1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_XADD 0xc0 /* exclusive add */
25 #define BPF_MOV 0xb0 /* mov reg to reg */
26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28 /* change endianness of a register */
29 #define BPF_END 0xd0 /* flags for endianness conversion: */
30 #define BPF_TO_LE 0x00 /* convert to little-endian */
31 #define BPF_TO_BE 0x08 /* convert to big-endian */
32 #define BPF_FROM_LE BPF_TO_LE
33 #define BPF_FROM_BE BPF_TO_BE
36 #define BPF_JNE 0x50 /* jump != */
37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
43 #define BPF_CALL 0x80 /* function call */
44 #define BPF_EXIT 0x90 /* function return */
46 /* Register numbers */
62 /* BPF has 10 general purpose 64-bit registers and stack frame. */
63 #define MAX_BPF_REG __MAX_BPF_REG
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74 struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data[0]; /* Arbitrary size */
79 struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
84 /* BPF syscall commands, see bpf(2) man-page for details. */
99 BPF_PROG_GET_FD_BY_ID,
100 BPF_MAP_GET_FD_BY_ID,
101 BPF_OBJ_GET_INFO_BY_FD,
103 BPF_RAW_TRACEPOINT_OPEN,
105 BPF_BTF_GET_FD_BY_ID,
107 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
110 BPF_MAP_LOOKUP_BATCH,
111 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
112 BPF_MAP_UPDATE_BATCH,
113 BPF_MAP_DELETE_BATCH,
116 BPF_LINK_GET_FD_BY_ID,
117 BPF_LINK_GET_NEXT_ID,
126 BPF_MAP_TYPE_PROG_ARRAY,
127 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
128 BPF_MAP_TYPE_PERCPU_HASH,
129 BPF_MAP_TYPE_PERCPU_ARRAY,
130 BPF_MAP_TYPE_STACK_TRACE,
131 BPF_MAP_TYPE_CGROUP_ARRAY,
132 BPF_MAP_TYPE_LRU_HASH,
133 BPF_MAP_TYPE_LRU_PERCPU_HASH,
134 BPF_MAP_TYPE_LPM_TRIE,
135 BPF_MAP_TYPE_ARRAY_OF_MAPS,
136 BPF_MAP_TYPE_HASH_OF_MAPS,
138 BPF_MAP_TYPE_SOCKMAP,
141 BPF_MAP_TYPE_SOCKHASH,
142 BPF_MAP_TYPE_CGROUP_STORAGE,
143 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
144 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
147 BPF_MAP_TYPE_SK_STORAGE,
148 BPF_MAP_TYPE_DEVMAP_HASH,
149 BPF_MAP_TYPE_STRUCT_OPS,
150 BPF_MAP_TYPE_RINGBUF,
153 /* Note that tracing related programs such as
154 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
155 * are not subject to a stable API since kernel internal data
156 * structures can change from release to release and may
157 * therefore break existing tracing BPF programs. Tracing BPF
158 * programs correspond to /a/ specific kernel which is to be
159 * analyzed, and not /a/ specific kernel /and/ all future ones.
162 BPF_PROG_TYPE_UNSPEC,
163 BPF_PROG_TYPE_SOCKET_FILTER,
164 BPF_PROG_TYPE_KPROBE,
165 BPF_PROG_TYPE_SCHED_CLS,
166 BPF_PROG_TYPE_SCHED_ACT,
167 BPF_PROG_TYPE_TRACEPOINT,
169 BPF_PROG_TYPE_PERF_EVENT,
170 BPF_PROG_TYPE_CGROUP_SKB,
171 BPF_PROG_TYPE_CGROUP_SOCK,
172 BPF_PROG_TYPE_LWT_IN,
173 BPF_PROG_TYPE_LWT_OUT,
174 BPF_PROG_TYPE_LWT_XMIT,
175 BPF_PROG_TYPE_SOCK_OPS,
176 BPF_PROG_TYPE_SK_SKB,
177 BPF_PROG_TYPE_CGROUP_DEVICE,
178 BPF_PROG_TYPE_SK_MSG,
179 BPF_PROG_TYPE_RAW_TRACEPOINT,
180 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
181 BPF_PROG_TYPE_LWT_SEG6LOCAL,
182 BPF_PROG_TYPE_LIRC_MODE2,
183 BPF_PROG_TYPE_SK_REUSEPORT,
184 BPF_PROG_TYPE_FLOW_DISSECTOR,
185 BPF_PROG_TYPE_CGROUP_SYSCTL,
186 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
187 BPF_PROG_TYPE_CGROUP_SOCKOPT,
188 BPF_PROG_TYPE_TRACING,
189 BPF_PROG_TYPE_STRUCT_OPS,
194 enum bpf_attach_type {
195 BPF_CGROUP_INET_INGRESS,
196 BPF_CGROUP_INET_EGRESS,
197 BPF_CGROUP_INET_SOCK_CREATE,
199 BPF_SK_SKB_STREAM_PARSER,
200 BPF_SK_SKB_STREAM_VERDICT,
203 BPF_CGROUP_INET4_BIND,
204 BPF_CGROUP_INET6_BIND,
205 BPF_CGROUP_INET4_CONNECT,
206 BPF_CGROUP_INET6_CONNECT,
207 BPF_CGROUP_INET4_POST_BIND,
208 BPF_CGROUP_INET6_POST_BIND,
209 BPF_CGROUP_UDP4_SENDMSG,
210 BPF_CGROUP_UDP6_SENDMSG,
214 BPF_CGROUP_UDP4_RECVMSG,
215 BPF_CGROUP_UDP6_RECVMSG,
216 BPF_CGROUP_GETSOCKOPT,
217 BPF_CGROUP_SETSOCKOPT,
224 BPF_CGROUP_INET4_GETPEERNAME,
225 BPF_CGROUP_INET6_GETPEERNAME,
226 BPF_CGROUP_INET4_GETSOCKNAME,
227 BPF_CGROUP_INET6_GETSOCKNAME,
229 __MAX_BPF_ATTACH_TYPE
232 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
235 BPF_LINK_TYPE_UNSPEC = 0,
236 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
237 BPF_LINK_TYPE_TRACING = 2,
238 BPF_LINK_TYPE_CGROUP = 3,
239 BPF_LINK_TYPE_ITER = 4,
240 BPF_LINK_TYPE_NETNS = 5,
245 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
247 * NONE(default): No further bpf programs allowed in the subtree.
249 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
250 * the program in this cgroup yields to sub-cgroup program.
252 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
253 * that cgroup program gets run in addition to the program in this cgroup.
255 * Only one program is allowed to be attached to a cgroup with
256 * NONE or BPF_F_ALLOW_OVERRIDE flag.
257 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
258 * release old program and attach the new one. Attach flags has to match.
260 * Multiple programs are allowed to be attached to a cgroup with
261 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
262 * (those that were attached first, run first)
263 * The programs of sub-cgroup are executed first, then programs of
264 * this cgroup and then programs of parent cgroup.
265 * When children program makes decision (like picking TCP CA or sock bind)
266 * parent program has a chance to override it.
268 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
269 * programs for a cgroup. Though it's possible to replace an old program at
270 * any position by also specifying BPF_F_REPLACE flag and position itself in
271 * replace_bpf_fd attribute. Old program at this position will be released.
273 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
274 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
276 * cgrp1 (MULTI progs A, B) ->
277 * cgrp2 (OVERRIDE prog C) ->
278 * cgrp3 (MULTI prog D) ->
279 * cgrp4 (OVERRIDE prog E) ->
280 * cgrp5 (NONE prog F)
281 * the event in cgrp5 triggers execution of F,D,A,B in that order.
282 * if prog F is detached, the execution is E,D,A,B
283 * if prog F and D are detached, the execution is E,A,B
284 * if prog F, E and D are detached, the execution is C,A,B
286 * All eligible programs are executed regardless of return code from
289 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
290 #define BPF_F_ALLOW_MULTI (1U << 1)
291 #define BPF_F_REPLACE (1U << 2)
293 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
294 * verifier will perform strict alignment checking as if the kernel
295 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
296 * and NET_IP_ALIGN defined to 2.
298 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
300 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
301 * verifier will allow any alignment whatsoever. On platforms
302 * with strict alignment requirements for loads ands stores (such
303 * as sparc and mips) the verifier validates that all loads and
304 * stores provably follow this requirement. This flag turns that
305 * checking and enforcement off.
307 * It is mostly used for testing when we want to validate the
308 * context and memory access aspects of the verifier, but because
309 * of an unaligned access the alignment check would trigger before
310 * the one we are interested in.
312 #define BPF_F_ANY_ALIGNMENT (1U << 1)
314 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
315 * Verifier does sub-register def/use analysis and identifies instructions whose
316 * def only matters for low 32-bit, high 32-bit is never referenced later
317 * through implicit zero extension. Therefore verifier notifies JIT back-ends
318 * that it is safe to ignore clearing high 32-bit for these instructions. This
319 * saves some back-ends a lot of code-gen. However such optimization is not
320 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
321 * hence hasn't used verifier's analysis result. But, we really want to have a
322 * way to be able to verify the correctness of the described optimization on
323 * x86_64 on which testsuites are frequently exercised.
325 * So, this flag is introduced. Once it is set, verifier will randomize high
326 * 32-bit for those instructions who has been identified as safe to ignore them.
327 * Then, if verifier is not doing correct analysis, such randomization will
328 * regress tests to expose bugs.
330 #define BPF_F_TEST_RND_HI32 (1U << 2)
332 /* The verifier internal test flag. Behavior is undefined */
333 #define BPF_F_TEST_STATE_FREQ (1U << 3)
335 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
338 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
339 * insn[0].imm: map fd map fd
340 * insn[1].imm: 0 offset into value
343 * ldimm64 rewrite: address of map address of map[0]+offset
344 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
346 #define BPF_PSEUDO_MAP_FD 1
347 #define BPF_PSEUDO_MAP_VALUE 2
349 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
350 * offset to another bpf function
352 #define BPF_PSEUDO_CALL 1
354 /* flags for BPF_MAP_UPDATE_ELEM command */
356 BPF_ANY = 0, /* create new element or update existing */
357 BPF_NOEXIST = 1, /* create new element if it didn't exist */
358 BPF_EXIST = 2, /* update existing element */
359 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
362 /* flags for BPF_MAP_CREATE command */
364 BPF_F_NO_PREALLOC = (1U << 0),
365 /* Instead of having one common LRU list in the
366 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
367 * which can scale and perform better.
368 * Note, the LRU nodes (including free nodes) cannot be moved
369 * across different LRU lists.
371 BPF_F_NO_COMMON_LRU = (1U << 1),
372 /* Specify numa node during map creation */
373 BPF_F_NUMA_NODE = (1U << 2),
375 /* Flags for accessing BPF object from syscall side. */
376 BPF_F_RDONLY = (1U << 3),
377 BPF_F_WRONLY = (1U << 4),
379 /* Flag for stack_map, store build_id+offset instead of pointer */
380 BPF_F_STACK_BUILD_ID = (1U << 5),
382 /* Zero-initialize hash function seed. This should only be used for testing. */
383 BPF_F_ZERO_SEED = (1U << 6),
385 /* Flags for accessing BPF object from program side. */
386 BPF_F_RDONLY_PROG = (1U << 7),
387 BPF_F_WRONLY_PROG = (1U << 8),
389 /* Clone map from listener for newly accepted socket */
390 BPF_F_CLONE = (1U << 9),
392 /* Enable memory-mapping BPF map */
393 BPF_F_MMAPABLE = (1U << 10),
396 /* Flags for BPF_PROG_QUERY. */
398 /* Query effective (directly attached + inherited from ancestor cgroups)
399 * programs that will be executed for events within a cgroup.
400 * attach_flags with this flag are returned only for directly attached programs.
402 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
404 /* type for BPF_ENABLE_STATS */
405 enum bpf_stats_type {
406 /* enabled run_time_ns and run_cnt */
407 BPF_STATS_RUN_TIME = 0,
410 enum bpf_stack_build_id_status {
411 /* user space need an empty entry to identify end of a trace */
412 BPF_STACK_BUILD_ID_EMPTY = 0,
413 /* with valid build_id and offset */
414 BPF_STACK_BUILD_ID_VALID = 1,
415 /* couldn't get build_id, fallback to ip */
416 BPF_STACK_BUILD_ID_IP = 2,
419 #define BPF_BUILD_ID_SIZE 20
420 struct bpf_stack_build_id {
422 unsigned char build_id[BPF_BUILD_ID_SIZE];
429 #define BPF_OBJ_NAME_LEN 16U
432 struct { /* anonymous struct used by BPF_MAP_CREATE command */
433 __u32 map_type; /* one of enum bpf_map_type */
434 __u32 key_size; /* size of key in bytes */
435 __u32 value_size; /* size of value in bytes */
436 __u32 max_entries; /* max number of entries in a map */
437 __u32 map_flags; /* BPF_MAP_CREATE related
438 * flags defined above.
440 __u32 inner_map_fd; /* fd pointing to the inner map */
441 __u32 numa_node; /* numa node (effective only if
442 * BPF_F_NUMA_NODE is set).
444 char map_name[BPF_OBJ_NAME_LEN];
445 __u32 map_ifindex; /* ifindex of netdev to create on */
446 __u32 btf_fd; /* fd pointing to a BTF type data */
447 __u32 btf_key_type_id; /* BTF type_id of the key */
448 __u32 btf_value_type_id; /* BTF type_id of the value */
449 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
450 * struct stored as the
455 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
460 __aligned_u64 next_key;
465 struct { /* struct used by BPF_MAP_*_BATCH commands */
466 __aligned_u64 in_batch; /* start batch,
467 * NULL to start from beginning
469 __aligned_u64 out_batch; /* output: next start batch */
471 __aligned_u64 values;
472 __u32 count; /* input/output:
473 * input: # of key/value
475 * output: # of filled elements
482 struct { /* anonymous struct used by BPF_PROG_LOAD command */
483 __u32 prog_type; /* one of enum bpf_prog_type */
486 __aligned_u64 license;
487 __u32 log_level; /* verbosity level of verifier */
488 __u32 log_size; /* size of user buffer */
489 __aligned_u64 log_buf; /* user supplied buffer */
490 __u32 kern_version; /* not used */
492 char prog_name[BPF_OBJ_NAME_LEN];
493 __u32 prog_ifindex; /* ifindex of netdev to prep for */
494 /* For some prog types expected attach type must be known at
495 * load time to verify attach type specific parts of prog
496 * (context accesses, allowed helpers, etc).
498 __u32 expected_attach_type;
499 __u32 prog_btf_fd; /* fd pointing to BTF type data */
500 __u32 func_info_rec_size; /* userspace bpf_func_info size */
501 __aligned_u64 func_info; /* func info */
502 __u32 func_info_cnt; /* number of bpf_func_info records */
503 __u32 line_info_rec_size; /* userspace bpf_line_info size */
504 __aligned_u64 line_info; /* line info */
505 __u32 line_info_cnt; /* number of bpf_line_info records */
506 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
507 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
510 struct { /* anonymous struct used by BPF_OBJ_* commands */
511 __aligned_u64 pathname;
516 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
517 __u32 target_fd; /* container object to attach to */
518 __u32 attach_bpf_fd; /* eBPF program to attach */
521 __u32 replace_bpf_fd; /* previously attached eBPF
522 * program to replace if
523 * BPF_F_REPLACE is used
527 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
530 __u32 data_size_in; /* input: len of data_in */
531 __u32 data_size_out; /* input/output: len of data_out
532 * returns ENOSPC if data_out
535 __aligned_u64 data_in;
536 __aligned_u64 data_out;
539 __u32 ctx_size_in; /* input: len of ctx_in */
540 __u32 ctx_size_out; /* input/output: len of ctx_out
541 * returns ENOSPC if ctx_out
544 __aligned_u64 ctx_in;
545 __aligned_u64 ctx_out;
548 struct { /* anonymous struct used by BPF_*_GET_*_ID */
560 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
566 struct { /* anonymous struct used by BPF_PROG_QUERY command */
567 __u32 target_fd; /* container object to query */
571 __aligned_u64 prog_ids;
575 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
580 struct { /* anonymous struct for BPF_BTF_LOAD */
582 __aligned_u64 btf_log_buf;
589 __u32 pid; /* input: pid */
590 __u32 fd; /* input: fd */
591 __u32 flags; /* input: flags */
592 __u32 buf_len; /* input/output: buf len */
593 __aligned_u64 buf; /* input/output:
594 * tp_name for tracepoint
596 * filename for uprobe
598 __u32 prog_id; /* output: prod_id */
599 __u32 fd_type; /* output: BPF_FD_TYPE_* */
600 __u64 probe_offset; /* output: probe_offset */
601 __u64 probe_addr; /* output: probe_addr */
604 struct { /* struct used by BPF_LINK_CREATE command */
605 __u32 prog_fd; /* eBPF program to attach */
606 __u32 target_fd; /* object to attach to */
607 __u32 attach_type; /* attach type */
608 __u32 flags; /* extra flags */
611 struct { /* struct used by BPF_LINK_UPDATE command */
612 __u32 link_fd; /* link fd */
613 /* new program fd to update link with */
615 __u32 flags; /* extra flags */
616 /* expected link's program fd; is specified only if
617 * BPF_F_REPLACE flag is set in flags */
621 struct { /* struct used by BPF_ENABLE_STATS command */
625 struct { /* struct used by BPF_ITER_CREATE command */
630 } __attribute__((aligned(8)));
632 /* The description below is an attempt at providing documentation to eBPF
633 * developers about the multiple available eBPF helper functions. It can be
634 * parsed and used to produce a manual page. The workflow is the following,
635 * and requires the rst2man utility:
637 * $ ./scripts/bpf_helpers_doc.py \
638 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
639 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
640 * $ man /tmp/bpf-helpers.7
642 * Note that in order to produce this external documentation, some RST
643 * formatting is used in the descriptions to get "bold" and "italics" in
644 * manual pages. Also note that the few trailing white spaces are
645 * intentional, removing them would break paragraphs for rst2man.
647 * Start of BPF helper function descriptions:
649 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
651 * Perform a lookup in *map* for an entry associated to *key*.
653 * Map value associated to *key*, or **NULL** if no entry was
656 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
658 * Add or update the value of the entry associated to *key* in
659 * *map* with *value*. *flags* is one of:
662 * The entry for *key* must not exist in the map.
664 * The entry for *key* must already exist in the map.
666 * No condition on the existence of the entry for *key*.
668 * Flag value **BPF_NOEXIST** cannot be used for maps of types
669 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
670 * elements always exist), the helper would return an error.
672 * 0 on success, or a negative error in case of failure.
674 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
676 * Delete entry with *key* from *map*.
678 * 0 on success, or a negative error in case of failure.
680 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
682 * For tracing programs, safely attempt to read *size* bytes from
683 * kernel space address *unsafe_ptr* and store the data in *dst*.
685 * Generally, use **bpf_probe_read_user**\ () or
686 * **bpf_probe_read_kernel**\ () instead.
688 * 0 on success, or a negative error in case of failure.
690 * u64 bpf_ktime_get_ns(void)
692 * Return the time elapsed since system boot, in nanoseconds.
693 * Does not include time the system was suspended.
694 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
698 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
700 * This helper is a "printk()-like" facility for debugging. It
701 * prints a message defined by format *fmt* (of size *fmt_size*)
702 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
703 * available. It can take up to three additional **u64**
704 * arguments (as an eBPF helpers, the total number of arguments is
707 * Each time the helper is called, it appends a line to the trace.
708 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
709 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
710 * The format of the trace is customizable, and the exact output
711 * one will get depends on the options set in
712 * *\/sys/kernel/debug/tracing/trace_options* (see also the
713 * *README* file under the same directory). However, it usually
714 * defaults to something like:
718 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
722 * * ``telnet`` is the name of the current task.
723 * * ``470`` is the PID of the current task.
724 * * ``001`` is the CPU number on which the task is
726 * * In ``.N..``, each character refers to a set of
727 * options (whether irqs are enabled, scheduling
728 * options, whether hard/softirqs are running, level of
729 * preempt_disabled respectively). **N** means that
730 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
732 * * ``419421.045894`` is a timestamp.
733 * * ``0x00000001`` is a fake value used by BPF for the
734 * instruction pointer register.
735 * * ``<formatted msg>`` is the message formatted with
738 * The conversion specifiers supported by *fmt* are similar, but
739 * more limited than for printk(). They are **%d**, **%i**,
740 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
741 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
742 * of field, padding with zeroes, etc.) is available, and the
743 * helper will return **-EINVAL** (but print nothing) if it
744 * encounters an unknown specifier.
746 * Also, note that **bpf_trace_printk**\ () is slow, and should
747 * only be used for debugging purposes. For this reason, a notice
748 * bloc (spanning several lines) is printed to kernel logs and
749 * states that the helper should not be used "for production use"
750 * the first time this helper is used (or more precisely, when
751 * **trace_printk**\ () buffers are allocated). For passing values
752 * to user space, perf events should be preferred.
754 * The number of bytes written to the buffer, or a negative error
755 * in case of failure.
757 * u32 bpf_get_prandom_u32(void)
759 * Get a pseudo-random number.
761 * From a security point of view, this helper uses its own
762 * pseudo-random internal state, and cannot be used to infer the
763 * seed of other random functions in the kernel. However, it is
764 * essential to note that the generator used by the helper is not
765 * cryptographically secure.
767 * A random 32-bit unsigned value.
769 * u32 bpf_get_smp_processor_id(void)
771 * Get the SMP (symmetric multiprocessing) processor id. Note that
772 * all programs run with preemption disabled, which means that the
773 * SMP processor id is stable during all the execution of the
776 * The SMP id of the processor running the program.
778 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
780 * Store *len* bytes from address *from* into the packet
781 * associated to *skb*, at *offset*. *flags* are a combination of
782 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
783 * checksum for the packet after storing the bytes) and
784 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
785 * **->swhash** and *skb*\ **->l4hash** to 0).
787 * A call to this helper is susceptible to change the underlying
788 * packet buffer. Therefore, at load time, all checks on pointers
789 * previously done by the verifier are invalidated and must be
790 * performed again, if the helper is used in combination with
791 * direct packet access.
793 * 0 on success, or a negative error in case of failure.
795 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
797 * Recompute the layer 3 (e.g. IP) checksum for the packet
798 * associated to *skb*. Computation is incremental, so the helper
799 * must know the former value of the header field that was
800 * modified (*from*), the new value of this field (*to*), and the
801 * number of bytes (2 or 4) for this field, stored in *size*.
802 * Alternatively, it is possible to store the difference between
803 * the previous and the new values of the header field in *to*, by
804 * setting *from* and *size* to 0. For both methods, *offset*
805 * indicates the location of the IP checksum within the packet.
807 * This helper works in combination with **bpf_csum_diff**\ (),
808 * which does not update the checksum in-place, but offers more
809 * flexibility and can handle sizes larger than 2 or 4 for the
810 * checksum to update.
812 * A call to this helper is susceptible to change the underlying
813 * packet buffer. Therefore, at load time, all checks on pointers
814 * previously done by the verifier are invalidated and must be
815 * performed again, if the helper is used in combination with
816 * direct packet access.
818 * 0 on success, or a negative error in case of failure.
820 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
822 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
823 * packet associated to *skb*. Computation is incremental, so the
824 * helper must know the former value of the header field that was
825 * modified (*from*), the new value of this field (*to*), and the
826 * number of bytes (2 or 4) for this field, stored on the lowest
827 * four bits of *flags*. Alternatively, it is possible to store
828 * the difference between the previous and the new values of the
829 * header field in *to*, by setting *from* and the four lowest
830 * bits of *flags* to 0. For both methods, *offset* indicates the
831 * location of the IP checksum within the packet. In addition to
832 * the size of the field, *flags* can be added (bitwise OR) actual
833 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
834 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
835 * for updates resulting in a null checksum the value is set to
836 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
837 * the checksum is to be computed against a pseudo-header.
839 * This helper works in combination with **bpf_csum_diff**\ (),
840 * which does not update the checksum in-place, but offers more
841 * flexibility and can handle sizes larger than 2 or 4 for the
842 * checksum to update.
844 * A call to this helper is susceptible to change the underlying
845 * packet buffer. Therefore, at load time, all checks on pointers
846 * previously done by the verifier are invalidated and must be
847 * performed again, if the helper is used in combination with
848 * direct packet access.
850 * 0 on success, or a negative error in case of failure.
852 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
854 * This special helper is used to trigger a "tail call", or in
855 * other words, to jump into another eBPF program. The same stack
856 * frame is used (but values on stack and in registers for the
857 * caller are not accessible to the callee). This mechanism allows
858 * for program chaining, either for raising the maximum number of
859 * available eBPF instructions, or to execute given programs in
860 * conditional blocks. For security reasons, there is an upper
861 * limit to the number of successive tail calls that can be
864 * Upon call of this helper, the program attempts to jump into a
865 * program referenced at index *index* in *prog_array_map*, a
866 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
867 * *ctx*, a pointer to the context.
869 * If the call succeeds, the kernel immediately runs the first
870 * instruction of the new program. This is not a function call,
871 * and it never returns to the previous program. If the call
872 * fails, then the helper has no effect, and the caller continues
873 * to run its subsequent instructions. A call can fail if the
874 * destination program for the jump does not exist (i.e. *index*
875 * is superior to the number of entries in *prog_array_map*), or
876 * if the maximum number of tail calls has been reached for this
877 * chain of programs. This limit is defined in the kernel by the
878 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
879 * which is currently set to 32.
881 * 0 on success, or a negative error in case of failure.
883 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
885 * Clone and redirect the packet associated to *skb* to another
886 * net device of index *ifindex*. Both ingress and egress
887 * interfaces can be used for redirection. The **BPF_F_INGRESS**
888 * value in *flags* is used to make the distinction (ingress path
889 * is selected if the flag is present, egress path otherwise).
890 * This is the only flag supported for now.
892 * In comparison with **bpf_redirect**\ () helper,
893 * **bpf_clone_redirect**\ () has the associated cost of
894 * duplicating the packet buffer, but this can be executed out of
895 * the eBPF program. Conversely, **bpf_redirect**\ () is more
896 * efficient, but it is handled through an action code where the
897 * redirection happens only after the eBPF program has returned.
899 * A call to this helper is susceptible to change the underlying
900 * packet buffer. Therefore, at load time, all checks on pointers
901 * previously done by the verifier are invalidated and must be
902 * performed again, if the helper is used in combination with
903 * direct packet access.
905 * 0 on success, or a negative error in case of failure.
907 * u64 bpf_get_current_pid_tgid(void)
909 * A 64-bit integer containing the current tgid and pid, and
911 * *current_task*\ **->tgid << 32 \|**
912 * *current_task*\ **->pid**.
914 * u64 bpf_get_current_uid_gid(void)
916 * A 64-bit integer containing the current GID and UID, and
917 * created as such: *current_gid* **<< 32 \|** *current_uid*.
919 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
921 * Copy the **comm** attribute of the current task into *buf* of
922 * *size_of_buf*. The **comm** attribute contains the name of
923 * the executable (excluding the path) for the current task. The
924 * *size_of_buf* must be strictly positive. On success, the
925 * helper makes sure that the *buf* is NUL-terminated. On failure,
926 * it is filled with zeroes.
928 * 0 on success, or a negative error in case of failure.
930 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
932 * Retrieve the classid for the current task, i.e. for the net_cls
933 * cgroup to which *skb* belongs.
935 * This helper can be used on TC egress path, but not on ingress.
937 * The net_cls cgroup provides an interface to tag network packets
938 * based on a user-provided identifier for all traffic coming from
939 * the tasks belonging to the related cgroup. See also the related
940 * kernel documentation, available from the Linux sources in file
941 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
943 * The Linux kernel has two versions for cgroups: there are
944 * cgroups v1 and cgroups v2. Both are available to users, who can
945 * use a mixture of them, but note that the net_cls cgroup is for
946 * cgroup v1 only. This makes it incompatible with BPF programs
947 * run on cgroups, which is a cgroup-v2-only feature (a socket can
948 * only hold data for one version of cgroups at a time).
950 * This helper is only available is the kernel was compiled with
951 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
952 * "**y**" or to "**m**".
954 * The classid, or 0 for the default unconfigured classid.
956 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
958 * Push a *vlan_tci* (VLAN tag control information) of protocol
959 * *vlan_proto* to the packet associated to *skb*, then update
960 * the checksum. Note that if *vlan_proto* is different from
961 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
962 * be **ETH_P_8021Q**.
964 * A call to this helper is susceptible to change the underlying
965 * packet buffer. Therefore, at load time, all checks on pointers
966 * previously done by the verifier are invalidated and must be
967 * performed again, if the helper is used in combination with
968 * direct packet access.
970 * 0 on success, or a negative error in case of failure.
972 * int bpf_skb_vlan_pop(struct sk_buff *skb)
974 * Pop a VLAN header from the packet associated to *skb*.
976 * A call to this helper is susceptible to change the underlying
977 * packet buffer. Therefore, at load time, all checks on pointers
978 * previously done by the verifier are invalidated and must be
979 * performed again, if the helper is used in combination with
980 * direct packet access.
982 * 0 on success, or a negative error in case of failure.
984 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
986 * Get tunnel metadata. This helper takes a pointer *key* to an
987 * empty **struct bpf_tunnel_key** of **size**, that will be
988 * filled with tunnel metadata for the packet associated to *skb*.
989 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
990 * indicates that the tunnel is based on IPv6 protocol instead of
993 * The **struct bpf_tunnel_key** is an object that generalizes the
994 * principal parameters used by various tunneling protocols into a
995 * single struct. This way, it can be used to easily make a
996 * decision based on the contents of the encapsulation header,
997 * "summarized" in this struct. In particular, it holds the IP
998 * address of the remote end (IPv4 or IPv6, depending on the case)
999 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1000 * this struct exposes the *key*\ **->tunnel_id**, which is
1001 * generally mapped to a VNI (Virtual Network Identifier), making
1002 * it programmable together with the **bpf_skb_set_tunnel_key**\
1005 * Let's imagine that the following code is part of a program
1006 * attached to the TC ingress interface, on one end of a GRE
1007 * tunnel, and is supposed to filter out all messages coming from
1008 * remote ends with IPv4 address other than 10.0.0.1:
1013 * struct bpf_tunnel_key key = {};
1015 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1017 * return TC_ACT_SHOT; // drop packet
1019 * if (key.remote_ipv4 != 0x0a000001)
1020 * return TC_ACT_SHOT; // drop packet
1022 * return TC_ACT_OK; // accept packet
1024 * This interface can also be used with all encapsulation devices
1025 * that can operate in "collect metadata" mode: instead of having
1026 * one network device per specific configuration, the "collect
1027 * metadata" mode only requires a single device where the
1028 * configuration can be extracted from this helper.
1030 * This can be used together with various tunnels such as VXLan,
1031 * Geneve, GRE or IP in IP (IPIP).
1033 * 0 on success, or a negative error in case of failure.
1035 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1037 * Populate tunnel metadata for packet associated to *skb.* The
1038 * tunnel metadata is set to the contents of *key*, of *size*. The
1039 * *flags* can be set to a combination of the following values:
1041 * **BPF_F_TUNINFO_IPV6**
1042 * Indicate that the tunnel is based on IPv6 protocol
1044 * **BPF_F_ZERO_CSUM_TX**
1045 * For IPv4 packets, add a flag to tunnel metadata
1046 * indicating that checksum computation should be skipped
1047 * and checksum set to zeroes.
1048 * **BPF_F_DONT_FRAGMENT**
1049 * Add a flag to tunnel metadata indicating that the
1050 * packet should not be fragmented.
1051 * **BPF_F_SEQ_NUMBER**
1052 * Add a flag to tunnel metadata indicating that a
1053 * sequence number should be added to tunnel header before
1054 * sending the packet. This flag was added for GRE
1055 * encapsulation, but might be used with other protocols
1056 * as well in the future.
1058 * Here is a typical usage on the transmit path:
1062 * struct bpf_tunnel_key key;
1064 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1065 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1067 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1068 * helper for additional information.
1070 * 0 on success, or a negative error in case of failure.
1072 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1074 * Read the value of a perf event counter. This helper relies on a
1075 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1076 * the perf event counter is selected when *map* is updated with
1077 * perf event file descriptors. The *map* is an array whose size
1078 * is the number of available CPUs, and each cell contains a value
1079 * relative to one CPU. The value to retrieve is indicated by
1080 * *flags*, that contains the index of the CPU to look up, masked
1081 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1082 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1083 * current CPU should be retrieved.
1085 * Note that before Linux 4.13, only hardware perf event can be
1088 * Also, be aware that the newer helper
1089 * **bpf_perf_event_read_value**\ () is recommended over
1090 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1091 * quirks where error and counter value are used as a return code
1092 * (which is wrong to do since ranges may overlap). This issue is
1093 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1094 * time provides more features over the **bpf_perf_event_read**\
1095 * () interface. Please refer to the description of
1096 * **bpf_perf_event_read_value**\ () for details.
1098 * The value of the perf event counter read from the map, or a
1099 * negative error code in case of failure.
1101 * int bpf_redirect(u32 ifindex, u64 flags)
1103 * Redirect the packet to another net device of index *ifindex*.
1104 * This helper is somewhat similar to **bpf_clone_redirect**\
1105 * (), except that the packet is not cloned, which provides
1106 * increased performance.
1108 * Except for XDP, both ingress and egress interfaces can be used
1109 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1110 * to make the distinction (ingress path is selected if the flag
1111 * is present, egress path otherwise). Currently, XDP only
1112 * supports redirection to the egress interface, and accepts no
1115 * The same effect can also be attained with the more generic
1116 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1117 * redirect target instead of providing it directly to the helper.
1119 * For XDP, the helper returns **XDP_REDIRECT** on success or
1120 * **XDP_ABORTED** on error. For other program types, the values
1121 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1124 * u32 bpf_get_route_realm(struct sk_buff *skb)
1126 * Retrieve the realm or the route, that is to say the
1127 * **tclassid** field of the destination for the *skb*. The
1128 * indentifier retrieved is a user-provided tag, similar to the
1129 * one used with the net_cls cgroup (see description for
1130 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1131 * held by a route (a destination entry), not by a task.
1133 * Retrieving this identifier works with the clsact TC egress hook
1134 * (see also **tc-bpf(8)**), or alternatively on conventional
1135 * classful egress qdiscs, but not on TC ingress path. In case of
1136 * clsact TC egress hook, this has the advantage that, internally,
1137 * the destination entry has not been dropped yet in the transmit
1138 * path. Therefore, the destination entry does not need to be
1139 * artificially held via **netif_keep_dst**\ () for a classful
1140 * qdisc until the *skb* is freed.
1142 * This helper is available only if the kernel was compiled with
1143 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1145 * The realm of the route for the packet associated to *skb*, or 0
1146 * if none was found.
1148 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1150 * Write raw *data* blob into a special BPF perf event held by
1151 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1152 * event must have the following attributes: **PERF_SAMPLE_RAW**
1153 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1154 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1156 * The *flags* are used to indicate the index in *map* for which
1157 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1158 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1159 * to indicate that the index of the current CPU core should be
1162 * The value to write, of *size*, is passed through eBPF stack and
1163 * pointed by *data*.
1165 * The context of the program *ctx* needs also be passed to the
1168 * On user space, a program willing to read the values needs to
1169 * call **perf_event_open**\ () on the perf event (either for
1170 * one or for all CPUs) and to store the file descriptor into the
1171 * *map*. This must be done before the eBPF program can send data
1172 * into it. An example is available in file
1173 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1174 * tree (the eBPF program counterpart is in
1175 * *samples/bpf/trace_output_kern.c*).
1177 * **bpf_perf_event_output**\ () achieves better performance
1178 * than **bpf_trace_printk**\ () for sharing data with user
1179 * space, and is much better suitable for streaming data from eBPF
1182 * Note that this helper is not restricted to tracing use cases
1183 * and can be used with programs attached to TC or XDP as well,
1184 * where it allows for passing data to user space listeners. Data
1187 * * Only custom structs,
1188 * * Only the packet payload, or
1189 * * A combination of both.
1191 * 0 on success, or a negative error in case of failure.
1193 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1195 * This helper was provided as an easy way to load data from a
1196 * packet. It can be used to load *len* bytes from *offset* from
1197 * the packet associated to *skb*, into the buffer pointed by
1200 * Since Linux 4.7, usage of this helper has mostly been replaced
1201 * by "direct packet access", enabling packet data to be
1202 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1203 * pointing respectively to the first byte of packet data and to
1204 * the byte after the last byte of packet data. However, it
1205 * remains useful if one wishes to read large quantities of data
1206 * at once from a packet into the eBPF stack.
1208 * 0 on success, or a negative error in case of failure.
1210 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1212 * Walk a user or a kernel stack and return its id. To achieve
1213 * this, the helper needs *ctx*, which is a pointer to the context
1214 * on which the tracing program is executed, and a pointer to a
1215 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1217 * The last argument, *flags*, holds the number of stack frames to
1218 * skip (from 0 to 255), masked with
1219 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1220 * a combination of the following flags:
1222 * **BPF_F_USER_STACK**
1223 * Collect a user space stack instead of a kernel stack.
1224 * **BPF_F_FAST_STACK_CMP**
1225 * Compare stacks by hash only.
1226 * **BPF_F_REUSE_STACKID**
1227 * If two different stacks hash into the same *stackid*,
1228 * discard the old one.
1230 * The stack id retrieved is a 32 bit long integer handle which
1231 * can be further combined with other data (including other stack
1232 * ids) and used as a key into maps. This can be useful for
1233 * generating a variety of graphs (such as flame graphs or off-cpu
1236 * For walking a stack, this helper is an improvement over
1237 * **bpf_probe_read**\ (), which can be used with unrolled loops
1238 * but is not efficient and consumes a lot of eBPF instructions.
1239 * Instead, **bpf_get_stackid**\ () can collect up to
1240 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1241 * this limit can be controlled with the **sysctl** program, and
1242 * that it should be manually increased in order to profile long
1243 * user stacks (such as stacks for Java programs). To do so, use:
1247 * # sysctl kernel.perf_event_max_stack=<new value>
1249 * The positive or null stack id on success, or a negative error
1250 * in case of failure.
1252 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1254 * Compute a checksum difference, from the raw buffer pointed by
1255 * *from*, of length *from_size* (that must be a multiple of 4),
1256 * towards the raw buffer pointed by *to*, of size *to_size*
1257 * (same remark). An optional *seed* can be added to the value
1258 * (this can be cascaded, the seed may come from a previous call
1261 * This is flexible enough to be used in several ways:
1263 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1264 * checksum, it can be used when pushing new data.
1265 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1266 * checksum, it can be used when removing data from a packet.
1267 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1268 * can be used to compute a diff. Note that *from_size* and
1269 * *to_size* do not need to be equal.
1271 * This helper can be used in combination with
1272 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1273 * which one can feed in the difference computed with
1274 * **bpf_csum_diff**\ ().
1276 * The checksum result, or a negative error code in case of
1279 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1281 * Retrieve tunnel options metadata for the packet associated to
1282 * *skb*, and store the raw tunnel option data to the buffer *opt*
1285 * This helper can be used with encapsulation devices that can
1286 * operate in "collect metadata" mode (please refer to the related
1287 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1288 * more details). A particular example where this can be used is
1289 * in combination with the Geneve encapsulation protocol, where it
1290 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1291 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1292 * the eBPF program. This allows for full customization of these
1295 * The size of the option data retrieved.
1297 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1299 * Set tunnel options metadata for the packet associated to *skb*
1300 * to the option data contained in the raw buffer *opt* of *size*.
1302 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1303 * helper for additional information.
1305 * 0 on success, or a negative error in case of failure.
1307 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1309 * Change the protocol of the *skb* to *proto*. Currently
1310 * supported are transition from IPv4 to IPv6, and from IPv6 to
1311 * IPv4. The helper takes care of the groundwork for the
1312 * transition, including resizing the socket buffer. The eBPF
1313 * program is expected to fill the new headers, if any, via
1314 * **skb_store_bytes**\ () and to recompute the checksums with
1315 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1316 * (). The main case for this helper is to perform NAT64
1317 * operations out of an eBPF program.
1319 * Internally, the GSO type is marked as dodgy so that headers are
1320 * checked and segments are recalculated by the GSO/GRO engine.
1321 * The size for GSO target is adapted as well.
1323 * All values for *flags* are reserved for future usage, and must
1326 * A call to this helper is susceptible to change the underlying
1327 * packet buffer. Therefore, at load time, all checks on pointers
1328 * previously done by the verifier are invalidated and must be
1329 * performed again, if the helper is used in combination with
1330 * direct packet access.
1332 * 0 on success, or a negative error in case of failure.
1334 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1336 * Change the packet type for the packet associated to *skb*. This
1337 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1338 * the eBPF program does not have a write access to *skb*\
1339 * **->pkt_type** beside this helper. Using a helper here allows
1340 * for graceful handling of errors.
1342 * The major use case is to change incoming *skb*s to
1343 * **PACKET_HOST** in a programmatic way instead of having to
1344 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1347 * Note that *type* only allows certain values. At this time, they
1352 * **PACKET_BROADCAST**
1353 * Send packet to all.
1354 * **PACKET_MULTICAST**
1355 * Send packet to group.
1356 * **PACKET_OTHERHOST**
1357 * Send packet to someone else.
1359 * 0 on success, or a negative error in case of failure.
1361 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1363 * Check whether *skb* is a descendant of the cgroup2 held by
1364 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1366 * The return value depends on the result of the test, and can be:
1368 * * 0, if the *skb* failed the cgroup2 descendant test.
1369 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1370 * * A negative error code, if an error occurred.
1372 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1374 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1375 * not set, in particular if the hash was cleared due to mangling,
1376 * recompute this hash. Later accesses to the hash can be done
1377 * directly with *skb*\ **->hash**.
1379 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1380 * prototype with **bpf_skb_change_proto**\ (), or calling
1381 * **bpf_skb_store_bytes**\ () with the
1382 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1383 * the hash and to trigger a new computation for the next call to
1384 * **bpf_get_hash_recalc**\ ().
1388 * u64 bpf_get_current_task(void)
1390 * A pointer to the current task struct.
1392 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1394 * Attempt in a safe way to write *len* bytes from the buffer
1395 * *src* to *dst* in memory. It only works for threads that are in
1396 * user context, and *dst* must be a valid user space address.
1398 * This helper should not be used to implement any kind of
1399 * security mechanism because of TOC-TOU attacks, but rather to
1400 * debug, divert, and manipulate execution of semi-cooperative
1403 * Keep in mind that this feature is meant for experiments, and it
1404 * has a risk of crashing the system and running programs.
1405 * Therefore, when an eBPF program using this helper is attached,
1406 * a warning including PID and process name is printed to kernel
1409 * 0 on success, or a negative error in case of failure.
1411 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1413 * Check whether the probe is being run is the context of a given
1414 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1415 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1417 * The return value depends on the result of the test, and can be:
1419 * * 0, if the *skb* task belongs to the cgroup2.
1420 * * 1, if the *skb* task does not belong to the cgroup2.
1421 * * A negative error code, if an error occurred.
1423 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1425 * Resize (trim or grow) the packet associated to *skb* to the
1426 * new *len*. The *flags* are reserved for future usage, and must
1429 * The basic idea is that the helper performs the needed work to
1430 * change the size of the packet, then the eBPF program rewrites
1431 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1432 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1433 * and others. This helper is a slow path utility intended for
1434 * replies with control messages. And because it is targeted for
1435 * slow path, the helper itself can afford to be slow: it
1436 * implicitly linearizes, unclones and drops offloads from the
1439 * A call to this helper is susceptible to change the underlying
1440 * packet buffer. Therefore, at load time, all checks on pointers
1441 * previously done by the verifier are invalidated and must be
1442 * performed again, if the helper is used in combination with
1443 * direct packet access.
1445 * 0 on success, or a negative error in case of failure.
1447 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1449 * Pull in non-linear data in case the *skb* is non-linear and not
1450 * all of *len* are part of the linear section. Make *len* bytes
1451 * from *skb* readable and writable. If a zero value is passed for
1452 * *len*, then the whole length of the *skb* is pulled.
1454 * This helper is only needed for reading and writing with direct
1457 * For direct packet access, testing that offsets to access
1458 * are within packet boundaries (test on *skb*\ **->data_end**) is
1459 * susceptible to fail if offsets are invalid, or if the requested
1460 * data is in non-linear parts of the *skb*. On failure the
1461 * program can just bail out, or in the case of a non-linear
1462 * buffer, use a helper to make the data available. The
1463 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1464 * the data. Another one consists in using **bpf_skb_pull_data**
1465 * to pull in once the non-linear parts, then retesting and
1466 * eventually access the data.
1468 * At the same time, this also makes sure the *skb* is uncloned,
1469 * which is a necessary condition for direct write. As this needs
1470 * to be an invariant for the write part only, the verifier
1471 * detects writes and adds a prologue that is calling
1472 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1473 * the very beginning in case it is indeed cloned.
1475 * A call to this helper is susceptible to change the underlying
1476 * packet buffer. Therefore, at load time, all checks on pointers
1477 * previously done by the verifier are invalidated and must be
1478 * performed again, if the helper is used in combination with
1479 * direct packet access.
1481 * 0 on success, or a negative error in case of failure.
1483 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1485 * Add the checksum *csum* into *skb*\ **->csum** in case the
1486 * driver has supplied a checksum for the entire packet into that
1487 * field. Return an error otherwise. This helper is intended to be
1488 * used in combination with **bpf_csum_diff**\ (), in particular
1489 * when the checksum needs to be updated after data has been
1490 * written into the packet through direct packet access.
1492 * The checksum on success, or a negative error code in case of
1495 * void bpf_set_hash_invalid(struct sk_buff *skb)
1497 * Invalidate the current *skb*\ **->hash**. It can be used after
1498 * mangling on headers through direct packet access, in order to
1499 * indicate that the hash is outdated and to trigger a
1500 * recalculation the next time the kernel tries to access this
1501 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1503 * int bpf_get_numa_node_id(void)
1505 * Return the id of the current NUMA node. The primary use case
1506 * for this helper is the selection of sockets for the local NUMA
1507 * node, when the program is attached to sockets using the
1508 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1509 * but the helper is also available to other eBPF program types,
1510 * similarly to **bpf_get_smp_processor_id**\ ().
1512 * The id of current NUMA node.
1514 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1516 * Grows headroom of packet associated to *skb* and adjusts the
1517 * offset of the MAC header accordingly, adding *len* bytes of
1518 * space. It automatically extends and reallocates memory as
1521 * This helper can be used on a layer 3 *skb* to push a MAC header
1522 * for redirection into a layer 2 device.
1524 * All values for *flags* are reserved for future usage, and must
1527 * A call to this helper is susceptible to change the underlying
1528 * packet buffer. Therefore, at load time, all checks on pointers
1529 * previously done by the verifier are invalidated and must be
1530 * performed again, if the helper is used in combination with
1531 * direct packet access.
1533 * 0 on success, or a negative error in case of failure.
1535 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1537 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1538 * it is possible to use a negative value for *delta*. This helper
1539 * can be used to prepare the packet for pushing or popping
1542 * A call to this helper is susceptible to change the underlying
1543 * packet buffer. Therefore, at load time, all checks on pointers
1544 * previously done by the verifier are invalidated and must be
1545 * performed again, if the helper is used in combination with
1546 * direct packet access.
1548 * 0 on success, or a negative error in case of failure.
1550 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1552 * Copy a NUL terminated string from an unsafe kernel address
1553 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
1556 * Generally, use **bpf_probe_read_user_str**\ () or
1557 * **bpf_probe_read_kernel_str**\ () instead.
1559 * On success, the strictly positive length of the string,
1560 * including the trailing NUL character. On error, a negative
1563 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1565 * If the **struct sk_buff** pointed by *skb* has a known socket,
1566 * retrieve the cookie (generated by the kernel) of this socket.
1567 * If no cookie has been set yet, generate a new cookie. Once
1568 * generated, the socket cookie remains stable for the life of the
1569 * socket. This helper can be useful for monitoring per socket
1570 * networking traffic statistics as it provides a global socket
1571 * identifier that can be assumed unique.
1573 * A 8-byte long non-decreasing number on success, or 0 if the
1574 * socket field is missing inside *skb*.
1576 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1578 * Equivalent to bpf_get_socket_cookie() helper that accepts
1579 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1581 * A 8-byte long non-decreasing number.
1583 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1585 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
1586 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1588 * A 8-byte long non-decreasing number.
1590 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1592 * The owner UID of the socket associated to *skb*. If the socket
1593 * is **NULL**, or if it is not a full socket (i.e. if it is a
1594 * time-wait or a request socket instead), **overflowuid** value
1595 * is returned (note that **overflowuid** might also be the actual
1596 * UID value for the socket).
1598 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1600 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1605 * int bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1607 * Emulate a call to **setsockopt()** on the socket associated to
1608 * *bpf_socket*, which must be a full socket. The *level* at
1609 * which the option resides and the name *optname* of the option
1610 * must be specified, see **setsockopt(2)** for more information.
1611 * The option value of length *optlen* is pointed by *optval*.
1613 * *bpf_socket* should be one of the following:
1615 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1616 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1617 * and **BPF_CGROUP_INET6_CONNECT**.
1619 * This helper actually implements a subset of **setsockopt()**.
1620 * It supports the following *level*\ s:
1622 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1623 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1624 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1625 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1626 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1627 * **TCP_BPF_SNDCWND_CLAMP**.
1628 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1629 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1631 * 0 on success, or a negative error in case of failure.
1633 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1635 * Grow or shrink the room for data in the packet associated to
1636 * *skb* by *len_diff*, and according to the selected *mode*.
1638 * There are two supported modes at this time:
1640 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1641 * (room space is added or removed below the layer 2 header).
1643 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1644 * (room space is added or removed below the layer 3 header).
1646 * The following flags are supported at this time:
1648 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1649 * Adjusting mss in this way is not allowed for datagrams.
1651 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1652 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1653 * Any new space is reserved to hold a tunnel header.
1654 * Configure skb offsets and other fields accordingly.
1656 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1657 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1658 * Use with ENCAP_L3 flags to further specify the tunnel type.
1660 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1661 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1662 * type; *len* is the length of the inner MAC header.
1664 * A call to this helper is susceptible to change the underlying
1665 * packet buffer. Therefore, at load time, all checks on pointers
1666 * previously done by the verifier are invalidated and must be
1667 * performed again, if the helper is used in combination with
1668 * direct packet access.
1670 * 0 on success, or a negative error in case of failure.
1672 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1674 * Redirect the packet to the endpoint referenced by *map* at
1675 * index *key*. Depending on its type, this *map* can contain
1676 * references to net devices (for forwarding packets through other
1677 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1678 * but this is only implemented for native XDP (with driver
1679 * support) as of this writing).
1681 * The lower two bits of *flags* are used as the return code if
1682 * the map lookup fails. This is so that the return value can be
1683 * one of the XDP program return codes up to **XDP_TX**, as chosen
1684 * by the caller. Any higher bits in the *flags* argument must be
1687 * See also **bpf_redirect**\ (), which only supports redirecting
1688 * to an ifindex, but doesn't require a map to do so.
1690 * **XDP_REDIRECT** on success, or the value of the two lower bits
1691 * of the *flags* argument on error.
1693 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1695 * Redirect the packet to the socket referenced by *map* (of type
1696 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1697 * egress interfaces can be used for redirection. The
1698 * **BPF_F_INGRESS** value in *flags* is used to make the
1699 * distinction (ingress path is selected if the flag is present,
1700 * egress path otherwise). This is the only flag supported for now.
1702 * **SK_PASS** on success, or **SK_DROP** on error.
1704 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1706 * Add an entry to, or update a *map* referencing sockets. The
1707 * *skops* is used as a new value for the entry associated to
1708 * *key*. *flags* is one of:
1711 * The entry for *key* must not exist in the map.
1713 * The entry for *key* must already exist in the map.
1715 * No condition on the existence of the entry for *key*.
1717 * If the *map* has eBPF programs (parser and verdict), those will
1718 * be inherited by the socket being added. If the socket is
1719 * already attached to eBPF programs, this results in an error.
1721 * 0 on success, or a negative error in case of failure.
1723 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1725 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1726 * *delta* (which can be positive or negative). Note that this
1727 * operation modifies the address stored in *xdp_md*\ **->data**,
1728 * so the latter must be loaded only after the helper has been
1731 * The use of *xdp_md*\ **->data_meta** is optional and programs
1732 * are not required to use it. The rationale is that when the
1733 * packet is processed with XDP (e.g. as DoS filter), it is
1734 * possible to push further meta data along with it before passing
1735 * to the stack, and to give the guarantee that an ingress eBPF
1736 * program attached as a TC classifier on the same device can pick
1737 * this up for further post-processing. Since TC works with socket
1738 * buffers, it remains possible to set from XDP the **mark** or
1739 * **priority** pointers, or other pointers for the socket buffer.
1740 * Having this scratch space generic and programmable allows for
1741 * more flexibility as the user is free to store whatever meta
1744 * A call to this helper is susceptible to change the underlying
1745 * packet buffer. Therefore, at load time, all checks on pointers
1746 * previously done by the verifier are invalidated and must be
1747 * performed again, if the helper is used in combination with
1748 * direct packet access.
1750 * 0 on success, or a negative error in case of failure.
1752 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1754 * Read the value of a perf event counter, and store it into *buf*
1755 * of size *buf_size*. This helper relies on a *map* of type
1756 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1757 * counter is selected when *map* is updated with perf event file
1758 * descriptors. The *map* is an array whose size is the number of
1759 * available CPUs, and each cell contains a value relative to one
1760 * CPU. The value to retrieve is indicated by *flags*, that
1761 * contains the index of the CPU to look up, masked with
1762 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1763 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1764 * current CPU should be retrieved.
1766 * This helper behaves in a way close to
1767 * **bpf_perf_event_read**\ () helper, save that instead of
1768 * just returning the value observed, it fills the *buf*
1769 * structure. This allows for additional data to be retrieved: in
1770 * particular, the enabled and running times (in *buf*\
1771 * **->enabled** and *buf*\ **->running**, respectively) are
1772 * copied. In general, **bpf_perf_event_read_value**\ () is
1773 * recommended over **bpf_perf_event_read**\ (), which has some
1774 * ABI issues and provides fewer functionalities.
1776 * These values are interesting, because hardware PMU (Performance
1777 * Monitoring Unit) counters are limited resources. When there are
1778 * more PMU based perf events opened than available counters,
1779 * kernel will multiplex these events so each event gets certain
1780 * percentage (but not all) of the PMU time. In case that
1781 * multiplexing happens, the number of samples or counter value
1782 * will not reflect the case compared to when no multiplexing
1783 * occurs. This makes comparison between different runs difficult.
1784 * Typically, the counter value should be normalized before
1785 * comparing to other experiments. The usual normalization is done
1790 * normalized_counter = counter * t_enabled / t_running
1792 * Where t_enabled is the time enabled for event and t_running is
1793 * the time running for event since last normalization. The
1794 * enabled and running times are accumulated since the perf event
1795 * open. To achieve scaling factor between two invocations of an
1796 * eBPF program, users can use CPU id as the key (which is
1797 * typical for perf array usage model) to remember the previous
1798 * value and do the calculation inside the eBPF program.
1800 * 0 on success, or a negative error in case of failure.
1802 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1804 * For en eBPF program attached to a perf event, retrieve the
1805 * value of the event counter associated to *ctx* and store it in
1806 * the structure pointed by *buf* and of size *buf_size*. Enabled
1807 * and running times are also stored in the structure (see
1808 * description of helper **bpf_perf_event_read_value**\ () for
1811 * 0 on success, or a negative error in case of failure.
1813 * int bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1815 * Emulate a call to **getsockopt()** on the socket associated to
1816 * *bpf_socket*, which must be a full socket. The *level* at
1817 * which the option resides and the name *optname* of the option
1818 * must be specified, see **getsockopt(2)** for more information.
1819 * The retrieved value is stored in the structure pointed by
1820 * *opval* and of length *optlen*.
1822 * *bpf_socket* should be one of the following:
1824 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1825 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1826 * and **BPF_CGROUP_INET6_CONNECT**.
1828 * This helper actually implements a subset of **getsockopt()**.
1829 * It supports the following *level*\ s:
1831 * * **IPPROTO_TCP**, which supports *optname*
1832 * **TCP_CONGESTION**.
1833 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1834 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1836 * 0 on success, or a negative error in case of failure.
1838 * int bpf_override_return(struct pt_regs *regs, u64 rc)
1840 * Used for error injection, this helper uses kprobes to override
1841 * the return value of the probed function, and to set it to *rc*.
1842 * The first argument is the context *regs* on which the kprobe
1845 * This helper works by setting the PC (program counter)
1846 * to an override function which is run in place of the original
1847 * probed function. This means the probed function is not run at
1848 * all. The replacement function just returns with the required
1851 * This helper has security implications, and thus is subject to
1852 * restrictions. It is only available if the kernel was compiled
1853 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1854 * option, and in this case it only works on functions tagged with
1855 * **ALLOW_ERROR_INJECTION** in the kernel code.
1857 * Also, the helper is only available for the architectures having
1858 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1859 * x86 architecture is the only one to support this feature.
1863 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1865 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1866 * for the full TCP socket associated to *bpf_sock_ops* to
1869 * The primary use of this field is to determine if there should
1870 * be calls to eBPF programs of type
1871 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1872 * code. A program of the same type can change its value, per
1873 * connection and as necessary, when the connection is
1874 * established. This field is directly accessible for reading, but
1875 * this helper must be used for updates in order to return an
1876 * error if an eBPF program tries to set a callback that is not
1877 * supported in the current kernel.
1879 * *argval* is a flag array which can combine these flags:
1881 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1882 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1883 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1884 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1886 * Therefore, this function can be used to clear a callback flag by
1887 * setting the appropriate bit to zero. e.g. to disable the RTO
1890 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1891 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1893 * Here are some examples of where one could call such eBPF
1897 * * When a packet is retransmitted.
1898 * * When the connection terminates.
1899 * * When a packet is sent.
1900 * * When a packet is received.
1902 * Code **-EINVAL** if the socket is not a full TCP socket;
1903 * otherwise, a positive number containing the bits that could not
1904 * be set is returned (which comes down to 0 if all bits were set
1907 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1909 * This helper is used in programs implementing policies at the
1910 * socket level. If the message *msg* is allowed to pass (i.e. if
1911 * the verdict eBPF program returns **SK_PASS**), redirect it to
1912 * the socket referenced by *map* (of type
1913 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1914 * egress interfaces can be used for redirection. The
1915 * **BPF_F_INGRESS** value in *flags* is used to make the
1916 * distinction (ingress path is selected if the flag is present,
1917 * egress path otherwise). This is the only flag supported for now.
1919 * **SK_PASS** on success, or **SK_DROP** on error.
1921 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1923 * For socket policies, apply the verdict of the eBPF program to
1924 * the next *bytes* (number of bytes) of message *msg*.
1926 * For example, this helper can be used in the following cases:
1928 * * A single **sendmsg**\ () or **sendfile**\ () system call
1929 * contains multiple logical messages that the eBPF program is
1930 * supposed to read and for which it should apply a verdict.
1931 * * An eBPF program only cares to read the first *bytes* of a
1932 * *msg*. If the message has a large payload, then setting up
1933 * and calling the eBPF program repeatedly for all bytes, even
1934 * though the verdict is already known, would create unnecessary
1937 * When called from within an eBPF program, the helper sets a
1938 * counter internal to the BPF infrastructure, that is used to
1939 * apply the last verdict to the next *bytes*. If *bytes* is
1940 * smaller than the current data being processed from a
1941 * **sendmsg**\ () or **sendfile**\ () system call, the first
1942 * *bytes* will be sent and the eBPF program will be re-run with
1943 * the pointer for start of data pointing to byte number *bytes*
1944 * **+ 1**. If *bytes* is larger than the current data being
1945 * processed, then the eBPF verdict will be applied to multiple
1946 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1949 * Note that if a socket closes with the internal counter holding
1950 * a non-zero value, this is not a problem because data is not
1951 * being buffered for *bytes* and is sent as it is received.
1955 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1957 * For socket policies, prevent the execution of the verdict eBPF
1958 * program for message *msg* until *bytes* (byte number) have been
1961 * This can be used when one needs a specific number of bytes
1962 * before a verdict can be assigned, even if the data spans
1963 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1964 * case would be a user calling **sendmsg**\ () repeatedly with
1965 * 1-byte long message segments. Obviously, this is bad for
1966 * performance, but it is still valid. If the eBPF program needs
1967 * *bytes* bytes to validate a header, this helper can be used to
1968 * prevent the eBPF program to be called again until *bytes* have
1973 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1975 * For socket policies, pull in non-linear data from user space
1976 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1977 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1980 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1981 * *msg* it can only parse data that the (**data**, **data_end**)
1982 * pointers have already consumed. For **sendmsg**\ () hooks this
1983 * is likely the first scatterlist element. But for calls relying
1984 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1985 * be the range (**0**, **0**) because the data is shared with
1986 * user space and by default the objective is to avoid allowing
1987 * user space to modify data while (or after) eBPF verdict is
1988 * being decided. This helper can be used to pull in data and to
1989 * set the start and end pointer to given values. Data will be
1990 * copied if necessary (i.e. if data was not linear and if start
1991 * and end pointers do not point to the same chunk).
1993 * A call to this helper is susceptible to change the underlying
1994 * packet buffer. Therefore, at load time, all checks on pointers
1995 * previously done by the verifier are invalidated and must be
1996 * performed again, if the helper is used in combination with
1997 * direct packet access.
1999 * All values for *flags* are reserved for future usage, and must
2002 * 0 on success, or a negative error in case of failure.
2004 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2006 * Bind the socket associated to *ctx* to the address pointed by
2007 * *addr*, of length *addr_len*. This allows for making outgoing
2008 * connection from the desired IP address, which can be useful for
2009 * example when all processes inside a cgroup should use one
2010 * single IP address on a host that has multiple IP configured.
2012 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2013 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2014 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2015 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2016 * behavior and lets the kernel efficiently pick up an unused
2017 * port as long as 4-tuple is unique. Passing non-zero port might
2018 * lead to degraded performance.
2020 * 0 on success, or a negative error in case of failure.
2022 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2024 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2025 * possible to both shrink and grow the packet tail.
2026 * Shrink done via *delta* being a negative integer.
2028 * A call to this helper is susceptible to change the underlying
2029 * packet buffer. Therefore, at load time, all checks on pointers
2030 * previously done by the verifier are invalidated and must be
2031 * performed again, if the helper is used in combination with
2032 * direct packet access.
2034 * 0 on success, or a negative error in case of failure.
2036 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2038 * Retrieve the XFRM state (IP transform framework, see also
2039 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2041 * The retrieved value is stored in the **struct bpf_xfrm_state**
2042 * pointed by *xfrm_state* and of length *size*.
2044 * All values for *flags* are reserved for future usage, and must
2047 * This helper is available only if the kernel was compiled with
2048 * **CONFIG_XFRM** configuration option.
2050 * 0 on success, or a negative error in case of failure.
2052 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2054 * Return a user or a kernel stack in bpf program provided buffer.
2055 * To achieve this, the helper needs *ctx*, which is a pointer
2056 * to the context on which the tracing program is executed.
2057 * To store the stacktrace, the bpf program provides *buf* with
2058 * a nonnegative *size*.
2060 * The last argument, *flags*, holds the number of stack frames to
2061 * skip (from 0 to 255), masked with
2062 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2063 * the following flags:
2065 * **BPF_F_USER_STACK**
2066 * Collect a user space stack instead of a kernel stack.
2067 * **BPF_F_USER_BUILD_ID**
2068 * Collect buildid+offset instead of ips for user stack,
2069 * only valid if **BPF_F_USER_STACK** is also specified.
2071 * **bpf_get_stack**\ () can collect up to
2072 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2073 * to sufficient large buffer size. Note that
2074 * this limit can be controlled with the **sysctl** program, and
2075 * that it should be manually increased in order to profile long
2076 * user stacks (such as stacks for Java programs). To do so, use:
2080 * # sysctl kernel.perf_event_max_stack=<new value>
2082 * A non-negative value equal to or less than *size* on success,
2083 * or a negative error in case of failure.
2085 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2087 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2088 * it provides an easy way to load *len* bytes from *offset*
2089 * from the packet associated to *skb*, into the buffer pointed
2090 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2091 * a fifth argument *start_header* exists in order to select a
2092 * base offset to start from. *start_header* can be one of:
2094 * **BPF_HDR_START_MAC**
2095 * Base offset to load data from is *skb*'s mac header.
2096 * **BPF_HDR_START_NET**
2097 * Base offset to load data from is *skb*'s network header.
2099 * In general, "direct packet access" is the preferred method to
2100 * access packet data, however, this helper is in particular useful
2101 * in socket filters where *skb*\ **->data** does not always point
2102 * to the start of the mac header and where "direct packet access"
2105 * 0 on success, or a negative error in case of failure.
2107 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2109 * Do FIB lookup in kernel tables using parameters in *params*.
2110 * If lookup is successful and result shows packet is to be
2111 * forwarded, the neighbor tables are searched for the nexthop.
2112 * If successful (ie., FIB lookup shows forwarding and nexthop
2113 * is resolved), the nexthop address is returned in ipv4_dst
2114 * or ipv6_dst based on family, smac is set to mac address of
2115 * egress device, dmac is set to nexthop mac address, rt_metric
2116 * is set to metric from route (IPv4/IPv6 only), and ifindex
2117 * is set to the device index of the nexthop from the FIB lookup.
2119 * *plen* argument is the size of the passed in struct.
2120 * *flags* argument can be a combination of one or more of the
2123 * **BPF_FIB_LOOKUP_DIRECT**
2124 * Do a direct table lookup vs full lookup using FIB
2126 * **BPF_FIB_LOOKUP_OUTPUT**
2127 * Perform lookup from an egress perspective (default is
2130 * *ctx* is either **struct xdp_md** for XDP programs or
2131 * **struct sk_buff** tc cls_act programs.
2133 * * < 0 if any input argument is invalid
2134 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2135 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2136 * packet is not forwarded or needs assist from full stack
2138 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2140 * Add an entry to, or update a sockhash *map* referencing sockets.
2141 * The *skops* is used as a new value for the entry associated to
2142 * *key*. *flags* is one of:
2145 * The entry for *key* must not exist in the map.
2147 * The entry for *key* must already exist in the map.
2149 * No condition on the existence of the entry for *key*.
2151 * If the *map* has eBPF programs (parser and verdict), those will
2152 * be inherited by the socket being added. If the socket is
2153 * already attached to eBPF programs, this results in an error.
2155 * 0 on success, or a negative error in case of failure.
2157 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2159 * This helper is used in programs implementing policies at the
2160 * socket level. If the message *msg* is allowed to pass (i.e. if
2161 * the verdict eBPF program returns **SK_PASS**), redirect it to
2162 * the socket referenced by *map* (of type
2163 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2164 * egress interfaces can be used for redirection. The
2165 * **BPF_F_INGRESS** value in *flags* is used to make the
2166 * distinction (ingress path is selected if the flag is present,
2167 * egress path otherwise). This is the only flag supported for now.
2169 * **SK_PASS** on success, or **SK_DROP** on error.
2171 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2173 * This helper is used in programs implementing policies at the
2174 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2175 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2176 * to the socket referenced by *map* (of type
2177 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2178 * egress interfaces can be used for redirection. The
2179 * **BPF_F_INGRESS** value in *flags* is used to make the
2180 * distinction (ingress path is selected if the flag is present,
2181 * egress otherwise). This is the only flag supported for now.
2183 * **SK_PASS** on success, or **SK_DROP** on error.
2185 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2187 * Encapsulate the packet associated to *skb* within a Layer 3
2188 * protocol header. This header is provided in the buffer at
2189 * address *hdr*, with *len* its size in bytes. *type* indicates
2190 * the protocol of the header and can be one of:
2192 * **BPF_LWT_ENCAP_SEG6**
2193 * IPv6 encapsulation with Segment Routing Header
2194 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2195 * the IPv6 header is computed by the kernel.
2196 * **BPF_LWT_ENCAP_SEG6_INLINE**
2197 * Only works if *skb* contains an IPv6 packet. Insert a
2198 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2200 * **BPF_LWT_ENCAP_IP**
2201 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2202 * must be IPv4 or IPv6, followed by zero or more
2203 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2204 * total bytes in all prepended headers. Please note that
2205 * if **skb_is_gso**\ (*skb*) is true, no more than two
2206 * headers can be prepended, and the inner header, if
2207 * present, should be either GRE or UDP/GUE.
2209 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2210 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2211 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2212 * **BPF_PROG_TYPE_LWT_XMIT**.
2214 * A call to this helper is susceptible to change the underlying
2215 * packet buffer. Therefore, at load time, all checks on pointers
2216 * previously done by the verifier are invalidated and must be
2217 * performed again, if the helper is used in combination with
2218 * direct packet access.
2220 * 0 on success, or a negative error in case of failure.
2222 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2224 * Store *len* bytes from address *from* into the packet
2225 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2226 * inside the outermost IPv6 Segment Routing Header can be
2227 * modified through this helper.
2229 * A call to this helper is susceptible to change the underlying
2230 * packet buffer. Therefore, at load time, all checks on pointers
2231 * previously done by the verifier are invalidated and must be
2232 * performed again, if the helper is used in combination with
2233 * direct packet access.
2235 * 0 on success, or a negative error in case of failure.
2237 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2239 * Adjust the size allocated to TLVs in the outermost IPv6
2240 * Segment Routing Header contained in the packet associated to
2241 * *skb*, at position *offset* by *delta* bytes. Only offsets
2242 * after the segments are accepted. *delta* can be as well
2243 * positive (growing) as negative (shrinking).
2245 * A call to this helper is susceptible to change the underlying
2246 * packet buffer. Therefore, at load time, all checks on pointers
2247 * previously done by the verifier are invalidated and must be
2248 * performed again, if the helper is used in combination with
2249 * direct packet access.
2251 * 0 on success, or a negative error in case of failure.
2253 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2255 * Apply an IPv6 Segment Routing action of type *action* to the
2256 * packet associated to *skb*. Each action takes a parameter
2257 * contained at address *param*, and of length *param_len* bytes.
2258 * *action* can be one of:
2260 * **SEG6_LOCAL_ACTION_END_X**
2261 * End.X action: Endpoint with Layer-3 cross-connect.
2262 * Type of *param*: **struct in6_addr**.
2263 * **SEG6_LOCAL_ACTION_END_T**
2264 * End.T action: Endpoint with specific IPv6 table lookup.
2265 * Type of *param*: **int**.
2266 * **SEG6_LOCAL_ACTION_END_B6**
2267 * End.B6 action: Endpoint bound to an SRv6 policy.
2268 * Type of *param*: **struct ipv6_sr_hdr**.
2269 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2270 * End.B6.Encap action: Endpoint bound to an SRv6
2271 * encapsulation policy.
2272 * Type of *param*: **struct ipv6_sr_hdr**.
2274 * A call to this helper is susceptible to change the underlying
2275 * packet buffer. Therefore, at load time, all checks on pointers
2276 * previously done by the verifier are invalidated and must be
2277 * performed again, if the helper is used in combination with
2278 * direct packet access.
2280 * 0 on success, or a negative error in case of failure.
2282 * int bpf_rc_repeat(void *ctx)
2284 * This helper is used in programs implementing IR decoding, to
2285 * report a successfully decoded repeat key message. This delays
2286 * the generation of a key up event for previously generated
2289 * Some IR protocols like NEC have a special IR message for
2290 * repeating last button, for when a button is held down.
2292 * The *ctx* should point to the lirc sample as passed into
2295 * This helper is only available is the kernel was compiled with
2296 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2301 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2303 * This helper is used in programs implementing IR decoding, to
2304 * report a successfully decoded key press with *scancode*,
2305 * *toggle* value in the given *protocol*. The scancode will be
2306 * translated to a keycode using the rc keymap, and reported as
2307 * an input key down event. After a period a key up event is
2308 * generated. This period can be extended by calling either
2309 * **bpf_rc_keydown**\ () again with the same values, or calling
2310 * **bpf_rc_repeat**\ ().
2312 * Some protocols include a toggle bit, in case the button was
2313 * released and pressed again between consecutive scancodes.
2315 * The *ctx* should point to the lirc sample as passed into
2318 * The *protocol* is the decoded protocol number (see
2319 * **enum rc_proto** for some predefined values).
2321 * This helper is only available is the kernel was compiled with
2322 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2327 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2329 * Return the cgroup v2 id of the socket associated with the *skb*.
2330 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2331 * helper for cgroup v1 by providing a tag resp. identifier that
2332 * can be matched on or used for map lookups e.g. to implement
2333 * policy. The cgroup v2 id of a given path in the hierarchy is
2334 * exposed in user space through the f_handle API in order to get
2335 * to the same 64-bit id.
2337 * This helper can be used on TC egress path, but not on ingress,
2338 * and is available only if the kernel was compiled with the
2339 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2341 * The id is returned or 0 in case the id could not be retrieved.
2343 * u64 bpf_get_current_cgroup_id(void)
2345 * A 64-bit integer containing the current cgroup id based
2346 * on the cgroup within which the current task is running.
2348 * void *bpf_get_local_storage(void *map, u64 flags)
2350 * Get the pointer to the local storage area.
2351 * The type and the size of the local storage is defined
2352 * by the *map* argument.
2353 * The *flags* meaning is specific for each map type,
2354 * and has to be 0 for cgroup local storage.
2356 * Depending on the BPF program type, a local storage area
2357 * can be shared between multiple instances of the BPF program,
2358 * running simultaneously.
2360 * A user should care about the synchronization by himself.
2361 * For example, by using the **BPF_STX_XADD** instruction to alter
2364 * A pointer to the local storage area.
2366 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2368 * Select a **SO_REUSEPORT** socket from a
2369 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2370 * It checks the selected socket is matching the incoming
2371 * request in the socket buffer.
2373 * 0 on success, or a negative error in case of failure.
2375 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2377 * Return id of cgroup v2 that is ancestor of cgroup associated
2378 * with the *skb* at the *ancestor_level*. The root cgroup is at
2379 * *ancestor_level* zero and each step down the hierarchy
2380 * increments the level. If *ancestor_level* == level of cgroup
2381 * associated with *skb*, then return value will be same as that
2382 * of **bpf_skb_cgroup_id**\ ().
2384 * The helper is useful to implement policies based on cgroups
2385 * that are upper in hierarchy than immediate cgroup associated
2388 * The format of returned id and helper limitations are same as in
2389 * **bpf_skb_cgroup_id**\ ().
2391 * The id is returned or 0 in case the id could not be retrieved.
2393 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2395 * Look for TCP socket matching *tuple*, optionally in a child
2396 * network namespace *netns*. The return value must be checked,
2397 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2399 * The *ctx* should point to the context of the program, such as
2400 * the skb or socket (depending on the hook in use). This is used
2401 * to determine the base network namespace for the lookup.
2403 * *tuple_size* must be one of:
2405 * **sizeof**\ (*tuple*\ **->ipv4**)
2406 * Look for an IPv4 socket.
2407 * **sizeof**\ (*tuple*\ **->ipv6**)
2408 * Look for an IPv6 socket.
2410 * If the *netns* is a negative signed 32-bit integer, then the
2411 * socket lookup table in the netns associated with the *ctx* will
2412 * will be used. For the TC hooks, this is the netns of the device
2413 * in the skb. For socket hooks, this is the netns of the socket.
2414 * If *netns* is any other signed 32-bit value greater than or
2415 * equal to zero then it specifies the ID of the netns relative to
2416 * the netns associated with the *ctx*. *netns* values beyond the
2417 * range of 32-bit integers are reserved for future use.
2419 * All values for *flags* are reserved for future usage, and must
2422 * This helper is available only if the kernel was compiled with
2423 * **CONFIG_NET** configuration option.
2425 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2426 * For sockets with reuseport option, the **struct bpf_sock**
2427 * result is from *reuse*\ **->socks**\ [] using the hash of the
2430 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2432 * Look for UDP socket matching *tuple*, optionally in a child
2433 * network namespace *netns*. The return value must be checked,
2434 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2436 * The *ctx* should point to the context of the program, such as
2437 * the skb or socket (depending on the hook in use). This is used
2438 * to determine the base network namespace for the lookup.
2440 * *tuple_size* must be one of:
2442 * **sizeof**\ (*tuple*\ **->ipv4**)
2443 * Look for an IPv4 socket.
2444 * **sizeof**\ (*tuple*\ **->ipv6**)
2445 * Look for an IPv6 socket.
2447 * If the *netns* is a negative signed 32-bit integer, then the
2448 * socket lookup table in the netns associated with the *ctx* will
2449 * will be used. For the TC hooks, this is the netns of the device
2450 * in the skb. For socket hooks, this is the netns of the socket.
2451 * If *netns* is any other signed 32-bit value greater than or
2452 * equal to zero then it specifies the ID of the netns relative to
2453 * the netns associated with the *ctx*. *netns* values beyond the
2454 * range of 32-bit integers are reserved for future use.
2456 * All values for *flags* are reserved for future usage, and must
2459 * This helper is available only if the kernel was compiled with
2460 * **CONFIG_NET** configuration option.
2462 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2463 * For sockets with reuseport option, the **struct bpf_sock**
2464 * result is from *reuse*\ **->socks**\ [] using the hash of the
2467 * int bpf_sk_release(struct bpf_sock *sock)
2469 * Release the reference held by *sock*. *sock* must be a
2470 * non-**NULL** pointer that was returned from
2471 * **bpf_sk_lookup_xxx**\ ().
2473 * 0 on success, or a negative error in case of failure.
2475 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2477 * Push an element *value* in *map*. *flags* is one of:
2480 * If the queue/stack is full, the oldest element is
2481 * removed to make room for this.
2483 * 0 on success, or a negative error in case of failure.
2485 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2487 * Pop an element from *map*.
2489 * 0 on success, or a negative error in case of failure.
2491 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2493 * Get an element from *map* without removing it.
2495 * 0 on success, or a negative error in case of failure.
2497 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2499 * For socket policies, insert *len* bytes into *msg* at offset
2502 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2503 * *msg* it may want to insert metadata or options into the *msg*.
2504 * This can later be read and used by any of the lower layer BPF
2507 * This helper may fail if under memory pressure (a malloc
2508 * fails) in these cases BPF programs will get an appropriate
2509 * error and BPF programs will need to handle them.
2511 * 0 on success, or a negative error in case of failure.
2513 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2515 * Will remove *len* bytes from a *msg* starting at byte *start*.
2516 * This may result in **ENOMEM** errors under certain situations if
2517 * an allocation and copy are required due to a full ring buffer.
2518 * However, the helper will try to avoid doing the allocation
2519 * if possible. Other errors can occur if input parameters are
2520 * invalid either due to *start* byte not being valid part of *msg*
2521 * payload and/or *pop* value being to large.
2523 * 0 on success, or a negative error in case of failure.
2525 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2527 * This helper is used in programs implementing IR decoding, to
2528 * report a successfully decoded pointer movement.
2530 * The *ctx* should point to the lirc sample as passed into
2533 * This helper is only available is the kernel was compiled with
2534 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2539 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2541 * Acquire a spinlock represented by the pointer *lock*, which is
2542 * stored as part of a value of a map. Taking the lock allows to
2543 * safely update the rest of the fields in that value. The
2544 * spinlock can (and must) later be released with a call to
2545 * **bpf_spin_unlock**\ (\ *lock*\ ).
2547 * Spinlocks in BPF programs come with a number of restrictions
2550 * * **bpf_spin_lock** objects are only allowed inside maps of
2551 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2552 * list could be extended in the future).
2553 * * BTF description of the map is mandatory.
2554 * * The BPF program can take ONE lock at a time, since taking two
2555 * or more could cause dead locks.
2556 * * Only one **struct bpf_spin_lock** is allowed per map element.
2557 * * When the lock is taken, calls (either BPF to BPF or helpers)
2559 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2560 * allowed inside a spinlock-ed region.
2561 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2562 * the lock, on all execution paths, before it returns.
2563 * * The BPF program can access **struct bpf_spin_lock** only via
2564 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2565 * helpers. Loading or storing data into the **struct
2566 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2567 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2568 * of the map value must be a struct and have **struct
2569 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2570 * Nested lock inside another struct is not allowed.
2571 * * The **struct bpf_spin_lock** *lock* field in a map value must
2572 * be aligned on a multiple of 4 bytes in that value.
2573 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2574 * the **bpf_spin_lock** field to user space.
2575 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2576 * a BPF program, do not update the **bpf_spin_lock** field.
2577 * * **bpf_spin_lock** cannot be on the stack or inside a
2578 * networking packet (it can only be inside of a map values).
2579 * * **bpf_spin_lock** is available to root only.
2580 * * Tracing programs and socket filter programs cannot use
2581 * **bpf_spin_lock**\ () due to insufficient preemption checks
2582 * (but this may change in the future).
2583 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2587 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2589 * Release the *lock* previously locked by a call to
2590 * **bpf_spin_lock**\ (\ *lock*\ ).
2594 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2596 * This helper gets a **struct bpf_sock** pointer such
2597 * that all the fields in this **bpf_sock** can be accessed.
2599 * A **struct bpf_sock** pointer on success, or **NULL** in
2602 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2604 * This helper gets a **struct bpf_tcp_sock** pointer from a
2605 * **struct bpf_sock** pointer.
2607 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2610 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
2612 * Set ECN (Explicit Congestion Notification) field of IP header
2613 * to **CE** (Congestion Encountered) if current value is **ECT**
2614 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2617 * 1 if the **CE** flag is set (either by the current helper call
2618 * or because it was already present), 0 if it is not set.
2620 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2622 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2623 * **bpf_sk_release**\ () is unnecessary and not allowed.
2625 * A **struct bpf_sock** pointer on success, or **NULL** in
2628 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2630 * Look for TCP socket matching *tuple*, optionally in a child
2631 * network namespace *netns*. The return value must be checked,
2632 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2634 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2635 * that it also returns timewait or request sockets. Use
2636 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2639 * This helper is available only if the kernel was compiled with
2640 * **CONFIG_NET** configuration option.
2642 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2643 * For sockets with reuseport option, the **struct bpf_sock**
2644 * result is from *reuse*\ **->socks**\ [] using the hash of the
2647 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2649 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2650 * the listening socket in *sk*.
2652 * *iph* points to the start of the IPv4 or IPv6 header, while
2653 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2654 * **sizeof**\ (**struct ip6hdr**).
2656 * *th* points to the start of the TCP header, while *th_len*
2657 * contains **sizeof**\ (**struct tcphdr**).
2659 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2662 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2664 * Get name of sysctl in /proc/sys/ and copy it into provided by
2665 * program buffer *buf* of size *buf_len*.
2667 * The buffer is always NUL terminated, unless it's zero-sized.
2669 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2670 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2671 * only (e.g. "tcp_mem").
2673 * Number of character copied (not including the trailing NUL).
2675 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2676 * truncated name in this case).
2678 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2680 * Get current value of sysctl as it is presented in /proc/sys
2681 * (incl. newline, etc), and copy it as a string into provided
2682 * by program buffer *buf* of size *buf_len*.
2684 * The whole value is copied, no matter what file position user
2685 * space issued e.g. sys_read at.
2687 * The buffer is always NUL terminated, unless it's zero-sized.
2689 * Number of character copied (not including the trailing NUL).
2691 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2692 * truncated name in this case).
2694 * **-EINVAL** if current value was unavailable, e.g. because
2695 * sysctl is uninitialized and read returns -EIO for it.
2697 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2699 * Get new value being written by user space to sysctl (before
2700 * the actual write happens) and copy it as a string into
2701 * provided by program buffer *buf* of size *buf_len*.
2703 * User space may write new value at file position > 0.
2705 * The buffer is always NUL terminated, unless it's zero-sized.
2707 * Number of character copied (not including the trailing NUL).
2709 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2710 * truncated name in this case).
2712 * **-EINVAL** if sysctl is being read.
2714 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2716 * Override new value being written by user space to sysctl with
2717 * value provided by program in buffer *buf* of size *buf_len*.
2719 * *buf* should contain a string in same form as provided by user
2720 * space on sysctl write.
2722 * User space may write new value at file position > 0. To override
2723 * the whole sysctl value file position should be set to zero.
2727 * **-E2BIG** if the *buf_len* is too big.
2729 * **-EINVAL** if sysctl is being read.
2731 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2733 * Convert the initial part of the string from buffer *buf* of
2734 * size *buf_len* to a long integer according to the given base
2735 * and save the result in *res*.
2737 * The string may begin with an arbitrary amount of white space
2738 * (as determined by **isspace**\ (3)) followed by a single
2739 * optional '**-**' sign.
2741 * Five least significant bits of *flags* encode base, other bits
2742 * are currently unused.
2744 * Base must be either 8, 10, 16 or 0 to detect it automatically
2745 * similar to user space **strtol**\ (3).
2747 * Number of characters consumed on success. Must be positive but
2748 * no more than *buf_len*.
2750 * **-EINVAL** if no valid digits were found or unsupported base
2753 * **-ERANGE** if resulting value was out of range.
2755 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2757 * Convert the initial part of the string from buffer *buf* of
2758 * size *buf_len* to an unsigned long integer according to the
2759 * given base and save the result in *res*.
2761 * The string may begin with an arbitrary amount of white space
2762 * (as determined by **isspace**\ (3)).
2764 * Five least significant bits of *flags* encode base, other bits
2765 * are currently unused.
2767 * Base must be either 8, 10, 16 or 0 to detect it automatically
2768 * similar to user space **strtoul**\ (3).
2770 * Number of characters consumed on success. Must be positive but
2771 * no more than *buf_len*.
2773 * **-EINVAL** if no valid digits were found or unsupported base
2776 * **-ERANGE** if resulting value was out of range.
2778 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2780 * Get a bpf-local-storage from a *sk*.
2782 * Logically, it could be thought of getting the value from
2783 * a *map* with *sk* as the **key**. From this
2784 * perspective, the usage is not much different from
2785 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2786 * helper enforces the key must be a full socket and the map must
2787 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2789 * Underneath, the value is stored locally at *sk* instead of
2790 * the *map*. The *map* is used as the bpf-local-storage
2791 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2792 * searched against all bpf-local-storages residing at *sk*.
2794 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2795 * used such that a new bpf-local-storage will be
2796 * created if one does not exist. *value* can be used
2797 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2798 * the initial value of a bpf-local-storage. If *value* is
2799 * **NULL**, the new bpf-local-storage will be zero initialized.
2801 * A bpf-local-storage pointer is returned on success.
2803 * **NULL** if not found or there was an error in adding
2804 * a new bpf-local-storage.
2806 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2808 * Delete a bpf-local-storage from a *sk*.
2812 * **-ENOENT** if the bpf-local-storage cannot be found.
2814 * int bpf_send_signal(u32 sig)
2816 * Send signal *sig* to the process of the current task.
2817 * The signal may be delivered to any of this process's threads.
2819 * 0 on success or successfully queued.
2821 * **-EBUSY** if work queue under nmi is full.
2823 * **-EINVAL** if *sig* is invalid.
2825 * **-EPERM** if no permission to send the *sig*.
2827 * **-EAGAIN** if bpf program can try again.
2829 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2831 * Try to issue a SYN cookie for the packet with corresponding
2832 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2834 * *iph* points to the start of the IPv4 or IPv6 header, while
2835 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2836 * **sizeof**\ (**struct ip6hdr**).
2838 * *th* points to the start of the TCP header, while *th_len*
2839 * contains the length of the TCP header.
2841 * On success, lower 32 bits hold the generated SYN cookie in
2842 * followed by 16 bits which hold the MSS value for that cookie,
2843 * and the top 16 bits are unused.
2845 * On failure, the returned value is one of the following:
2847 * **-EINVAL** SYN cookie cannot be issued due to error
2849 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2851 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2853 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2855 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2857 * Write raw *data* blob into a special BPF perf event held by
2858 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2859 * event must have the following attributes: **PERF_SAMPLE_RAW**
2860 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2861 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2863 * The *flags* are used to indicate the index in *map* for which
2864 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2865 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2866 * to indicate that the index of the current CPU core should be
2869 * The value to write, of *size*, is passed through eBPF stack and
2870 * pointed by *data*.
2872 * *ctx* is a pointer to in-kernel struct sk_buff.
2874 * This helper is similar to **bpf_perf_event_output**\ () but
2875 * restricted to raw_tracepoint bpf programs.
2877 * 0 on success, or a negative error in case of failure.
2879 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2881 * Safely attempt to read *size* bytes from user space address
2882 * *unsafe_ptr* and store the data in *dst*.
2884 * 0 on success, or a negative error in case of failure.
2886 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2888 * Safely attempt to read *size* bytes from kernel space address
2889 * *unsafe_ptr* and store the data in *dst*.
2891 * 0 on success, or a negative error in case of failure.
2893 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2895 * Copy a NUL terminated string from an unsafe user address
2896 * *unsafe_ptr* to *dst*. The *size* should include the
2897 * terminating NUL byte. In case the string length is smaller than
2898 * *size*, the target is not padded with further NUL bytes. If the
2899 * string length is larger than *size*, just *size*-1 bytes are
2900 * copied and the last byte is set to NUL.
2902 * On success, the length of the copied string is returned. This
2903 * makes this helper useful in tracing programs for reading
2904 * strings, and more importantly to get its length at runtime. See
2905 * the following snippet:
2909 * SEC("kprobe/sys_open")
2910 * void bpf_sys_open(struct pt_regs *ctx)
2912 * char buf[PATHLEN]; // PATHLEN is defined to 256
2913 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2916 * // Consume buf, for example push it to
2917 * // userspace via bpf_perf_event_output(); we
2918 * // can use res (the string length) as event
2919 * // size, after checking its boundaries.
2922 * In comparison, using **bpf_probe_read_user**\ () helper here
2923 * instead to read the string would require to estimate the length
2924 * at compile time, and would often result in copying more memory
2927 * Another useful use case is when parsing individual process
2928 * arguments or individual environment variables navigating
2929 * *current*\ **->mm->arg_start** and *current*\
2930 * **->mm->env_start**: using this helper and the return value,
2931 * one can quickly iterate at the right offset of the memory area.
2933 * On success, the strictly positive length of the string,
2934 * including the trailing NUL character. On error, a negative
2937 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2939 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
2940 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
2942 * On success, the strictly positive length of the string, including
2943 * the trailing NUL character. On error, a negative value.
2945 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2947 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
2948 * *rcv_nxt* is the ack_seq to be sent out.
2950 * 0 on success, or a negative error in case of failure.
2952 * int bpf_send_signal_thread(u32 sig)
2954 * Send signal *sig* to the thread corresponding to the current task.
2956 * 0 on success or successfully queued.
2958 * **-EBUSY** if work queue under nmi is full.
2960 * **-EINVAL** if *sig* is invalid.
2962 * **-EPERM** if no permission to send the *sig*.
2964 * **-EAGAIN** if bpf program can try again.
2966 * u64 bpf_jiffies64(void)
2968 * Obtain the 64bit jiffies
2970 * The 64 bit jiffies
2972 * int bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
2974 * For an eBPF program attached to a perf event, retrieve the
2975 * branch records (**struct perf_branch_entry**) associated to *ctx*
2976 * and store it in the buffer pointed by *buf* up to size
2979 * On success, number of bytes written to *buf*. On error, a
2982 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
2983 * instead return the number of bytes required to store all the
2984 * branch entries. If this flag is set, *buf* may be NULL.
2986 * **-EINVAL** if arguments invalid or **size** not a multiple
2987 * of **sizeof**\ (**struct perf_branch_entry**\ ).
2989 * **-ENOENT** if architecture does not support branch records.
2991 * int bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
2993 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
2994 * *namespace* will be returned in *nsdata*.
2996 * 0 on success, or one of the following in case of failure:
2998 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
2999 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3001 * **-ENOENT** if pidns does not exists for the current task.
3003 * int bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3005 * Write raw *data* blob into a special BPF perf event held by
3006 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3007 * event must have the following attributes: **PERF_SAMPLE_RAW**
3008 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3009 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3011 * The *flags* are used to indicate the index in *map* for which
3012 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3013 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3014 * to indicate that the index of the current CPU core should be
3017 * The value to write, of *size*, is passed through eBPF stack and
3018 * pointed by *data*.
3020 * *ctx* is a pointer to in-kernel struct xdp_buff.
3022 * This helper is similar to **bpf_perf_eventoutput**\ () but
3023 * restricted to raw_tracepoint bpf programs.
3025 * 0 on success, or a negative error in case of failure.
3027 * u64 bpf_get_netns_cookie(void *ctx)
3029 * Retrieve the cookie (generated by the kernel) of the network
3030 * namespace the input *ctx* is associated with. The network
3031 * namespace cookie remains stable for its lifetime and provides
3032 * a global identifier that can be assumed unique. If *ctx* is
3033 * NULL, then the helper returns the cookie for the initial
3034 * network namespace. The cookie itself is very similar to that
3035 * of **bpf_get_socket_cookie**\ () helper, but for network
3036 * namespaces instead of sockets.
3038 * A 8-byte long opaque number.
3040 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3042 * Return id of cgroup v2 that is ancestor of the cgroup associated
3043 * with the current task at the *ancestor_level*. The root cgroup
3044 * is at *ancestor_level* zero and each step down the hierarchy
3045 * increments the level. If *ancestor_level* == level of cgroup
3046 * associated with the current task, then return value will be the
3047 * same as that of **bpf_get_current_cgroup_id**\ ().
3049 * The helper is useful to implement policies based on cgroups
3050 * that are upper in hierarchy than immediate cgroup associated
3051 * with the current task.
3053 * The format of returned id and helper limitations are same as in
3054 * **bpf_get_current_cgroup_id**\ ().
3056 * The id is returned or 0 in case the id could not be retrieved.
3058 * int bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags)
3060 * Assign the *sk* to the *skb*. When combined with appropriate
3061 * routing configuration to receive the packet towards the socket,
3062 * will cause *skb* to be delivered to the specified socket.
3063 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3064 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3065 * interfere with successful delivery to the socket.
3067 * This operation is only valid from TC ingress path.
3069 * The *flags* argument must be zero.
3071 * 0 on success, or a negative error in case of failure:
3073 * **-EINVAL** if specified *flags* are not supported.
3075 * **-ENOENT** if the socket is unavailable for assignment.
3077 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3079 * **-EOPNOTSUPP** if the operation is not supported, for example
3080 * a call from outside of TC ingress.
3082 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3085 * u64 bpf_ktime_get_boot_ns(void)
3087 * Return the time elapsed since system boot, in nanoseconds.
3088 * Does include the time the system was suspended.
3089 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
3093 * int bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3095 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3096 * out the format string.
3097 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3098 * the format string itself. The *data* and *data_len* are format string
3099 * arguments. The *data* are a **u64** array and corresponding format string
3100 * values are stored in the array. For strings and pointers where pointees
3101 * are accessed, only the pointer values are stored in the *data* array.
3102 * The *data_len* is the size of *data* in bytes.
3104 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3105 * Reading kernel memory may fail due to either invalid address or
3106 * valid address but requiring a major memory fault. If reading kernel memory
3107 * fails, the string for **%s** will be an empty string, and the ip
3108 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
3109 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
3111 * 0 on success, or a negative error in case of failure:
3113 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3114 * by returning 1 from bpf program.
3116 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3118 * **-E2BIG** if *fmt* contains too many format specifiers.
3120 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3122 * int bpf_seq_write(struct seq_file *m, const void *data, u32 len)
3124 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
3125 * The *m* represents the seq_file. The *data* and *len* represent the
3126 * data to write in bytes.
3128 * 0 on success, or a negative error in case of failure:
3130 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3132 * u64 bpf_sk_cgroup_id(struct bpf_sock *sk)
3134 * Return the cgroup v2 id of the socket *sk*.
3136 * *sk* must be a non-**NULL** pointer to a full socket, e.g. one
3137 * returned from **bpf_sk_lookup_xxx**\ (),
3138 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3139 * same as in **bpf_skb_cgroup_id**\ ().
3141 * This helper is available only if the kernel was compiled with
3142 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3144 * The id is returned or 0 in case the id could not be retrieved.
3146 * u64 bpf_sk_ancestor_cgroup_id(struct bpf_sock *sk, int ancestor_level)
3148 * Return id of cgroup v2 that is ancestor of cgroup associated
3149 * with the *sk* at the *ancestor_level*. The root cgroup is at
3150 * *ancestor_level* zero and each step down the hierarchy
3151 * increments the level. If *ancestor_level* == level of cgroup
3152 * associated with *sk*, then return value will be same as that
3153 * of **bpf_sk_cgroup_id**\ ().
3155 * The helper is useful to implement policies based on cgroups
3156 * that are upper in hierarchy than immediate cgroup associated
3159 * The format of returned id and helper limitations are same as in
3160 * **bpf_sk_cgroup_id**\ ().
3162 * The id is returned or 0 in case the id could not be retrieved.
3164 * void *bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
3166 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
3167 * If BPF_RB_NO_WAKEUP is specified in *flags*, no notification of
3168 * new data availability is sent.
3169 * IF BPF_RB_FORCE_WAKEUP is specified in *flags*, notification of
3170 * new data availability is sent unconditionally.
3175 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3177 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3179 * Valid pointer with *size* bytes of memory available; NULL,
3182 * void bpf_ringbuf_submit(void *data, u64 flags)
3184 * Submit reserved ring buffer sample, pointed to by *data*.
3185 * If BPF_RB_NO_WAKEUP is specified in *flags*, no notification of
3186 * new data availability is sent.
3187 * IF BPF_RB_FORCE_WAKEUP is specified in *flags*, notification of
3188 * new data availability is sent unconditionally.
3190 * Nothing. Always succeeds.
3192 * void bpf_ringbuf_discard(void *data, u64 flags)
3194 * Discard reserved ring buffer sample, pointed to by *data*.
3195 * If BPF_RB_NO_WAKEUP is specified in *flags*, no notification of
3196 * new data availability is sent.
3197 * IF BPF_RB_FORCE_WAKEUP is specified in *flags*, notification of
3198 * new data availability is sent unconditionally.
3200 * Nothing. Always succeeds.
3202 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3204 * Query various characteristics of provided ring buffer. What
3205 * exactly is queries is determined by *flags*:
3206 * - BPF_RB_AVAIL_DATA - amount of data not yet consumed;
3207 * - BPF_RB_RING_SIZE - the size of ring buffer;
3208 * - BPF_RB_CONS_POS - consumer position (can wrap around);
3209 * - BPF_RB_PROD_POS - producer(s) position (can wrap around);
3210 * Data returned is just a momentary snapshots of actual values
3211 * and could be inaccurate, so this facility should be used to
3212 * power heuristics and for reporting, not to make 100% correct
3215 * Requested value, or 0, if flags are not recognized.
3217 #define __BPF_FUNC_MAPPER(FN) \
3219 FN(map_lookup_elem), \
3220 FN(map_update_elem), \
3221 FN(map_delete_elem), \
3225 FN(get_prandom_u32), \
3226 FN(get_smp_processor_id), \
3227 FN(skb_store_bytes), \
3228 FN(l3_csum_replace), \
3229 FN(l4_csum_replace), \
3231 FN(clone_redirect), \
3232 FN(get_current_pid_tgid), \
3233 FN(get_current_uid_gid), \
3234 FN(get_current_comm), \
3235 FN(get_cgroup_classid), \
3236 FN(skb_vlan_push), \
3238 FN(skb_get_tunnel_key), \
3239 FN(skb_set_tunnel_key), \
3240 FN(perf_event_read), \
3242 FN(get_route_realm), \
3243 FN(perf_event_output), \
3244 FN(skb_load_bytes), \
3247 FN(skb_get_tunnel_opt), \
3248 FN(skb_set_tunnel_opt), \
3249 FN(skb_change_proto), \
3250 FN(skb_change_type), \
3251 FN(skb_under_cgroup), \
3252 FN(get_hash_recalc), \
3253 FN(get_current_task), \
3254 FN(probe_write_user), \
3255 FN(current_task_under_cgroup), \
3256 FN(skb_change_tail), \
3257 FN(skb_pull_data), \
3259 FN(set_hash_invalid), \
3260 FN(get_numa_node_id), \
3261 FN(skb_change_head), \
3262 FN(xdp_adjust_head), \
3263 FN(probe_read_str), \
3264 FN(get_socket_cookie), \
3265 FN(get_socket_uid), \
3268 FN(skb_adjust_room), \
3270 FN(sk_redirect_map), \
3271 FN(sock_map_update), \
3272 FN(xdp_adjust_meta), \
3273 FN(perf_event_read_value), \
3274 FN(perf_prog_read_value), \
3276 FN(override_return), \
3277 FN(sock_ops_cb_flags_set), \
3278 FN(msg_redirect_map), \
3279 FN(msg_apply_bytes), \
3280 FN(msg_cork_bytes), \
3281 FN(msg_pull_data), \
3283 FN(xdp_adjust_tail), \
3284 FN(skb_get_xfrm_state), \
3286 FN(skb_load_bytes_relative), \
3288 FN(sock_hash_update), \
3289 FN(msg_redirect_hash), \
3290 FN(sk_redirect_hash), \
3291 FN(lwt_push_encap), \
3292 FN(lwt_seg6_store_bytes), \
3293 FN(lwt_seg6_adjust_srh), \
3294 FN(lwt_seg6_action), \
3297 FN(skb_cgroup_id), \
3298 FN(get_current_cgroup_id), \
3299 FN(get_local_storage), \
3300 FN(sk_select_reuseport), \
3301 FN(skb_ancestor_cgroup_id), \
3302 FN(sk_lookup_tcp), \
3303 FN(sk_lookup_udp), \
3305 FN(map_push_elem), \
3307 FN(map_peek_elem), \
3308 FN(msg_push_data), \
3310 FN(rc_pointer_rel), \
3315 FN(skb_ecn_set_ce), \
3316 FN(get_listener_sock), \
3317 FN(skc_lookup_tcp), \
3318 FN(tcp_check_syncookie), \
3319 FN(sysctl_get_name), \
3320 FN(sysctl_get_current_value), \
3321 FN(sysctl_get_new_value), \
3322 FN(sysctl_set_new_value), \
3325 FN(sk_storage_get), \
3326 FN(sk_storage_delete), \
3328 FN(tcp_gen_syncookie), \
3330 FN(probe_read_user), \
3331 FN(probe_read_kernel), \
3332 FN(probe_read_user_str), \
3333 FN(probe_read_kernel_str), \
3335 FN(send_signal_thread), \
3337 FN(read_branch_records), \
3338 FN(get_ns_current_pid_tgid), \
3340 FN(get_netns_cookie), \
3341 FN(get_current_ancestor_cgroup_id), \
3343 FN(ktime_get_boot_ns), \
3347 FN(sk_ancestor_cgroup_id), \
3348 FN(ringbuf_output), \
3349 FN(ringbuf_reserve), \
3350 FN(ringbuf_submit), \
3351 FN(ringbuf_discard), \
3354 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
3355 * function eBPF program intends to call
3357 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
3359 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
3362 #undef __BPF_ENUM_FN
3364 /* All flags used by eBPF helper functions, placed here. */
3366 /* BPF_FUNC_skb_store_bytes flags. */
3368 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3369 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3372 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3373 * First 4 bits are for passing the header field size.
3376 BPF_F_HDR_FIELD_MASK = 0xfULL,
3379 /* BPF_FUNC_l4_csum_replace flags. */
3381 BPF_F_PSEUDO_HDR = (1ULL << 4),
3382 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3383 BPF_F_MARK_ENFORCE = (1ULL << 6),
3386 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
3388 BPF_F_INGRESS = (1ULL << 0),
3391 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
3393 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3396 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
3398 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3399 BPF_F_USER_STACK = (1ULL << 8),
3400 /* flags used by BPF_FUNC_get_stackid only. */
3401 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3402 BPF_F_REUSE_STACKID = (1ULL << 10),
3403 /* flags used by BPF_FUNC_get_stack only. */
3404 BPF_F_USER_BUILD_ID = (1ULL << 11),
3407 /* BPF_FUNC_skb_set_tunnel_key flags. */
3409 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3410 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3411 BPF_F_SEQ_NUMBER = (1ULL << 3),
3414 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3415 * BPF_FUNC_perf_event_read_value flags.
3418 BPF_F_INDEX_MASK = 0xffffffffULL,
3419 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
3420 /* BPF_FUNC_perf_event_output for sk_buff input context. */
3421 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3424 /* Current network namespace */
3426 BPF_F_CURRENT_NETNS = (-1L),
3429 /* BPF_FUNC_skb_adjust_room flags. */
3431 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3432 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3433 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3434 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3435 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
3439 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3440 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3443 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
3444 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3445 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
3447 /* BPF_FUNC_sysctl_get_name flags. */
3449 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3452 /* BPF_FUNC_sk_storage_get flags */
3454 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0),
3457 /* BPF_FUNC_read_branch_records flags. */
3459 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3462 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
3463 * BPF_FUNC_bpf_ringbuf_output flags.
3466 BPF_RB_NO_WAKEUP = (1ULL << 0),
3467 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
3470 /* BPF_FUNC_bpf_ringbuf_query flags */
3472 BPF_RB_AVAIL_DATA = 0,
3473 BPF_RB_RING_SIZE = 1,
3474 BPF_RB_CONS_POS = 2,
3475 BPF_RB_PROD_POS = 3,
3478 /* BPF ring buffer constants */
3480 BPF_RINGBUF_BUSY_BIT = (1U << 31),
3481 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
3482 BPF_RINGBUF_HDR_SZ = 8,
3485 /* Mode for BPF_FUNC_skb_adjust_room helper. */
3486 enum bpf_adj_room_mode {
3491 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3492 enum bpf_hdr_start_off {
3497 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3498 enum bpf_lwt_encap_mode {
3500 BPF_LWT_ENCAP_SEG6_INLINE,
3504 #define __bpf_md_ptr(type, name) \
3508 } __attribute__((aligned(8)))
3510 /* user accessible mirror of in-kernel sk_buff.
3511 * new fields can only be added to the end of this structure
3517 __u32 queue_mapping;
3523 __u32 ingress_ifindex;
3533 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
3535 __u32 remote_ip4; /* Stored in network byte order */
3536 __u32 local_ip4; /* Stored in network byte order */
3537 __u32 remote_ip6[4]; /* Stored in network byte order */
3538 __u32 local_ip6[4]; /* Stored in network byte order */
3539 __u32 remote_port; /* Stored in network byte order */
3540 __u32 local_port; /* stored in host byte order */
3544 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
3548 __bpf_md_ptr(struct bpf_sock *, sk);
3552 struct bpf_tunnel_key {
3556 __u32 remote_ipv6[4];
3560 __u16 tunnel_ext; /* Padding, future use. */
3564 /* user accessible mirror of in-kernel xfrm_state.
3565 * new fields can only be added to the end of this structure
3567 struct bpf_xfrm_state {
3569 __u32 spi; /* Stored in network byte order */
3571 __u16 ext; /* Padding, future use. */
3573 __u32 remote_ipv4; /* Stored in network byte order */
3574 __u32 remote_ipv6[4]; /* Stored in network byte order */
3578 /* Generic BPF return codes which all BPF program types may support.
3579 * The values are binary compatible with their TC_ACT_* counter-part to
3580 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3583 * XDP is handled seprately, see XDP_*.
3591 /* >127 are reserved for prog type specific return codes.
3593 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3594 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3595 * changed and should be routed based on its new L3 header.
3596 * (This is an L3 redirect, as opposed to L2 redirect
3597 * represented by BPF_REDIRECT above).
3599 BPF_LWT_REROUTE = 128,
3609 /* IP address also allows 1 and 2 bytes access */
3612 __u32 src_port; /* host byte order */
3613 __u32 dst_port; /* network byte order */
3617 __s32 rx_queue_mapping;
3620 struct bpf_tcp_sock {
3621 __u32 snd_cwnd; /* Sending congestion window */
3622 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3624 __u32 snd_ssthresh; /* Slow start size threshold */
3625 __u32 rcv_nxt; /* What we want to receive next */
3626 __u32 snd_nxt; /* Next sequence we send */
3627 __u32 snd_una; /* First byte we want an ack for */
3628 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3629 __u32 ecn_flags; /* ECN status bits. */
3630 __u32 rate_delivered; /* saved rate sample: packets delivered */
3631 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3632 __u32 packets_out; /* Packets which are "in flight" */
3633 __u32 retrans_out; /* Retransmitted packets out */
3634 __u32 total_retrans; /* Total retransmits for entire connection */
3635 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3636 * total number of segments in.
3638 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3639 * total number of data segments in.
3641 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3642 * The total number of segments sent.
3644 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3645 * total number of data segments sent.
3647 __u32 lost_out; /* Lost packets */
3648 __u32 sacked_out; /* SACK'd packets */
3649 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3650 * sum(delta(rcv_nxt)), or how many bytes
3653 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3654 * sum(delta(snd_una)), or how many bytes
3657 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3658 * total number of DSACK blocks received
3660 __u32 delivered; /* Total data packets delivered incl. rexmits */
3661 __u32 delivered_ce; /* Like the above but only ECE marked packets */
3662 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
3665 struct bpf_sock_tuple {
3682 struct bpf_xdp_sock {
3686 #define XDP_PACKET_HEADROOM 256
3688 /* User return codes for XDP prog type.
3689 * A valid XDP program must return one of these defined values. All other
3690 * return codes are reserved for future use. Unknown return codes will
3691 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
3701 /* user accessible metadata for XDP packet hook
3702 * new fields must be added to the end of this structure
3708 /* Below access go through struct xdp_rxq_info */
3709 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3710 __u32 rx_queue_index; /* rxq->queue_index */
3712 __u32 egress_ifindex; /* txq->dev->ifindex */
3720 /* user accessible metadata for SK_MSG packet hook, new fields must
3721 * be added to the end of this structure
3724 __bpf_md_ptr(void *, data);
3725 __bpf_md_ptr(void *, data_end);
3728 __u32 remote_ip4; /* Stored in network byte order */
3729 __u32 local_ip4; /* Stored in network byte order */
3730 __u32 remote_ip6[4]; /* Stored in network byte order */
3731 __u32 local_ip6[4]; /* Stored in network byte order */
3732 __u32 remote_port; /* Stored in network byte order */
3733 __u32 local_port; /* stored in host byte order */
3734 __u32 size; /* Total size of sk_msg */
3736 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
3739 struct sk_reuseport_md {
3741 * Start of directly accessible data. It begins from
3742 * the tcp/udp header.
3744 __bpf_md_ptr(void *, data);
3745 /* End of directly accessible data */
3746 __bpf_md_ptr(void *, data_end);
3748 * Total length of packet (starting from the tcp/udp header).
3749 * Note that the directly accessible bytes (data_end - data)
3750 * could be less than this "len". Those bytes could be
3751 * indirectly read by a helper "bpf_skb_load_bytes()".
3755 * Eth protocol in the mac header (network byte order). e.g.
3756 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3759 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3760 __u32 bind_inany; /* Is sock bound to an INANY address? */
3761 __u32 hash; /* A hash of the packet 4 tuples */
3764 #define BPF_TAG_SIZE 8
3766 struct bpf_prog_info {
3769 __u8 tag[BPF_TAG_SIZE];
3770 __u32 jited_prog_len;
3771 __u32 xlated_prog_len;
3772 __aligned_u64 jited_prog_insns;
3773 __aligned_u64 xlated_prog_insns;
3774 __u64 load_time; /* ns since boottime */
3775 __u32 created_by_uid;
3777 __aligned_u64 map_ids;
3778 char name[BPF_OBJ_NAME_LEN];
3780 __u32 gpl_compatible:1;
3781 __u32 :31; /* alignment pad */
3784 __u32 nr_jited_ksyms;
3785 __u32 nr_jited_func_lens;
3786 __aligned_u64 jited_ksyms;
3787 __aligned_u64 jited_func_lens;
3789 __u32 func_info_rec_size;
3790 __aligned_u64 func_info;
3793 __aligned_u64 line_info;
3794 __aligned_u64 jited_line_info;
3795 __u32 nr_jited_line_info;
3796 __u32 line_info_rec_size;
3797 __u32 jited_line_info_rec_size;
3799 __aligned_u64 prog_tags;
3802 } __attribute__((aligned(8)));
3804 struct bpf_map_info {
3811 char name[BPF_OBJ_NAME_LEN];
3813 __u32 btf_vmlinux_value_type_id;
3817 __u32 btf_key_type_id;
3818 __u32 btf_value_type_id;
3819 } __attribute__((aligned(8)));
3821 struct bpf_btf_info {
3825 } __attribute__((aligned(8)));
3827 struct bpf_link_info {
3833 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
3834 __u32 tp_name_len; /* in/out: tp_name buffer len */
3848 } __attribute__((aligned(8)));
3850 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3851 * by user and intended to be used by socket (e.g. to bind to, depends on
3852 * attach attach type).
3854 struct bpf_sock_addr {
3855 __u32 user_family; /* Allows 4-byte read, but no write. */
3856 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3857 * Stored in network byte order.
3859 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3860 * Stored in network byte order.
3862 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
3863 * Stored in network byte order
3865 __u32 family; /* Allows 4-byte read, but no write */
3866 __u32 type; /* Allows 4-byte read, but no write */
3867 __u32 protocol; /* Allows 4-byte read, but no write */
3868 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3869 * Stored in network byte order.
3871 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3872 * Stored in network byte order.
3874 __bpf_md_ptr(struct bpf_sock *, sk);
3877 /* User bpf_sock_ops struct to access socket values and specify request ops
3878 * and their replies.
3879 * Some of this fields are in network (bigendian) byte order and may need
3880 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3881 * New fields can only be added at the end of this structure
3883 struct bpf_sock_ops {
3886 __u32 args[4]; /* Optionally passed to bpf program */
3887 __u32 reply; /* Returned by bpf program */
3888 __u32 replylong[4]; /* Optionally returned by bpf prog */
3891 __u32 remote_ip4; /* Stored in network byte order */
3892 __u32 local_ip4; /* Stored in network byte order */
3893 __u32 remote_ip6[4]; /* Stored in network byte order */
3894 __u32 local_ip6[4]; /* Stored in network byte order */
3895 __u32 remote_port; /* Stored in network byte order */
3896 __u32 local_port; /* stored in host byte order */
3897 __u32 is_fullsock; /* Some TCP fields are only valid if
3898 * there is a full socket. If not, the
3899 * fields read as zero.
3902 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
3903 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3912 __u32 rate_delivered;
3913 __u32 rate_interval_us;
3916 __u32 total_retrans;
3920 __u32 data_segs_out;
3924 __u64 bytes_received;
3926 __bpf_md_ptr(struct bpf_sock *, sk);
3929 /* Definitions for bpf_sock_ops_cb_flags */
3931 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
3932 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
3933 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
3934 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
3935 /* Mask of all currently supported cb flags */
3936 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF,
3939 /* List of known BPF sock_ops operators.
3940 * New entries can only be added at the end
3944 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3945 * -1 if default value should be used
3947 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3948 * window (in packets) or -1 if default
3949 * value should be used
3951 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3952 * active connection is initialized
3954 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3955 * active connection is
3958 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3959 * passive connection is
3962 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3965 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3966 * based on the path and may be
3967 * dependent on the congestion control
3968 * algorithm. In general it indicates
3969 * a congestion threshold. RTTs above
3970 * this indicate congestion
3972 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3973 * Arg1: value of icsk_retransmits
3974 * Arg2: value of icsk_rto
3975 * Arg3: whether RTO has expired
3977 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3978 * Arg1: sequence number of 1st byte
3980 * Arg3: return value of
3981 * tcp_transmit_skb (0 => success)
3983 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3987 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3988 * socket transition to LISTEN state.
3990 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3994 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3995 * changes between the TCP and BPF versions. Ideally this should never happen.
3996 * If it does, we need to add code to convert them before calling
3997 * the BPF sock_ops function.
4000 BPF_TCP_ESTABLISHED = 1,
4010 BPF_TCP_CLOSING, /* Now a valid state */
4011 BPF_TCP_NEW_SYN_RECV,
4013 BPF_TCP_MAX_STATES /* Leave at the end! */
4017 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4018 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
4021 struct bpf_perf_event_value {
4028 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
4029 BPF_DEVCG_ACC_READ = (1ULL << 1),
4030 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
4034 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
4035 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
4038 struct bpf_cgroup_dev_ctx {
4039 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
4045 struct bpf_raw_tracepoint_args {
4049 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
4050 * OUTPUT: Do lookup from egress perspective; default is ingress
4053 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
4054 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
4058 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
4059 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
4060 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
4061 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
4062 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
4063 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
4064 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
4065 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
4066 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
4069 struct bpf_fib_lookup {
4070 /* input: network family for lookup (AF_INET, AF_INET6)
4071 * output: network family of egress nexthop
4075 /* set if lookup is to consider L4 data - e.g., FIB rules */
4080 /* total length of packet from network header - used for MTU check */
4083 /* input: L3 device index for lookup
4084 * output: device index from FIB lookup
4089 /* inputs to lookup */
4090 __u8 tos; /* AF_INET */
4091 __be32 flowinfo; /* AF_INET6, flow_label + priority */
4093 /* output: metric of fib result (IPv4/IPv6 only) */
4099 __u32 ipv6_src[4]; /* in6_addr; network order */
4102 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
4103 * network header. output: bpf_fib_lookup sets to gateway address
4104 * if FIB lookup returns gateway route
4108 __u32 ipv6_dst[4]; /* in6_addr; network order */
4112 __be16 h_vlan_proto;
4114 __u8 smac[6]; /* ETH_ALEN */
4115 __u8 dmac[6]; /* ETH_ALEN */
4118 enum bpf_task_fd_type {
4119 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
4120 BPF_FD_TYPE_TRACEPOINT, /* tp name */
4121 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
4122 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
4123 BPF_FD_TYPE_UPROBE, /* filename + offset */
4124 BPF_FD_TYPE_URETPROBE, /* filename + offset */
4128 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
4129 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
4130 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
4133 struct bpf_flow_keys {
4136 __u16 addr_proto; /* ETH_P_* of valid addrs */
4150 __u32 ipv6_src[4]; /* in6_addr; network order */
4151 __u32 ipv6_dst[4]; /* in6_addr; network order */
4158 struct bpf_func_info {
4163 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
4164 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
4166 struct bpf_line_info {
4168 __u32 file_name_off;
4173 struct bpf_spin_lock {
4178 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
4179 * Allows 1,2,4-byte read, but no write.
4181 __u32 file_pos; /* Sysctl file position to read from, write to.
4182 * Allows 1,2,4-byte read an 4-byte write.
4186 struct bpf_sockopt {
4187 __bpf_md_ptr(struct bpf_sock *, sk);
4188 __bpf_md_ptr(void *, optval);
4189 __bpf_md_ptr(void *, optval_end);
4197 struct bpf_pidns_info {
4201 #endif /* _UAPI__LINUX_BPF_H__ */
git.samba.org / sfrench / cifs-2.6.git / blob