple times,
3502 * the total number of bytes will be reserved.
3504 * This helper can only be called during
3505 * BPF_SOCK_OPS_HDR_OPT_LEN_CB.
3508 * 0 on success, or negative error in case of failure:
3510 * **-EINVAL** if param is invalid
3512 * **-ENOSPC** Not enough space in the header.
3514 * **-EPERM** This helper cannot be used under the
3515 * current sock_ops->op.
3517 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
3519 * Get a bpf_local_storage from an *inode*.
3521 * Logically, it could be thought of as getting the value from
3522 * a *map* with *inode* as the **key**. From this
3523 * perspective, the usage is not much different from
3524 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
3525 * helper enforces the key must be an inode and the map must also
3526 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
3528 * Underneath, the value is stored locally at *inode* instead of
3529 * the *map*. The *map* is used as the bpf-local-storage
3530 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3531 * searched against all bpf_local_storage residing at *inode*.
3533 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
3534 * used such that a new bpf_local_storage will be
3535 * created if one does not exist. *value* can be used
3536 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
3537 * the initial value of a bpf_local_storage. If *value* is
3538 * **NULL**, the new bpf_local_storage will be zero initialized.
3540 * A bpf_local_storage pointer is returned on success.
3542 * **NULL** if not found or there was an error in adding
3543 * a new bpf_local_storage.
3545 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
3547 * Delete a bpf_local_storage from an *inode*.
3551 * **-ENOENT** if the bpf_local_storage cannot be found.
3553 * long bpf_d_path(struct path *path, char *buf, u32 sz)
3555 * Return full path for given 'struct path' object, which
3556 * needs to be the kernel BTF 'path' object. The path is
3557 * returned in the provided buffer 'buf' of size 'sz' and
3558 * is zero terminated.
3561 * On success, the strictly positive length of the string,
3562 * including the trailing NUL character. On error, a negative
3565 #define __BPF_FUNC_MAPPER(FN) \
3567 FN(map_lookup_elem), \
3568 FN(map_update_elem), \
3569 FN(map_delete_elem), \
3573 FN(get_prandom_u32), \
3574 FN(get_smp_processor_id), \
3575 FN(skb_store_bytes), \
3576 FN(l3_csum_replace), \
3577 FN(l4_csum_replace), \
3579 FN(clone_redirect), \
3580 FN(get_current_pid_tgid), \
3581 FN(get_current_uid_gid), \
3582 FN(get_current_comm), \
3583 FN(get_cgroup_classid), \
3584 FN(skb_vlan_push), \
3586 FN(skb_get_tunnel_key), \
3587 FN(skb_set_tunnel_key), \
3588 FN(perf_event_read), \
3590 FN(get_route_realm), \
3591 FN(perf_event_output), \
3592 FN(skb_load_bytes), \
3595 FN(skb_get_tunnel_opt), \
3596 FN(skb_set_tunnel_opt), \
3597 FN(skb_change_proto), \
3598 FN(skb_change_type), \
3599 FN(skb_under_cgroup), \
3600 FN(get_hash_recalc), \
3601 FN(get_current_task), \
3602 FN(probe_write_user), \
3603 FN(current_task_under_cgroup), \
3604 FN(skb_change_tail), \
3605 FN(skb_pull_data), \
3607 FN(set_hash_invalid), \
3608 FN(get_numa_node_id), \
3609 FN(skb_change_head), \
3610 FN(xdp_adjust_head), \
3611 FN(probe_read_str), \
3612 FN(get_socket_cookie), \
3613 FN(get_socket_uid), \
3616 FN(skb_adjust_room), \
3618 FN(sk_redirect_map), \
3619 FN(sock_map_update), \
3620 FN(xdp_adjust_meta), \
3621 FN(perf_event_read_value), \
3622 FN(perf_prog_read_value), \
3624 FN(override_return), \
3625 FN(sock_ops_cb_flags_set), \
3626 FN(msg_redirect_map), \
3627 FN(msg_apply_bytes), \
3628 FN(msg_cork_bytes), \
3629 FN(msg_pull_data), \
3631 FN(xdp_adjust_tail), \
3632 FN(skb_get_xfrm_state), \
3634 FN(skb_load_bytes_relative), \
3636 FN(sock_hash_update), \
3637 FN(msg_redirect_hash), \
3638 FN(sk_redirect_hash), \
3639 FN(lwt_push_encap), \
3640 FN(lwt_seg6_store_bytes), \
3641 FN(lwt_seg6_adjust_srh), \
3642 FN(lwt_seg6_action), \
3645 FN(skb_cgroup_id), \
3646 FN(get_current_cgroup_id), \
3647 FN(get_local_storage), \
3648 FN(sk_select_reuseport), \
3649 FN(skb_ancestor_cgroup_id), \
3650 FN(sk_lookup_tcp), \
3651 FN(sk_lookup_udp), \
3653 FN(map_push_elem), \
3655 FN(map_peek_elem), \
3656 FN(msg_push_data), \
3658 FN(rc_pointer_rel), \
3663 FN(skb_ecn_set_ce), \
3664 FN(get_listener_sock), \
3665 FN(skc_lookup_tcp), \
3666 FN(tcp_check_syncookie), \
3667 FN(sysctl_get_name), \
3668 FN(sysctl_get_current_value), \
3669 FN(sysctl_get_new_value), \
3670 FN(sysctl_set_new_value), \
3673 FN(sk_storage_get), \
3674 FN(sk_storage_delete), \
3676 FN(tcp_gen_syncookie), \
3678 FN(probe_read_user), \
3679 FN(probe_read_kernel), \
3680 FN(probe_read_user_str), \
3681 FN(probe_read_kernel_str), \
3683 FN(send_signal_thread), \
3685 FN(read_branch_records), \
3686 FN(get_ns_current_pid_tgid), \
3688 FN(get_netns_cookie), \
3689 FN(get_current_ancestor_cgroup_id), \
3691 FN(ktime_get_boot_ns), \
3695 FN(sk_ancestor_cgroup_id), \
3696 FN(ringbuf_output), \
3697 FN(ringbuf_reserve), \
3698 FN(ringbuf_submit), \
3699 FN(ringbuf_discard), \
3700 FN(ringbuf_query), \
3702 FN(skc_to_tcp6_sock), \
3703 FN(skc_to_tcp_sock), \
3704 FN(skc_to_tcp_timewait_sock), \
3705 FN(skc_to_tcp_request_sock), \
3706 FN(skc_to_udp6_sock), \
3707 FN(get_task_stack), \
3709 FN(store_hdr_opt), \
3710 FN(reserve_hdr_opt), \
3711 FN(inode_storage_get), \
3712 FN(inode_storage_delete), \
3716 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
3717 * function eBPF program intends to call
3719 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
3721 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
3724 #undef __BPF_ENUM_FN
3726 /* All flags used by eBPF helper functions, placed here. */
3728 /* BPF_FUNC_skb_store_bytes flags. */
3730 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3731 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3734 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3735 * First 4 bits are for passing the header field size.
3738 BPF_F_HDR_FIELD_MASK = 0xfULL,
3741 /* BPF_FUNC_l4_csum_replace flags. */
3743 BPF_F_PSEUDO_HDR = (1ULL << 4),
3744 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3745 BPF_F_MARK_ENFORCE = (1ULL << 6),
3748 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
3750 BPF_F_INGRESS = (1ULL << 0),
3753 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
3755 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3758 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
3760 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3761 BPF_F_USER_STACK = (1ULL << 8),
3762 /* flags used by BPF_FUNC_get_stackid only. */
3763 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3764 BPF_F_REUSE_STACKID = (1ULL << 10),
3765 /* flags used by BPF_FUNC_get_stack only. */
3766 BPF_F_USER_BUILD_ID = (1ULL << 11),
3769 /* BPF_FUNC_skb_set_tunnel_key flags. */
3771 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3772 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3773 BPF_F_SEQ_NUMBER = (1ULL << 3),
3776 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3777 * BPF_FUNC_perf_event_read_value flags.
3780 BPF_F_INDEX_MASK = 0xffffffffULL,
3781 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
3782 /* BPF_FUNC_perf_event_output for sk_buff input context. */
3783 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3786 /* Current network namespace */
3788 BPF_F_CURRENT_NETNS = (-1L),
3791 /* BPF_FUNC_csum_level level values. */
3793 BPF_CSUM_LEVEL_QUERY,
3796 BPF_CSUM_LEVEL_RESET,
3799 /* BPF_FUNC_skb_adjust_room flags. */
3801 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3802 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3803 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3804 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3805 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
3806 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
3810 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3811 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3814 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
3815 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3816 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
3818 /* BPF_FUNC_sysctl_get_name flags. */
3820 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3823 /* BPF_FUNC_<kernel_obj>_storage_get flags */
3825 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
3826 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
3827 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
3829 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
3832 /* BPF_FUNC_read_branch_records flags. */
3834 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3837 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
3838 * BPF_FUNC_bpf_ringbuf_output flags.
3841 BPF_RB_NO_WAKEUP = (1ULL << 0),
3842 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
3845 /* BPF_FUNC_bpf_ringbuf_query flags */
3847 BPF_RB_AVAIL_DATA = 0,
3848 BPF_RB_RING_SIZE = 1,
3849 BPF_RB_CONS_POS = 2,
3850 BPF_RB_PROD_POS = 3,
3853 /* BPF ring buffer constants */
3855 BPF_RINGBUF_BUSY_BIT = (1U << 31),
3856 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
3857 BPF_RINGBUF_HDR_SZ = 8,
3860 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
3862 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
3863 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
3866 /* Mode for BPF_FUNC_skb_adjust_room helper. */
3867 enum bpf_adj_room_mode {
3872 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3873 enum bpf_hdr_start_off {
3878 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3879 enum bpf_lwt_encap_mode {
3881 BPF_LWT_ENCAP_SEG6_INLINE,
3885 #define __bpf_md_ptr(type, name) \
3889 } __attribute__((aligned(8)))
3891 /* user accessible mirror of in-kernel sk_buff.
3892 * new fields can only be added to the end of this structure
3898 __u32 queue_mapping;
3904 __u32 ingress_ifindex;
3914 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
3916 __u32 remote_ip4; /* Stored in network byte order */
3917 __u32 local_ip4; /* Stored in network byte order */
3918 __u32 remote_ip6[4]; /* Stored in network byte order */
3919 __u32 local_ip6[4]; /* Stored in network byte order */
3920 __u32 remote_port; /* Stored in network byte order */
3921 __u32 local_port; /* stored in host byte order */
3925 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
3929 __bpf_md_ptr(struct bpf_sock *, sk);
3933 struct bpf_tunnel_key {
3937 __u32 remote_ipv6[4];
3941 __u16 tunnel_ext; /* Padding, future use. */
3945 /* user accessible mirror of in-kernel xfrm_state.
3946 * new fields can only be added to the end of this structure
3948 struct bpf_xfrm_state {
3950 __u32 spi; /* Stored in network byte order */
3952 __u16 ext; /* Padding, future use. */
3954 __u32 remote_ipv4; /* Stored in network byte order */
3955 __u32 remote_ipv6[4]; /* Stored in network byte order */
3959 /* Generic BPF return codes which all BPF program types may support.
3960 * The values are binary compatible with their TC_ACT_* counter-part to
3961 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3964 * XDP is handled seprately, see XDP_*.
3972 /* >127 are reserved for prog type specific return codes.
3974 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3975 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3976 * changed and should be routed based on its new L3 header.
3977 * (This is an L3 redirect, as opposed to L2 redirect
3978 * represented by BPF_REDIRECT above).
3980 BPF_LWT_REROUTE = 128,
3990 /* IP address also allows 1 and 2 bytes access */
3993 __u32 src_port; /* host byte order */
3994 __u32 dst_port; /* network byte order */
3998 __s32 rx_queue_mapping;
4001 struct bpf_tcp_sock {
4002 __u32 snd_cwnd; /* Sending congestion window */
4003 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
4005 __u32 snd_ssthresh; /* Slow start size threshold */
4006 __u32 rcv_nxt; /* What we want to receive next */
4007 __u32 snd_nxt; /* Next sequence we send */
4008 __u32 snd_una; /* First byte we want an ack for */
4009 __u32 mss_cache; /* Cached effective mss, not including SACKS */
4010 __u32 ecn_flags; /* ECN status bits. */
4011 __u32 rate_delivered; /* saved rate sample: packets delivered */
4012 __u32 rate_interval_us; /* saved rate sample: time elapsed */
4013 __u32 packets_out; /* Packets which are "in flight" */
4014 __u32 retrans_out; /* Retransmitted packets out */
4015 __u32 total_retrans; /* Total retransmits for entire connection */
4016 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
4017 * total number of segments in.
4019 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
4020 * total number of data segments in.
4022 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
4023 * The total number of segments sent.
4025 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
4026 * total number of data segments sent.
4028 __u32 lost_out; /* Lost packets */
4029 __u32 sacked_out; /* SACK'd packets */
4030 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
4031 * sum(delta(rcv_nxt)), or how many bytes
4034 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
4035 * sum(delta(snd_una)), or how many bytes
4038 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
4039 * total number of DSACK blocks received
4041 __u32 delivered; /* Total data packets delivered incl. rexmits */
4042 __u32 delivered_ce; /* Like the above but only ECE marked packets */
4043 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
4046 struct bpf_sock_tuple {
4063 struct bpf_xdp_sock {
4067 #define XDP_PACKET_HEADROOM 256
4069 /* User return codes for XDP prog type.
4070 * A valid XDP program must return one of these defined values. All other
4071 * return codes are reserved for future use. Unknown return codes will
4072 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
4082 /* user accessible metadata for XDP packet hook
4083 * new fields must be added to the end of this structure
4089 /* Below access go through struct xdp_rxq_info */
4090 __u32 ingress_ifindex; /* rxq->dev->ifindex */
4091 __u32 rx_queue_index; /* rxq->queue_index */
4093 __u32 egress_ifindex; /* txq->dev->ifindex */
4096 /* DEVMAP map-value layout
4098 * The struct data-layout of map-value is a configuration interface.
4099 * New members can only be added to the end of this structure.
4101 struct bpf_devmap_val {
4102 __u32 ifindex; /* device index */
4104 int fd; /* prog fd on map write */
4105 __u32 id; /* prog id on map read */
4109 /* CPUMAP map-value layout
4111 * The struct data-layout of map-value is a configuration interface.
4112 * New members can only be added to the end of this structure.
4114 struct bpf_cpumap_val {
4115 __u32 qsize; /* queue size to remote target CPU */
4117 int fd; /* prog fd on map write */
4118 __u32 id; /* prog id on map read */
4127 /* user accessible metadata for SK_MSG packet hook, new fields must
4128 * be added to the end of this structure
4131 __bpf_md_ptr(void *, data);
4132 __bpf_md_ptr(void *, data_end);
4135 __u32 remote_ip4; /* Stored in network byte order */
4136 __u32 local_ip4; /* Stored in network byte order */
4137 __u32 remote_ip6[4]; /* Stored in network byte order */
4138 __u32 local_ip6[4]; /* Stored in network byte order */
4139 __u32 remote_port; /* Stored in network byte order */
4140 __u32 local_port; /* stored in host byte order */
4141 __u32 size; /* Total size of sk_msg */
4143 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
4146 struct sk_reuseport_md {
4148 * Start of directly accessible data. It begins from
4149 * the tcp/udp header.
4151 __bpf_md_ptr(void *, data);
4152 /* End of directly accessible data */
4153 __bpf_md_ptr(void *, data_end);
4155 * Total length of packet (starting from the tcp/udp header).
4156 * Note that the directly accessible bytes (data_end - data)
4157 * could be less than this "len". Those bytes could be
4158 * indirectly read by a helper "bpf_skb_load_bytes()".
4162 * Eth protocol in the mac header (network byte order). e.g.
4163 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
4166 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
4167 __u32 bind_inany; /* Is sock bound to an INANY address? */
4168 __u32 hash; /* A hash of the packet 4 tuples */
4171 #define BPF_TAG_SIZE 8
4173 struct bpf_prog_info {
4176 __u8 tag[BPF_TAG_SIZE];
4177 __u32 jited_prog_len;
4178 __u32 xlated_prog_len;
4179 __aligned_u64 jited_prog_insns;
4180 __aligned_u64 xlated_prog_insns;
4181 __u64 load_time; /* ns since boottime */
4182 __u32 created_by_uid;
4184 __aligned_u64 map_ids;
4185 char name[BPF_OBJ_NAME_LEN];
4187 __u32 gpl_compatible:1;
4188 __u32 :31; /* alignment pad */
4191 __u32 nr_jited_ksyms;
4192 __u32 nr_jited_func_lens;
4193 __aligned_u64 jited_ksyms;
4194 __aligned_u64 jited_func_lens;
4196 __u32 func_info_rec_size;
4197 __aligned_u64 func_info;
4200 __aligned_u64 line_info;
4201 __aligned_u64 jited_line_info;
4202 __u32 nr_jited_line_info;
4203 __u32 line_info_rec_size;
4204 __u32 jited_line_info_rec_size;
4206 __aligned_u64 prog_tags;
4209 } __attribute__((aligned(8)));
4211 struct bpf_map_info {
4218 char name[BPF_OBJ_NAME_LEN];
4220 __u32 btf_vmlinux_value_type_id;
4224 __u32 btf_key_type_id;
4225 __u32 btf_value_type_id;
4226 } __attribute__((aligned(8)));
4228 struct bpf_btf_info {
4232 } __attribute__((aligned(8)));
4234 struct bpf_link_info {
4240 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
4241 __u32 tp_name_len; /* in/out: tp_name buffer len */
4251 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
4252 __u32 target_name_len; /* in/out: target_name buffer len */
4265 } __attribute__((aligned(8)));
4267 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
4268 * by user and intended to be used by socket (e.g. to bind to, depends on
4271 struct bpf_sock_addr {
4272 __u32 user_family; /* Allows 4-byte read, but no write. */
4273 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4274 * Stored in network byte order.
4276 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4277 * Stored in network byte order.
4279 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
4280 * Stored in network byte order
4282 __u32 family; /* Allows 4-byte read, but no write */
4283 __u32 type; /* Allows 4-byte read, but no write */
4284 __u32 protocol; /* Allows 4-byte read, but no write */
4285 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4286 * Stored in network byte order.
4288 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4289 * Stored in network byte order.
4291 __bpf_md_ptr(struct bpf_sock *, sk);
4294 /* User bpf_sock_ops struct to access socket values and specify request ops
4295 * and their replies.
4296 * Some of this fields are in network (bigendian) byte order and may need
4297 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
4298 * New fields can only be added at the end of this structure
4300 struct bpf_sock_ops {
4303 __u32 args[4]; /* Optionally passed to bpf program */
4304 __u32 reply; /* Returned by bpf program */
4305 __u32 replylong[4]; /* Optionally returned by bpf prog */
4308 __u32 remote_ip4; /* Stored in network byte order */
4309 __u32 local_ip4; /* Stored in network byte order */
4310 __u32 remote_ip6[4]; /* Stored in network byte order */
4311 __u32 local_ip6[4]; /* Stored in network byte order */
4312 __u32 remote_port; /* Stored in network byte order */
4313 __u32 local_port; /* stored in host byte order */
4314 __u32 is_fullsock; /* Some TCP fields are only valid if
4315 * there is a full socket. If not, the
4316 * fields read as zero.
4319 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
4320 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
4329 __u32 rate_delivered;
4330 __u32 rate_interval_us;
4333 __u32 total_retrans;
4337 __u32 data_segs_out;
4341 __u64 bytes_received;
4343 __bpf_md_ptr(struct bpf_sock *, sk);
4344 /* [skb_data, skb_data_end) covers the whole TCP header.
4346 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
4347 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
4348 * header has not been written.
4349 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
4350 * been written so far.
4351 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
4353 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
4356 * bpf_load_hdr_opt() can also be used to read a particular option.
4358 __bpf_md_ptr(void *, skb_data);
4359 __bpf_md_ptr(void *, skb_data_end);
4360 __u32 skb_len; /* The total length of a packet.
4361 * It includes the header, options,
4364 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
4365 * an easy way to check for tcp_flags
4366 * without parsing skb_data.
4368 * In particular, the skb_tcp_flags
4369 * will still be available in
4370 * BPF_SOCK_OPS_HDR_OPT_LEN even though
4371 * the outgoing header has not
4376 /* Definitions for bpf_sock_ops_cb_flags */
4378 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
4379 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
4380 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
4381 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
4382 /* Call bpf for all received TCP headers. The bpf prog will be
4383 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4385 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4386 * for the header option related helpers that will be useful
4387 * to the bpf programs.
4389 * It could be used at the client/active side (i.e. connect() side)
4390 * when the server told it that the server was in syncookie
4391 * mode and required the active side to resend the bpf-written
4392 * options. The active side can keep writing the bpf-options until
4393 * it received a valid packet from the server side to confirm
4394 * the earlier packet (and options) has been received. The later
4395 * example patch is using it like this at the active side when the
4396 * server is in syncookie mode.
4398 * The bpf prog will usually turn this off in the common cases.
4400 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
4401 /* Call bpf when kernel has received a header option that
4402 * the kernel cannot handle. The bpf prog will be called under
4403 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
4405 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4406 * for the header option related helpers that will be useful
4407 * to the bpf programs.
4409 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
4410 /* Call bpf when the kernel is writing header options for the
4411 * outgoing packet. The bpf prog will first be called
4412 * to reserve space in a skb under
4413 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
4414 * the bpf prog will be called to write the header option(s)
4415 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4417 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
4418 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
4419 * related helpers that will be useful to the bpf programs.
4421 * The kernel gets its chance to reserve space and write
4422 * options first before the BPF program does.
4424 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
4425 /* Mask of all currently supported cb flags */
4426 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
4429 /* List of known BPF sock_ops operators.
4430 * New entries can only be added at the end
4434 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
4435 * -1 if default value should be used
4437 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
4438 * window (in packets) or -1 if default
4439 * value should be used
4441 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
4442 * active connection is initialized
4444 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
4445 * active connection is
4448 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
4449 * passive connection is
4452 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
4455 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
4456 * based on the path and may be
4457 * dependent on the congestion control
4458 * algorithm. In general it indicates
4459 * a congestion threshold. RTTs above
4460 * this indicate congestion
4462 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
4463 * Arg1: value of icsk_retransmits
4464 * Arg2: value of icsk_rto
4465 * Arg3: whether RTO has expired
4467 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
4468 * Arg1: sequence number of 1st byte
4470 * Arg3: return value of
4471 * tcp_transmit_skb (0 => success)
4473 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
4477 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
4478 * socket transition to LISTEN state.
4480 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
4482 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
4483 * It will be called to handle
4484 * the packets received at
4485 * an already established
4488 * sock_ops->skb_data:
4489 * Referring to the received skb.
4490 * It covers the TCP header only.
4492 * bpf_load_hdr_opt() can also
4493 * be used to search for a
4494 * particular option.
4496 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
4497 * header option later in
4498 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4499 * Arg1: bool want_cookie. (in
4500 * writing SYNACK only)
4502 * sock_ops->skb_data:
4503 * Not available because no header has
4506 * sock_ops->skb_tcp_flags:
4507 * The tcp_flags of the
4508 * outgoing skb. (e.g. SYN, ACK, FIN).
4510 * bpf_reserve_hdr_opt() should
4511 * be used to reserve space.
4513 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
4514 * Arg1: bool want_cookie. (in
4515 * writing SYNACK only)
4517 * sock_ops->skb_data:
4518 * Referring to the outgoing skb.
4519 * It covers the TCP header
4520 * that has already been written
4521 * by the kernel and the
4522 * earlier bpf-progs.
4524 * sock_ops->skb_tcp_flags:
4525 * The tcp_flags of the outgoing
4526 * skb. (e.g. SYN, ACK, FIN).
4528 * bpf_store_hdr_opt() should
4529 * be used to write the
4532 * bpf_load_hdr_opt() can also
4533 * be used to search for a
4534 * particular option that
4535 * has already been written
4536 * by the kernel or the
4537 * earlier bpf-progs.
4541 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
4542 * changes between the TCP and BPF versions. Ideally this should never happen.
4543 * If it does, we need to add code to convert them before calling
4544 * the BPF sock_ops function.
4547 BPF_TCP_ESTABLISHED = 1,
4557 BPF_TCP_CLOSING, /* Now a valid state */
4558 BPF_TCP_NEW_SYN_RECV,
4560 BPF_TCP_MAX_STATES /* Leave at the end! */
4564 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4565 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
4566 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
4567 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
4568 /* Copy the SYN pkt to optval
4570 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
4571 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
4572 * to only getting from the saved_syn. It can either get the
4575 * 1. the just-received SYN packet (only available when writing the
4576 * SYNACK). It will be useful when it is not necessary to
4577 * save the SYN packet for latter use. It is also the only way
4578 * to get the SYN during syncookie mode because the syn
4579 * packet cannot be saved during syncookie.
4583 * 2. the earlier saved syn which was done by
4584 * bpf_setsockopt(TCP_SAVE_SYN).
4586 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
4587 * SYN packet is obtained.
4589 * If the bpf-prog does not need the IP[46] header, the
4590 * bpf-prog can avoid parsing the IP header by using
4591 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
4592 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
4594 * >0: Total number of bytes copied
4595 * -ENOSPC: Not enough space in optval. Only optlen number of
4597 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
4598 * is not saved by setsockopt(TCP_SAVE_SYN).
4600 TCP_BPF_SYN = 1005, /* Copy the TCP header */
4601 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
4602 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
4606 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
4609 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
4610 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4613 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
4614 * total option spaces
4615 * required for an established
4616 * sk in order to calculate the
4617 * MSS. No skb is actually
4620 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
4621 * when sending a SYN.
4625 struct bpf_perf_event_value {
4632 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
4633 BPF_DEVCG_ACC_READ = (1ULL << 1),
4634 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
4638 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
4639 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
4642 struct bpf_cgroup_dev_ctx {
4643 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
4649 struct bpf_raw_tracepoint_args {
4653 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
4654 * OUTPUT: Do lookup from egress perspective; default is ingress
4657 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
4658 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
4662 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
4663 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
4664 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
4665 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
4666 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
4667 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
4668 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
4669 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
4670 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
4673 struct bpf_fib_lookup {
4674 /* input: network family for lookup (AF_INET, AF_INET6)
4675 * output: network family of egress nexthop
4679 /* set if lookup is to consider L4 data - e.g., FIB rules */
4684 /* total length of packet from network header - used for MTU check */
4687 /* input: L3 device index for lookup
4688 * output: device index from FIB lookup
4693 /* inputs to lookup */
4694 __u8 tos; /* AF_INET */
4695 __be32 flowinfo; /* AF_INET6, flow_label + priority */
4697 /* output: metric of fib result (IPv4/IPv6 only) */
4703 __u32 ipv6_src[4]; /* in6_addr; network order */
4706 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
4707 * network header. output: bpf_fib_lookup sets to gateway address
4708 * if FIB lookup returns gateway route
4712 __u32 ipv6_dst[4]; /* in6_addr; network order */
4716 __be16 h_vlan_proto;
4718 __u8 smac[6]; /* ETH_ALEN */
4719 __u8 dmac[6]; /* ETH_ALEN */
4722 enum bpf_task_fd_type {
4723 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
4724 BPF_FD_TYPE_TRACEPOINT, /* tp name */
4725 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
4726 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
4727 BPF_FD_TYPE_UPROBE, /* filename + offset */
4728 BPF_FD_TYPE_URETPROBE, /* filename + offset */
4732 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
4733 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
4734 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
4737 struct bpf_flow_keys {
4740 __u16 addr_proto; /* ETH_P_* of valid addrs */
4754 __u32 ipv6_src[4]; /* in6_addr; network order */
4755 __u32 ipv6_dst[4]; /* in6_addr; network order */
4762 struct bpf_func_info {
4767 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
4768 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
4770 struct bpf_line_info {
4772 __u32 file_name_off;
4777 struct bpf_spin_lock {
4782 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
4783 * Allows 1,2,4-byte read, but no write.
4785 __u32 file_pos; /* Sysctl file position to read from, write to.
4786 * Allows 1,2,4-byte read an 4-byte write.
4790 struct bpf_sockopt {
4791 __bpf_md_ptr(struct bpf_sock *, sk);
4792 __bpf_md_ptr(void *, optval);
4793 __bpf_md_ptr(void *, optval_end);
4801 struct bpf_pidns_info {
4806 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
4807 struct bpf_sk_lookup {
4808 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
4810 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
4811 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
4812 __u32 remote_ip4; /* Network byte order */
4813 __u32 remote_ip6[4]; /* Network byte order */
4814 __u32 remote_port; /* Network byte order */
4815 __u32 local_ip4; /* Network byte order */
4816 __u32 local_ip6[4]; /* Network byte order */
4817 __u32 local_port; /* Host byte order */
4820 #endif /* _UAPI__LINUX_BPF_H__ */