1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_XADD 0xc0 /* exclusive add */
25 #define BPF_MOV 0xb0 /* mov reg to reg */
26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28 /* change endianness of a register */
29 #define BPF_END 0xd0 /* flags for endianness conversion: */
30 #define BPF_TO_LE 0x00 /* convert to little-endian */
31 #define BPF_TO_BE 0x08 /* convert to big-endian */
32 #define BPF_FROM_LE BPF_TO_LE
33 #define BPF_FROM_BE BPF_TO_BE
36 #define BPF_JNE 0x50 /* jump != */
37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
43 #define BPF_CALL 0x80 /* function call */
44 #define BPF_EXIT 0x90 /* function return */
46 /* Register numbers */
62 /* BPF has 10 general purpose 64-bit registers and stack frame. */
63 #define MAX_BPF_REG __MAX_BPF_REG
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74 struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data[0]; /* Arbitrary size */
79 struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
84 union bpf_iter_link_info {
90 /* BPF syscall commands, see bpf(2) man-page for details. */
103 BPF_PROG_GET_NEXT_ID,
105 BPF_PROG_GET_FD_BY_ID,
106 BPF_MAP_GET_FD_BY_ID,
107 BPF_OBJ_GET_INFO_BY_FD,
109 BPF_RAW_TRACEPOINT_OPEN,
111 BPF_BTF_GET_FD_BY_ID,
113 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
116 BPF_MAP_LOOKUP_BATCH,
117 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
118 BPF_MAP_UPDATE_BATCH,
119 BPF_MAP_DELETE_BATCH,
122 BPF_LINK_GET_FD_BY_ID,
123 BPF_LINK_GET_NEXT_ID,
134 BPF_MAP_TYPE_PROG_ARRAY,
135 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
136 BPF_MAP_TYPE_PERCPU_HASH,
137 BPF_MAP_TYPE_PERCPU_ARRAY,
138 BPF_MAP_TYPE_STACK_TRACE,
139 BPF_MAP_TYPE_CGROUP_ARRAY,
140 BPF_MAP_TYPE_LRU_HASH,
141 BPF_MAP_TYPE_LRU_PERCPU_HASH,
142 BPF_MAP_TYPE_LPM_TRIE,
143 BPF_MAP_TYPE_ARRAY_OF_MAPS,
144 BPF_MAP_TYPE_HASH_OF_MAPS,
146 BPF_MAP_TYPE_SOCKMAP,
149 BPF_MAP_TYPE_SOCKHASH,
150 BPF_MAP_TYPE_CGROUP_STORAGE,
151 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
152 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
155 BPF_MAP_TYPE_SK_STORAGE,
156 BPF_MAP_TYPE_DEVMAP_HASH,
157 BPF_MAP_TYPE_STRUCT_OPS,
158 BPF_MAP_TYPE_RINGBUF,
159 BPF_MAP_TYPE_INODE_STORAGE,
162 /* Note that tracing related programs such as
163 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
164 * are not subject to a stable API since kernel internal data
165 * structures can change from release to release and may
166 * therefore break existing tracing BPF programs. Tracing BPF
167 * programs correspond to /a/ specific kernel which is to be
168 * analyzed, and not /a/ specific kernel /and/ all future ones.
171 BPF_PROG_TYPE_UNSPEC,
172 BPF_PROG_TYPE_SOCKET_FILTER,
173 BPF_PROG_TYPE_KPROBE,
174 BPF_PROG_TYPE_SCHED_CLS,
175 BPF_PROG_TYPE_SCHED_ACT,
176 BPF_PROG_TYPE_TRACEPOINT,
178 BPF_PROG_TYPE_PERF_EVENT,
179 BPF_PROG_TYPE_CGROUP_SKB,
180 BPF_PROG_TYPE_CGROUP_SOCK,
181 BPF_PROG_TYPE_LWT_IN,
182 BPF_PROG_TYPE_LWT_OUT,
183 BPF_PROG_TYPE_LWT_XMIT,
184 BPF_PROG_TYPE_SOCK_OPS,
185 BPF_PROG_TYPE_SK_SKB,
186 BPF_PROG_TYPE_CGROUP_DEVICE,
187 BPF_PROG_TYPE_SK_MSG,
188 BPF_PROG_TYPE_RAW_TRACEPOINT,
189 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
190 BPF_PROG_TYPE_LWT_SEG6LOCAL,
191 BPF_PROG_TYPE_LIRC_MODE2,
192 BPF_PROG_TYPE_SK_REUSEPORT,
193 BPF_PROG_TYPE_FLOW_DISSECTOR,
194 BPF_PROG_TYPE_CGROUP_SYSCTL,
195 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
196 BPF_PROG_TYPE_CGROUP_SOCKOPT,
197 BPF_PROG_TYPE_TRACING,
198 BPF_PROG_TYPE_STRUCT_OPS,
201 BPF_PROG_TYPE_SK_LOOKUP,
204 enum bpf_attach_type {
205 BPF_CGROUP_INET_INGRESS,
206 BPF_CGROUP_INET_EGRESS,
207 BPF_CGROUP_INET_SOCK_CREATE,
209 BPF_SK_SKB_STREAM_PARSER,
210 BPF_SK_SKB_STREAM_VERDICT,
213 BPF_CGROUP_INET4_BIND,
214 BPF_CGROUP_INET6_BIND,
215 BPF_CGROUP_INET4_CONNECT,
216 BPF_CGROUP_INET6_CONNECT,
217 BPF_CGROUP_INET4_POST_BIND,
218 BPF_CGROUP_INET6_POST_BIND,
219 BPF_CGROUP_UDP4_SENDMSG,
220 BPF_CGROUP_UDP6_SENDMSG,
224 BPF_CGROUP_UDP4_RECVMSG,
225 BPF_CGROUP_UDP6_RECVMSG,
226 BPF_CGROUP_GETSOCKOPT,
227 BPF_CGROUP_SETSOCKOPT,
234 BPF_CGROUP_INET4_GETPEERNAME,
235 BPF_CGROUP_INET6_GETPEERNAME,
236 BPF_CGROUP_INET4_GETSOCKNAME,
237 BPF_CGROUP_INET6_GETSOCKNAME,
239 BPF_CGROUP_INET_SOCK_RELEASE,
243 __MAX_BPF_ATTACH_TYPE
246 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
249 BPF_LINK_TYPE_UNSPEC = 0,
250 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
251 BPF_LINK_TYPE_TRACING = 2,
252 BPF_LINK_TYPE_CGROUP = 3,
253 BPF_LINK_TYPE_ITER = 4,
254 BPF_LINK_TYPE_NETNS = 5,
255 BPF_LINK_TYPE_XDP = 6,
260 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
262 * NONE(default): No further bpf programs allowed in the subtree.
264 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
265 * the program in this cgroup yields to sub-cgroup program.
267 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
268 * that cgroup program gets run in addition to the program in this cgroup.
270 * Only one program is allowed to be attached to a cgroup with
271 * NONE or BPF_F_ALLOW_OVERRIDE flag.
272 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
273 * release old program and attach the new one. Attach flags has to match.
275 * Multiple programs are allowed to be attached to a cgroup with
276 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
277 * (those that were attached first, run first)
278 * The programs of sub-cgroup are executed first, then programs of
279 * this cgroup and then programs of parent cgroup.
280 * When children program makes decision (like picking TCP CA or sock bind)
281 * parent program has a chance to override it.
283 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
284 * programs for a cgroup. Though it's possible to replace an old program at
285 * any position by also specifying BPF_F_REPLACE flag and position itself in
286 * replace_bpf_fd attribute. Old program at this position will be released.
288 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
289 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
291 * cgrp1 (MULTI progs A, B) ->
292 * cgrp2 (OVERRIDE prog C) ->
293 * cgrp3 (MULTI prog D) ->
294 * cgrp4 (OVERRIDE prog E) ->
295 * cgrp5 (NONE prog F)
296 * the event in cgrp5 triggers execution of F,D,A,B in that order.
297 * if prog F is detached, the execution is E,D,A,B
298 * if prog F and D are detached, the execution is E,A,B
299 * if prog F, E and D are detached, the execution is C,A,B
301 * All eligible programs are executed regardless of return code from
304 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
305 #define BPF_F_ALLOW_MULTI (1U << 1)
306 #define BPF_F_REPLACE (1U << 2)
308 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
309 * verifier will perform strict alignment checking as if the kernel
310 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
311 * and NET_IP_ALIGN defined to 2.
313 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
315 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
316 * verifier will allow any alignment whatsoever. On platforms
317 * with strict alignment requirements for loads ands stores (such
318 * as sparc and mips) the verifier validates that all loads and
319 * stores provably follow this requirement. This flag turns that
320 * checking and enforcement off.
322 * It is mostly used for testing when we want to validate the
323 * context and memory access aspects of the verifier, but because
324 * of an unaligned access the alignment check would trigger before
325 * the one we are interested in.
327 #define BPF_F_ANY_ALIGNMENT (1U << 1)
329 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
330 * Verifier does sub-register def/use analysis and identifies instructions whose
331 * def only matters for low 32-bit, high 32-bit is never referenced later
332 * through implicit zero extension. Therefore verifier notifies JIT back-ends
333 * that it is safe to ignore clearing high 32-bit for these instructions. This
334 * saves some back-ends a lot of code-gen. However such optimization is not
335 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
336 * hence hasn't used verifier's analysis result. But, we really want to have a
337 * way to be able to verify the correctness of the described optimization on
338 * x86_64 on which testsuites are frequently exercised.
340 * So, this flag is introduced. Once it is set, verifier will randomize high
341 * 32-bit for those instructions who has been identified as safe to ignore them.
342 * Then, if verifier is not doing correct analysis, such randomization will
343 * regress tests to expose bugs.
345 #define BPF_F_TEST_RND_HI32 (1U << 2)
347 /* The verifier internal test flag. Behavior is undefined */
348 #define BPF_F_TEST_STATE_FREQ (1U << 3)
350 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
351 * restrict map and helper usage for such programs. Sleepable BPF programs can
352 * only be attached to hooks where kernel execution context allows sleeping.
353 * Such programs are allowed to use helpers that may sleep like
354 * bpf_copy_from_user().
356 #define BPF_F_SLEEPABLE (1U << 4)
358 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
359 * the following extensions:
361 * insn[0].src_reg: BPF_PSEUDO_MAP_FD
362 * insn[0].imm: map fd
366 * ldimm64 rewrite: address of map
367 * verifier type: CONST_PTR_TO_MAP
369 #define BPF_PSEUDO_MAP_FD 1
370 /* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE
371 * insn[0].imm: map fd
372 * insn[1].imm: offset into value
375 * ldimm64 rewrite: address of map[0]+offset
376 * verifier type: PTR_TO_MAP_VALUE
378 #define BPF_PSEUDO_MAP_VALUE 2
379 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID
380 * insn[0].imm: kernel btd id of VAR
384 * ldimm64 rewrite: address of the kernel variable
385 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
388 #define BPF_PSEUDO_BTF_ID 3
390 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
391 * offset to another bpf function
393 #define BPF_PSEUDO_CALL 1
395 /* flags for BPF_MAP_UPDATE_ELEM command */
397 BPF_ANY = 0, /* create new element or update existing */
398 BPF_NOEXIST = 1, /* create new element if it didn't exist */
399 BPF_EXIST = 2, /* update existing element */
400 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
403 /* flags for BPF_MAP_CREATE command */
405 BPF_F_NO_PREALLOC = (1U << 0),
406 /* Instead of having one common LRU list in the
407 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
408 * which can scale and perform better.
409 * Note, the LRU nodes (including free nodes) cannot be moved
410 * across different LRU lists.
412 BPF_F_NO_COMMON_LRU = (1U << 1),
413 /* Specify numa node during map creation */
414 BPF_F_NUMA_NODE = (1U << 2),
416 /* Flags for accessing BPF object from syscall side. */
417 BPF_F_RDONLY = (1U << 3),
418 BPF_F_WRONLY = (1U << 4),
420 /* Flag for stack_map, store build_id+offset instead of pointer */
421 BPF_F_STACK_BUILD_ID = (1U << 5),
423 /* Zero-initialize hash function seed. This should only be used for testing. */
424 BPF_F_ZERO_SEED = (1U << 6),
426 /* Flags for accessing BPF object from program side. */
427 BPF_F_RDONLY_PROG = (1U << 7),
428 BPF_F_WRONLY_PROG = (1U << 8),
430 /* Clone map from listener for newly accepted socket */
431 BPF_F_CLONE = (1U << 9),
433 /* Enable memory-mapping BPF map */
434 BPF_F_MMAPABLE = (1U << 10),
436 /* Share perf_event among processes */
437 BPF_F_PRESERVE_ELEMS = (1U << 11),
440 /* Flags for BPF_PROG_QUERY. */
442 /* Query effective (directly attached + inherited from ancestor cgroups)
443 * programs that will be executed for events within a cgroup.
444 * attach_flags with this flag are returned only for directly attached programs.
446 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
448 /* Flags for BPF_PROG_TEST_RUN */
450 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
451 #define BPF_F_TEST_RUN_ON_CPU (1U << 0)
453 /* type for BPF_ENABLE_STATS */
454 enum bpf_stats_type {
455 /* enabled run_time_ns and run_cnt */
456 BPF_STATS_RUN_TIME = 0,
459 enum bpf_stack_build_id_status {
460 /* user space need an empty entry to identify end of a trace */
461 BPF_STACK_BUILD_ID_EMPTY = 0,
462 /* with valid build_id and offset */
463 BPF_STACK_BUILD_ID_VALID = 1,
464 /* couldn't get build_id, fallback to ip */
465 BPF_STACK_BUILD_ID_IP = 2,
468 #define BPF_BUILD_ID_SIZE 20
469 struct bpf_stack_build_id {
471 unsigned char build_id[BPF_BUILD_ID_SIZE];
478 #define BPF_OBJ_NAME_LEN 16U
481 struct { /* anonymous struct used by BPF_MAP_CREATE command */
482 __u32 map_type; /* one of enum bpf_map_type */
483 __u32 key_size; /* size of key in bytes */
484 __u32 value_size; /* size of value in bytes */
485 __u32 max_entries; /* max number of entries in a map */
486 __u32 map_flags; /* BPF_MAP_CREATE related
487 * flags defined above.
489 __u32 inner_map_fd; /* fd pointing to the inner map */
490 __u32 numa_node; /* numa node (effective only if
491 * BPF_F_NUMA_NODE is set).
493 char map_name[BPF_OBJ_NAME_LEN];
494 __u32 map_ifindex; /* ifindex of netdev to create on */
495 __u32 btf_fd; /* fd pointing to a BTF type data */
496 __u32 btf_key_type_id; /* BTF type_id of the key */
497 __u32 btf_value_type_id; /* BTF type_id of the value */
498 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
499 * struct stored as the
504 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
509 __aligned_u64 next_key;
514 struct { /* struct used by BPF_MAP_*_BATCH commands */
515 __aligned_u64 in_batch; /* start batch,
516 * NULL to start from beginning
518 __aligned_u64 out_batch; /* output: next start batch */
520 __aligned_u64 values;
521 __u32 count; /* input/output:
522 * input: # of key/value
524 * output: # of filled elements
531 struct { /* anonymous struct used by BPF_PROG_LOAD command */
532 __u32 prog_type; /* one of enum bpf_prog_type */
535 __aligned_u64 license;
536 __u32 log_level; /* verbosity level of verifier */
537 __u32 log_size; /* size of user buffer */
538 __aligned_u64 log_buf; /* user supplied buffer */
539 __u32 kern_version; /* not used */
541 char prog_name[BPF_OBJ_NAME_LEN];
542 __u32 prog_ifindex; /* ifindex of netdev to prep for */
543 /* For some prog types expected attach type must be known at
544 * load time to verify attach type specific parts of prog
545 * (context accesses, allowed helpers, etc).
547 __u32 expected_attach_type;
548 __u32 prog_btf_fd; /* fd pointing to BTF type data */
549 __u32 func_info_rec_size; /* userspace bpf_func_info size */
550 __aligned_u64 func_info; /* func info */
551 __u32 func_info_cnt; /* number of bpf_func_info records */
552 __u32 line_info_rec_size; /* userspace bpf_line_info size */
553 __aligned_u64 line_info; /* line info */
554 __u32 line_info_cnt; /* number of bpf_line_info records */
555 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
556 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
559 struct { /* anonymous struct used by BPF_OBJ_* commands */
560 __aligned_u64 pathname;
565 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
566 __u32 target_fd; /* container object to attach to */
567 __u32 attach_bpf_fd; /* eBPF program to attach */
570 __u32 replace_bpf_fd; /* previously attached eBPF
571 * program to replace if
572 * BPF_F_REPLACE is used
576 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
579 __u32 data_size_in; /* input: len of data_in */
580 __u32 data_size_out; /* input/output: len of data_out
581 * returns ENOSPC if data_out
584 __aligned_u64 data_in;
585 __aligned_u64 data_out;
588 __u32 ctx_size_in; /* input: len of ctx_in */
589 __u32 ctx_size_out; /* input/output: len of ctx_out
590 * returns ENOSPC if ctx_out
593 __aligned_u64 ctx_in;
594 __aligned_u64 ctx_out;
599 struct { /* anonymous struct used by BPF_*_GET_*_ID */
611 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
617 struct { /* anonymous struct used by BPF_PROG_QUERY command */
618 __u32 target_fd; /* container object to query */
622 __aligned_u64 prog_ids;
626 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
631 struct { /* anonymous struct for BPF_BTF_LOAD */
633 __aligned_u64 btf_log_buf;
640 __u32 pid; /* input: pid */
641 __u32 fd; /* input: fd */
642 __u32 flags; /* input: flags */
643 __u32 buf_len; /* input/output: buf len */
644 __aligned_u64 buf; /* input/output:
645 * tp_name for tracepoint
647 * filename for uprobe
649 __u32 prog_id; /* output: prod_id */
650 __u32 fd_type; /* output: BPF_FD_TYPE_* */
651 __u64 probe_offset; /* output: probe_offset */
652 __u64 probe_addr; /* output: probe_addr */
655 struct { /* struct used by BPF_LINK_CREATE command */
656 __u32 prog_fd; /* eBPF program to attach */
658 __u32 target_fd; /* object to attach to */
659 __u32 target_ifindex; /* target ifindex */
661 __u32 attach_type; /* attach type */
662 __u32 flags; /* extra flags */
664 __u32 target_btf_id; /* btf_id of target to attach to */
666 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
667 __u32 iter_info_len; /* iter_info length */
672 struct { /* struct used by BPF_LINK_UPDATE command */
673 __u32 link_fd; /* link fd */
674 /* new program fd to update link with */
676 __u32 flags; /* extra flags */
677 /* expected link's program fd; is specified only if
678 * BPF_F_REPLACE flag is set in flags */
686 struct { /* struct used by BPF_ENABLE_STATS command */
690 struct { /* struct used by BPF_ITER_CREATE command */
695 struct { /* struct used by BPF_PROG_BIND_MAP command */
698 __u32 flags; /* extra flags */
701 } __attribute__((aligned(8)));
703 /* The description below is an attempt at providing documentation to eBPF
704 * developers about the multiple available eBPF helper functions. It can be
705 * parsed and used to produce a manual page. The workflow is the following,
706 * and requires the rst2man utility:
708 * $ ./scripts/bpf_helpers_doc.py \
709 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
710 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
711 * $ man /tmp/bpf-helpers.7
713 * Note that in order to produce this external documentation, some RST
714 * formatting is used in the descriptions to get "bold" and "italics" in
715 * manual pages. Also note that the few trailing white spaces are
716 * intentional, removing them would break paragraphs for rst2man.
718 * Start of BPF helper function descriptions:
720 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
722 * Perform a lookup in *map* for an entry associated to *key*.
724 * Map value associated to *key*, or **NULL** if no entry was
727 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
729 * Add or update the value of the entry associated to *key* in
730 * *map* with *value*. *flags* is one of:
733 * The entry for *key* must not exist in the map.
735 * The entry for *key* must already exist in the map.
737 * No condition on the existence of the entry for *key*.
739 * Flag value **BPF_NOEXIST** cannot be used for maps of types
740 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
741 * elements always exist), the helper would return an error.
743 * 0 on success, or a negative error in case of failure.
745 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
747 * Delete entry with *key* from *map*.
749 * 0 on success, or a negative error in case of failure.
751 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
753 * For tracing programs, safely attempt to read *size* bytes from
754 * kernel space address *unsafe_ptr* and store the data in *dst*.
756 * Generally, use **bpf_probe_read_user**\ () or
757 * **bpf_probe_read_kernel**\ () instead.
759 * 0 on success, or a negative error in case of failure.
761 * u64 bpf_ktime_get_ns(void)
763 * Return the time elapsed since system boot, in nanoseconds.
764 * Does not include time the system was suspended.
765 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
769 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
771 * This helper is a "printk()-like" facility for debugging. It
772 * prints a message defined by format *fmt* (of size *fmt_size*)
773 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
774 * available. It can take up to three additional **u64**
775 * arguments (as an eBPF helpers, the total number of arguments is
778 * Each time the helper is called, it appends a line to the trace.
779 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
780 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
781 * The format of the trace is customizable, and the exact output
782 * one will get depends on the options set in
783 * *\/sys/kernel/debug/tracing/trace_options* (see also the
784 * *README* file under the same directory). However, it usually
785 * defaults to something like:
789 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
793 * * ``telnet`` is the name of the current task.
794 * * ``470`` is the PID of the current task.
795 * * ``001`` is the CPU number on which the task is
797 * * In ``.N..``, each character refers to a set of
798 * options (whether irqs are enabled, scheduling
799 * options, whether hard/softirqs are running, level of
800 * preempt_disabled respectively). **N** means that
801 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
803 * * ``419421.045894`` is a timestamp.
804 * * ``0x00000001`` is a fake value used by BPF for the
805 * instruction pointer register.
806 * * ``<formatted msg>`` is the message formatted with
809 * The conversion specifiers supported by *fmt* are similar, but
810 * more limited than for printk(). They are **%d**, **%i**,
811 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
812 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
813 * of field, padding with zeroes, etc.) is available, and the
814 * helper will return **-EINVAL** (but print nothing) if it
815 * encounters an unknown specifier.
817 * Also, note that **bpf_trace_printk**\ () is slow, and should
818 * only be used for debugging purposes. For this reason, a notice
819 * block (spanning several lines) is printed to kernel logs and
820 * states that the helper should not be used "for production use"
821 * the first time this helper is used (or more precisely, when
822 * **trace_printk**\ () buffers are allocated). For passing values
823 * to user space, perf events should be preferred.
825 * The number of bytes written to the buffer, or a negative error
826 * in case of failure.
828 * u32 bpf_get_prandom_u32(void)
830 * Get a pseudo-random number.
832 * From a security point of view, this helper uses its own
833 * pseudo-random internal state, and cannot be used to infer the
834 * seed of other random functions in the kernel. However, it is
835 * essential to note that the generator used by the helper is not
836 * cryptographically secure.
838 * A random 32-bit unsigned value.
840 * u32 bpf_get_smp_processor_id(void)
842 * Get the SMP (symmetric multiprocessing) processor id. Note that
843 * all programs run with preemption disabled, which means that the
844 * SMP processor id is stable during all the execution of the
847 * The SMP id of the processor running the program.
849 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
851 * Store *len* bytes from address *from* into the packet
852 * associated to *skb*, at *offset*. *flags* are a combination of
853 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
854 * checksum for the packet after storing the bytes) and
855 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
856 * **->swhash** and *skb*\ **->l4hash** to 0).
858 * A call to this helper is susceptible to change the underlying
859 * packet buffer. Therefore, at load time, all checks on pointers
860 * previously done by the verifier are invalidated and must be
861 * performed again, if the helper is used in combination with
862 * direct packet access.
864 * 0 on success, or a negative error in case of failure.
866 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
868 * Recompute the layer 3 (e.g. IP) checksum for the packet
869 * associated to *skb*. Computation is incremental, so the helper
870 * must know the former value of the header field that was
871 * modified (*from*), the new value of this field (*to*), and the
872 * number of bytes (2 or 4) for this field, stored in *size*.
873 * Alternatively, it is possible to store the difference between
874 * the previous and the new values of the header field in *to*, by
875 * setting *from* and *size* to 0. For both methods, *offset*
876 * indicates the location of the IP checksum within the packet.
878 * This helper works in combination with **bpf_csum_diff**\ (),
879 * which does not update the checksum in-place, but offers more
880 * flexibility and can handle sizes larger than 2 or 4 for the
881 * checksum to update.
883 * A call to this helper is susceptible to change the underlying
884 * packet buffer. Therefore, at load time, all checks on pointers
885 * previously done by the verifier are invalidated and must be
886 * performed again, if the helper is used in combination with
887 * direct packet access.
889 * 0 on success, or a negative error in case of failure.
891 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
893 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
894 * packet associated to *skb*. Computation is incremental, so the
895 * helper must know the former value of the header field that was
896 * modified (*from*), the new value of this field (*to*), and the
897 * number of bytes (2 or 4) for this field, stored on the lowest
898 * four bits of *flags*. Alternatively, it is possible to store
899 * the difference between the previous and the new values of the
900 * header field in *to*, by setting *from* and the four lowest
901 * bits of *flags* to 0. For both methods, *offset* indicates the
902 * location of the IP checksum within the packet. In addition to
903 * the size of the field, *flags* can be added (bitwise OR) actual
904 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
905 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
906 * for updates resulting in a null checksum the value is set to
907 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
908 * the checksum is to be computed against a pseudo-header.
910 * This helper works in combination with **bpf_csum_diff**\ (),
911 * which does not update the checksum in-place, but offers more
912 * flexibility and can handle sizes larger than 2 or 4 for the
913 * checksum to update.
915 * A call to this helper is susceptible to change the underlying
916 * packet buffer. Therefore, at load time, all checks on pointers
917 * previously done by the verifier are invalidated and must be
918 * performed again, if the helper is used in combination with
919 * direct packet access.
921 * 0 on success, or a negative error in case of failure.
923 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
925 * This special helper is used to trigger a "tail call", or in
926 * other words, to jump into another eBPF program. The same stack
927 * frame is used (but values on stack and in registers for the
928 * caller are not accessible to the callee). This mechanism allows
929 * for program chaining, either for raising the maximum number of
930 * available eBPF instructions, or to execute given programs in
931 * conditional blocks. For security reasons, there is an upper
932 * limit to the number of successive tail calls that can be
935 * Upon call of this helper, the program attempts to jump into a
936 * program referenced at index *index* in *prog_array_map*, a
937 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
938 * *ctx*, a pointer to the context.
940 * If the call succeeds, the kernel immediately runs the first
941 * instruction of the new program. This is not a function call,
942 * and it never returns to the previous program. If the call
943 * fails, then the helper has no effect, and the caller continues
944 * to run its subsequent instructions. A call can fail if the
945 * destination program for the jump does not exist (i.e. *index*
946 * is superior to the number of entries in *prog_array_map*), or
947 * if the maximum number of tail calls has been reached for this
948 * chain of programs. This limit is defined in the kernel by the
949 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
950 * which is currently set to 32.
952 * 0 on success, or a negative error in case of failure.
954 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
956 * Clone and redirect the packet associated to *skb* to another
957 * net device of index *ifindex*. Both ingress and egress
958 * interfaces can be used for redirection. The **BPF_F_INGRESS**
959 * value in *flags* is used to make the distinction (ingress path
960 * is selected if the flag is present, egress path otherwise).
961 * This is the only flag supported for now.
963 * In comparison with **bpf_redirect**\ () helper,
964 * **bpf_clone_redirect**\ () has the associated cost of
965 * duplicating the packet buffer, but this can be executed out of
966 * the eBPF program. Conversely, **bpf_redirect**\ () is more
967 * efficient, but it is handled through an action code where the
968 * redirection happens only after the eBPF program has returned.
970 * A call to this helper is susceptible to change the underlying
971 * packet buffer. Therefore, at load time, all checks on pointers
972 * previously done by the verifier are invalidated and must be
973 * performed again, if the helper is used in combination with
974 * direct packet access.
976 * 0 on success, or a negative error in case of failure.
978 * u64 bpf_get_current_pid_tgid(void)
980 * A 64-bit integer containing the current tgid and pid, and
982 * *current_task*\ **->tgid << 32 \|**
983 * *current_task*\ **->pid**.
985 * u64 bpf_get_current_uid_gid(void)
987 * A 64-bit integer containing the current GID and UID, and
988 * created as such: *current_gid* **<< 32 \|** *current_uid*.
990 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
992 * Copy the **comm** attribute of the current task into *buf* of
993 * *size_of_buf*. The **comm** attribute contains the name of
994 * the executable (excluding the path) for the current task. The
995 * *size_of_buf* must be strictly positive. On success, the
996 * helper makes sure that the *buf* is NUL-terminated. On failure,
997 * it is filled with zeroes.
999 * 0 on success, or a negative error in case of failure.
1001 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1003 * Retrieve the classid for the current task, i.e. for the net_cls
1004 * cgroup to which *skb* belongs.
1006 * This helper can be used on TC egress path, but not on ingress.
1008 * The net_cls cgroup provides an interface to tag network packets
1009 * based on a user-provided identifier for all traffic coming from
1010 * the tasks belonging to the related cgroup. See also the related
1011 * kernel documentation, available from the Linux sources in file
1012 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1014 * The Linux kernel has two versions for cgroups: there are
1015 * cgroups v1 and cgroups v2. Both are available to users, who can
1016 * use a mixture of them, but note that the net_cls cgroup is for
1017 * cgroup v1 only. This makes it incompatible with BPF programs
1018 * run on cgroups, which is a cgroup-v2-only feature (a socket can
1019 * only hold data for one version of cgroups at a time).
1021 * This helper is only available is the kernel was compiled with
1022 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1023 * "**y**" or to "**m**".
1025 * The classid, or 0 for the default unconfigured classid.
1027 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1029 * Push a *vlan_tci* (VLAN tag control information) of protocol
1030 * *vlan_proto* to the packet associated to *skb*, then update
1031 * the checksum. Note that if *vlan_proto* is different from
1032 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1033 * be **ETH_P_8021Q**.
1035 * A call to this helper is susceptible to change the underlying
1036 * packet buffer. Therefore, at load time, all checks on pointers
1037 * previously done by the verifier are invalidated and must be
1038 * performed again, if the helper is used in combination with
1039 * direct packet access.
1041 * 0 on success, or a negative error in case of failure.
1043 * long bpf_skb_vlan_pop(struct sk_buff *skb)
1045 * Pop a VLAN header from the packet associated to *skb*.
1047 * A call to this helper is susceptible to change the underlying
1048 * packet buffer. Therefore, at load time, all checks on pointers
1049 * previously done by the verifier are invalidated and must be
1050 * performed again, if the helper is used in combination with
1051 * direct packet access.
1053 * 0 on success, or a negative error in case of failure.
1055 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1057 * Get tunnel metadata. This helper takes a pointer *key* to an
1058 * empty **struct bpf_tunnel_key** of **size**, that will be
1059 * filled with tunnel metadata for the packet associated to *skb*.
1060 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1061 * indicates that the tunnel is based on IPv6 protocol instead of
1064 * The **struct bpf_tunnel_key** is an object that generalizes the
1065 * principal parameters used by various tunneling protocols into a
1066 * single struct. This way, it can be used to easily make a
1067 * decision based on the contents of the encapsulation header,
1068 * "summarized" in this struct. In particular, it holds the IP
1069 * address of the remote end (IPv4 or IPv6, depending on the case)
1070 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1071 * this struct exposes the *key*\ **->tunnel_id**, which is
1072 * generally mapped to a VNI (Virtual Network Identifier), making
1073 * it programmable together with the **bpf_skb_set_tunnel_key**\
1076 * Let's imagine that the following code is part of a program
1077 * attached to the TC ingress interface, on one end of a GRE
1078 * tunnel, and is supposed to filter out all messages coming from
1079 * remote ends with IPv4 address other than 10.0.0.1:
1084 * struct bpf_tunnel_key key = {};
1086 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1088 * return TC_ACT_SHOT; // drop packet
1090 * if (key.remote_ipv4 != 0x0a000001)
1091 * return TC_ACT_SHOT; // drop packet
1093 * return TC_ACT_OK; // accept packet
1095 * This interface can also be used with all encapsulation devices
1096 * that can operate in "collect metadata" mode: instead of having
1097 * one network device per specific configuration, the "collect
1098 * metadata" mode only requires a single device where the
1099 * configuration can be extracted from this helper.
1101 * This can be used together with various tunnels such as VXLan,
1102 * Geneve, GRE or IP in IP (IPIP).
1104 * 0 on success, or a negative error in case of failure.
1106 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1108 * Populate tunnel metadata for packet associated to *skb.* The
1109 * tunnel metadata is set to the contents of *key*, of *size*. The
1110 * *flags* can be set to a combination of the following values:
1112 * **BPF_F_TUNINFO_IPV6**
1113 * Indicate that the tunnel is based on IPv6 protocol
1115 * **BPF_F_ZERO_CSUM_TX**
1116 * For IPv4 packets, add a flag to tunnel metadata
1117 * indicating that checksum computation should be skipped
1118 * and checksum set to zeroes.
1119 * **BPF_F_DONT_FRAGMENT**
1120 * Add a flag to tunnel metadata indicating that the
1121 * packet should not be fragmented.
1122 * **BPF_F_SEQ_NUMBER**
1123 * Add a flag to tunnel metadata indicating that a
1124 * sequence number should be added to tunnel header before
1125 * sending the packet. This flag was added for GRE
1126 * encapsulation, but might be used with other protocols
1127 * as well in the future.
1129 * Here is a typical usage on the transmit path:
1133 * struct bpf_tunnel_key key;
1135 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1136 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1138 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1139 * helper for additional information.
1141 * 0 on success, or a negative error in case of failure.
1143 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1145 * Read the value of a perf event counter. This helper relies on a
1146 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1147 * the perf event counter is selected when *map* is updated with
1148 * perf event file descriptors. The *map* is an array whose size
1149 * is the number of available CPUs, and each cell contains a value
1150 * relative to one CPU. The value to retrieve is indicated by
1151 * *flags*, that contains the index of the CPU to look up, masked
1152 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1153 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1154 * current CPU should be retrieved.
1156 * Note that before Linux 4.13, only hardware perf event can be
1159 * Also, be aware that the newer helper
1160 * **bpf_perf_event_read_value**\ () is recommended over
1161 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1162 * quirks where error and counter value are used as a return code
1163 * (which is wrong to do since ranges may overlap). This issue is
1164 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1165 * time provides more features over the **bpf_perf_event_read**\
1166 * () interface. Please refer to the description of
1167 * **bpf_perf_event_read_value**\ () for details.
1169 * The value of the perf event counter read from the map, or a
1170 * negative error code in case of failure.
1172 * long bpf_redirect(u32 ifindex, u64 flags)
1174 * Redirect the packet to another net device of index *ifindex*.
1175 * This helper is somewhat similar to **bpf_clone_redirect**\
1176 * (), except that the packet is not cloned, which provides
1177 * increased performance.
1179 * Except for XDP, both ingress and egress interfaces can be used
1180 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1181 * to make the distinction (ingress path is selected if the flag
1182 * is present, egress path otherwise). Currently, XDP only
1183 * supports redirection to the egress interface, and accepts no
1186 * The same effect can also be attained with the more generic
1187 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1188 * redirect target instead of providing it directly to the helper.
1190 * For XDP, the helper returns **XDP_REDIRECT** on success or
1191 * **XDP_ABORTED** on error. For other program types, the values
1192 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1195 * u32 bpf_get_route_realm(struct sk_buff *skb)
1197 * Retrieve the realm or the route, that is to say the
1198 * **tclassid** field of the destination for the *skb*. The
1199 * identifier retrieved is a user-provided tag, similar to the
1200 * one used with the net_cls cgroup (see description for
1201 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1202 * held by a route (a destination entry), not by a task.
1204 * Retrieving this identifier works with the clsact TC egress hook
1205 * (see also **tc-bpf(8)**), or alternatively on conventional
1206 * classful egress qdiscs, but not on TC ingress path. In case of
1207 * clsact TC egress hook, this has the advantage that, internally,
1208 * the destination entry has not been dropped yet in the transmit
1209 * path. Therefore, the destination entry does not need to be
1210 * artificially held via **netif_keep_dst**\ () for a classful
1211 * qdisc until the *skb* is freed.
1213 * This helper is available only if the kernel was compiled with
1214 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1216 * The realm of the route for the packet associated to *skb*, or 0
1217 * if none was found.
1219 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1221 * Write raw *data* blob into a special BPF perf event held by
1222 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1223 * event must have the following attributes: **PERF_SAMPLE_RAW**
1224 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1225 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1227 * The *flags* are used to indicate the index in *map* for which
1228 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1229 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1230 * to indicate that the index of the current CPU core should be
1233 * The value to write, of *size*, is passed through eBPF stack and
1234 * pointed by *data*.
1236 * The context of the program *ctx* needs also be passed to the
1239 * On user space, a program willing to read the values needs to
1240 * call **perf_event_open**\ () on the perf event (either for
1241 * one or for all CPUs) and to store the file descriptor into the
1242 * *map*. This must be done before the eBPF program can send data
1243 * into it. An example is available in file
1244 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1245 * tree (the eBPF program counterpart is in
1246 * *samples/bpf/trace_output_kern.c*).
1248 * **bpf_perf_event_output**\ () achieves better performance
1249 * than **bpf_trace_printk**\ () for sharing data with user
1250 * space, and is much better suitable for streaming data from eBPF
1253 * Note that this helper is not restricted to tracing use cases
1254 * and can be used with programs attached to TC or XDP as well,
1255 * where it allows for passing data to user space listeners. Data
1258 * * Only custom structs,
1259 * * Only the packet payload, or
1260 * * A combination of both.
1262 * 0 on success, or a negative error in case of failure.
1264 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1266 * This helper was provided as an easy way to load data from a
1267 * packet. It can be used to load *len* bytes from *offset* from
1268 * the packet associated to *skb*, into the buffer pointed by
1271 * Since Linux 4.7, usage of this helper has mostly been replaced
1272 * by "direct packet access", enabling packet data to be
1273 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1274 * pointing respectively to the first byte of packet data and to
1275 * the byte after the last byte of packet data. However, it
1276 * remains useful if one wishes to read large quantities of data
1277 * at once from a packet into the eBPF stack.
1279 * 0 on success, or a negative error in case of failure.
1281 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1283 * Walk a user or a kernel stack and return its id. To achieve
1284 * this, the helper needs *ctx*, which is a pointer to the context
1285 * on which the tracing program is executed, and a pointer to a
1286 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1288 * The last argument, *flags*, holds the number of stack frames to
1289 * skip (from 0 to 255), masked with
1290 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1291 * a combination of the following flags:
1293 * **BPF_F_USER_STACK**
1294 * Collect a user space stack instead of a kernel stack.
1295 * **BPF_F_FAST_STACK_CMP**
1296 * Compare stacks by hash only.
1297 * **BPF_F_REUSE_STACKID**
1298 * If two different stacks hash into the same *stackid*,
1299 * discard the old one.
1301 * The stack id retrieved is a 32 bit long integer handle which
1302 * can be further combined with other data (including other stack
1303 * ids) and used as a key into maps. This can be useful for
1304 * generating a variety of graphs (such as flame graphs or off-cpu
1307 * For walking a stack, this helper is an improvement over
1308 * **bpf_probe_read**\ (), which can be used with unrolled loops
1309 * but is not efficient and consumes a lot of eBPF instructions.
1310 * Instead, **bpf_get_stackid**\ () can collect up to
1311 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1312 * this limit can be controlled with the **sysctl** program, and
1313 * that it should be manually increased in order to profile long
1314 * user stacks (such as stacks for Java programs). To do so, use:
1318 * # sysctl kernel.perf_event_max_stack=<new value>
1320 * The positive or null stack id on success, or a negative error
1321 * in case of failure.
1323 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1325 * Compute a checksum difference, from the raw buffer pointed by
1326 * *from*, of length *from_size* (that must be a multiple of 4),
1327 * towards the raw buffer pointed by *to*, of size *to_size*
1328 * (same remark). An optional *seed* can be added to the value
1329 * (this can be cascaded, the seed may come from a previous call
1332 * This is flexible enough to be used in several ways:
1334 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1335 * checksum, it can be used when pushing new data.
1336 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1337 * checksum, it can be used when removing data from a packet.
1338 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1339 * can be used to compute a diff. Note that *from_size* and
1340 * *to_size* do not need to be equal.
1342 * This helper can be used in combination with
1343 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1344 * which one can feed in the difference computed with
1345 * **bpf_csum_diff**\ ().
1347 * The checksum result, or a negative error code in case of
1350 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1352 * Retrieve tunnel options metadata for the packet associated to
1353 * *skb*, and store the raw tunnel option data to the buffer *opt*
1356 * This helper can be used with encapsulation devices that can
1357 * operate in "collect metadata" mode (please refer to the related
1358 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1359 * more details). A particular example where this can be used is
1360 * in combination with the Geneve encapsulation protocol, where it
1361 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1362 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1363 * the eBPF program. This allows for full customization of these
1366 * The size of the option data retrieved.
1368 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1370 * Set tunnel options metadata for the packet associated to *skb*
1371 * to the option data contained in the raw buffer *opt* of *size*.
1373 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1374 * helper for additional information.
1376 * 0 on success, or a negative error in case of failure.
1378 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1380 * Change the protocol of the *skb* to *proto*. Currently
1381 * supported are transition from IPv4 to IPv6, and from IPv6 to
1382 * IPv4. The helper takes care of the groundwork for the
1383 * transition, including resizing the socket buffer. The eBPF
1384 * program is expected to fill the new headers, if any, via
1385 * **skb_store_bytes**\ () and to recompute the checksums with
1386 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1387 * (). The main case for this helper is to perform NAT64
1388 * operations out of an eBPF program.
1390 * Internally, the GSO type is marked as dodgy so that headers are
1391 * checked and segments are recalculated by the GSO/GRO engine.
1392 * The size for GSO target is adapted as well.
1394 * All values for *flags* are reserved for future usage, and must
1397 * A call to this helper is susceptible to change the underlying
1398 * packet buffer. Therefore, at load time, all checks on pointers
1399 * previously done by the verifier are invalidated and must be
1400 * performed again, if the helper is used in combination with
1401 * direct packet access.
1403 * 0 on success, or a negative error in case of failure.
1405 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
1407 * Change the packet type for the packet associated to *skb*. This
1408 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1409 * the eBPF program does not have a write access to *skb*\
1410 * **->pkt_type** beside this helper. Using a helper here allows
1411 * for graceful handling of errors.
1413 * The major use case is to change incoming *skb*s to
1414 * **PACKET_HOST** in a programmatic way instead of having to
1415 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1418 * Note that *type* only allows certain values. At this time, they
1423 * **PACKET_BROADCAST**
1424 * Send packet to all.
1425 * **PACKET_MULTICAST**
1426 * Send packet to group.
1427 * **PACKET_OTHERHOST**
1428 * Send packet to someone else.
1430 * 0 on success, or a negative error in case of failure.
1432 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1434 * Check whether *skb* is a descendant of the cgroup2 held by
1435 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1437 * The return value depends on the result of the test, and can be:
1439 * * 0, if the *skb* failed the cgroup2 descendant test.
1440 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1441 * * A negative error code, if an error occurred.
1443 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1445 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1446 * not set, in particular if the hash was cleared due to mangling,
1447 * recompute this hash. Later accesses to the hash can be done
1448 * directly with *skb*\ **->hash**.
1450 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1451 * prototype with **bpf_skb_change_proto**\ (), or calling
1452 * **bpf_skb_store_bytes**\ () with the
1453 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1454 * the hash and to trigger a new computation for the next call to
1455 * **bpf_get_hash_recalc**\ ().
1459 * u64 bpf_get_current_task(void)
1461 * A pointer to the current task struct.
1463 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
1465 * Attempt in a safe way to write *len* bytes from the buffer
1466 * *src* to *dst* in memory. It only works for threads that are in
1467 * user context, and *dst* must be a valid user space address.
1469 * This helper should not be used to implement any kind of
1470 * security mechanism because of TOC-TOU attacks, but rather to
1471 * debug, divert, and manipulate execution of semi-cooperative
1474 * Keep in mind that this feature is meant for experiments, and it
1475 * has a risk of crashing the system and running programs.
1476 * Therefore, when an eBPF program using this helper is attached,
1477 * a warning including PID and process name is printed to kernel
1480 * 0 on success, or a negative error in case of failure.
1482 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1484 * Check whether the probe is being run is the context of a given
1485 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1486 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1488 * The return value depends on the result of the test, and can be:
1490 * * 0, if current task belongs to the cgroup2.
1491 * * 1, if current task does not belong to the cgroup2.
1492 * * A negative error code, if an error occurred.
1494 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1496 * Resize (trim or grow) the packet associated to *skb* to the
1497 * new *len*. The *flags* are reserved for future usage, and must
1500 * The basic idea is that the helper performs the needed work to
1501 * change the size of the packet, then the eBPF program rewrites
1502 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1503 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1504 * and others. This helper is a slow path utility intended for
1505 * replies with control messages. And because it is targeted for
1506 * slow path, the helper itself can afford to be slow: it
1507 * implicitly linearizes, unclones and drops offloads from the
1510 * A call to this helper is susceptible to change the underlying
1511 * packet buffer. Therefore, at load time, all checks on pointers
1512 * previously done by the verifier are invalidated and must be
1513 * performed again, if the helper is used in combination with
1514 * direct packet access.
1516 * 0 on success, or a negative error in case of failure.
1518 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1520 * Pull in non-linear data in case the *skb* is non-linear and not
1521 * all of *len* are part of the linear section. Make *len* bytes
1522 * from *skb* readable and writable. If a zero value is passed for
1523 * *len*, then the whole length of the *skb* is pulled.
1525 * This helper is only needed for reading and writing with direct
1528 * For direct packet access, testing that offsets to access
1529 * are within packet boundaries (test on *skb*\ **->data_end**) is
1530 * susceptible to fail if offsets are invalid, or if the requested
1531 * data is in non-linear parts of the *skb*. On failure the
1532 * program can just bail out, or in the case of a non-linear
1533 * buffer, use a helper to make the data available. The
1534 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1535 * the data. Another one consists in using **bpf_skb_pull_data**
1536 * to pull in once the non-linear parts, then retesting and
1537 * eventually access the data.
1539 * At the same time, this also makes sure the *skb* is uncloned,
1540 * which is a necessary condition for direct write. As this needs
1541 * to be an invariant for the write part only, the verifier
1542 * detects writes and adds a prologue that is calling
1543 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1544 * the very beginning in case it is indeed cloned.
1546 * A call to this helper is susceptible to change the underlying
1547 * packet buffer. Therefore, at load time, all checks on pointers
1548 * previously done by the verifier are invalidated and must be
1549 * performed again, if the helper is used in combination with
1550 * direct packet access.
1552 * 0 on success, or a negative error in case of failure.
1554 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1556 * Add the checksum *csum* into *skb*\ **->csum** in case the
1557 * driver has supplied a checksum for the entire packet into that
1558 * field. Return an error otherwise. This helper is intended to be
1559 * used in combination with **bpf_csum_diff**\ (), in particular
1560 * when the checksum needs to be updated after data has been
1561 * written into the packet through direct packet access.
1563 * The checksum on success, or a negative error code in case of
1566 * void bpf_set_hash_invalid(struct sk_buff *skb)
1568 * Invalidate the current *skb*\ **->hash**. It can be used after
1569 * mangling on headers through direct packet access, in order to
1570 * indicate that the hash is outdated and to trigger a
1571 * recalculation the next time the kernel tries to access this
1572 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1574 * long bpf_get_numa_node_id(void)
1576 * Return the id of the current NUMA node. The primary use case
1577 * for this helper is the selection of sockets for the local NUMA
1578 * node, when the program is attached to sockets using the
1579 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1580 * but the helper is also available to other eBPF program types,
1581 * similarly to **bpf_get_smp_processor_id**\ ().
1583 * The id of current NUMA node.
1585 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1587 * Grows headroom of packet associated to *skb* and adjusts the
1588 * offset of the MAC header accordingly, adding *len* bytes of
1589 * space. It automatically extends and reallocates memory as
1592 * This helper can be used on a layer 3 *skb* to push a MAC header
1593 * for redirection into a layer 2 device.
1595 * All values for *flags* are reserved for future usage, and must
1598 * A call to this helper is susceptible to change the underlying
1599 * packet buffer. Therefore, at load time, all checks on pointers
1600 * previously done by the verifier are invalidated and must be
1601 * performed again, if the helper is used in combination with
1602 * direct packet access.
1604 * 0 on success, or a negative error in case of failure.
1606 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1608 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1609 * it is possible to use a negative value for *delta*. This helper
1610 * can be used to prepare the packet for pushing or popping
1613 * A call to this helper is susceptible to change the underlying
1614 * packet buffer. Therefore, at load time, all checks on pointers
1615 * previously done by the verifier are invalidated and must be
1616 * performed again, if the helper is used in combination with
1617 * direct packet access.
1619 * 0 on success, or a negative error in case of failure.
1621 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1623 * Copy a NUL terminated string from an unsafe kernel address
1624 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
1627 * Generally, use **bpf_probe_read_user_str**\ () or
1628 * **bpf_probe_read_kernel_str**\ () instead.
1630 * On success, the strictly positive length of the string,
1631 * including the trailing NUL character. On error, a negative
1634 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1636 * If the **struct sk_buff** pointed by *skb* has a known socket,
1637 * retrieve the cookie (generated by the kernel) of this socket.
1638 * If no cookie has been set yet, generate a new cookie. Once
1639 * generated, the socket cookie remains stable for the life of the
1640 * socket. This helper can be useful for monitoring per socket
1641 * networking traffic statistics as it provides a global socket
1642 * identifier that can be assumed unique.
1644 * A 8-byte long non-decreasing number on success, or 0 if the
1645 * socket field is missing inside *skb*.
1647 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1649 * Equivalent to bpf_get_socket_cookie() helper that accepts
1650 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1652 * A 8-byte long non-decreasing number.
1654 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1656 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
1657 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1659 * A 8-byte long non-decreasing number.
1661 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1663 * The owner UID of the socket associated to *skb*. If the socket
1664 * is **NULL**, or if it is not a full socket (i.e. if it is a
1665 * time-wait or a request socket instead), **overflowuid** value
1666 * is returned (note that **overflowuid** might also be the actual
1667 * UID value for the socket).
1669 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
1671 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1676 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1678 * Emulate a call to **setsockopt()** on the socket associated to
1679 * *bpf_socket*, which must be a full socket. The *level* at
1680 * which the option resides and the name *optname* of the option
1681 * must be specified, see **setsockopt(2)** for more information.
1682 * The option value of length *optlen* is pointed by *optval*.
1684 * *bpf_socket* should be one of the following:
1686 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1687 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1688 * and **BPF_CGROUP_INET6_CONNECT**.
1690 * This helper actually implements a subset of **setsockopt()**.
1691 * It supports the following *level*\ s:
1693 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1694 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1695 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
1696 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
1697 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1698 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1699 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
1700 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
1701 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**.
1702 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1703 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1705 * 0 on success, or a negative error in case of failure.
1707 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1709 * Grow or shrink the room for data in the packet associated to
1710 * *skb* by *len_diff*, and according to the selected *mode*.
1712 * By default, the helper will reset any offloaded checksum
1713 * indicator of the skb to CHECKSUM_NONE. This can be avoided
1714 * by the following flag:
1716 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
1717 * checksum data of the skb to CHECKSUM_NONE.
1719 * There are two supported modes at this time:
1721 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1722 * (room space is added or removed below the layer 2 header).
1724 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1725 * (room space is added or removed below the layer 3 header).
1727 * The following flags are supported at this time:
1729 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1730 * Adjusting mss in this way is not allowed for datagrams.
1732 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1733 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1734 * Any new space is reserved to hold a tunnel header.
1735 * Configure skb offsets and other fields accordingly.
1737 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1738 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1739 * Use with ENCAP_L3 flags to further specify the tunnel type.
1741 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1742 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1743 * type; *len* is the length of the inner MAC header.
1745 * A call to this helper is susceptible to change the underlying
1746 * packet buffer. Therefore, at load time, all checks on pointers
1747 * previously done by the verifier are invalidated and must be
1748 * performed again, if the helper is used in combination with
1749 * direct packet access.
1751 * 0 on success, or a negative error in case of failure.
1753 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1755 * Redirect the packet to the endpoint referenced by *map* at
1756 * index *key*. Depending on its type, this *map* can contain
1757 * references to net devices (for forwarding packets through other
1758 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1759 * but this is only implemented for native XDP (with driver
1760 * support) as of this writing).
1762 * The lower two bits of *flags* are used as the return code if
1763 * the map lookup fails. This is so that the return value can be
1764 * one of the XDP program return codes up to **XDP_TX**, as chosen
1765 * by the caller. Any higher bits in the *flags* argument must be
1768 * See also **bpf_redirect**\ (), which only supports redirecting
1769 * to an ifindex, but doesn't require a map to do so.
1771 * **XDP_REDIRECT** on success, or the value of the two lower bits
1772 * of the *flags* argument on error.
1774 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1776 * Redirect the packet to the socket referenced by *map* (of type
1777 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1778 * egress interfaces can be used for redirection. The
1779 * **BPF_F_INGRESS** value in *flags* is used to make the
1780 * distinction (ingress path is selected if the flag is present,
1781 * egress path otherwise). This is the only flag supported for now.
1783 * **SK_PASS** on success, or **SK_DROP** on error.
1785 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1787 * Add an entry to, or update a *map* referencing sockets. The
1788 * *skops* is used as a new value for the entry associated to
1789 * *key*. *flags* is one of:
1792 * The entry for *key* must not exist in the map.
1794 * The entry for *key* must already exist in the map.
1796 * No condition on the existence of the entry for *key*.
1798 * If the *map* has eBPF programs (parser and verdict), those will
1799 * be inherited by the socket being added. If the socket is
1800 * already attached to eBPF programs, this results in an error.
1802 * 0 on success, or a negative error in case of failure.
1804 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1806 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1807 * *delta* (which can be positive or negative). Note that this
1808 * operation modifies the address stored in *xdp_md*\ **->data**,
1809 * so the latter must be loaded only after the helper has been
1812 * The use of *xdp_md*\ **->data_meta** is optional and programs
1813 * are not required to use it. The rationale is that when the
1814 * packet is processed with XDP (e.g. as DoS filter), it is
1815 * possible to push further meta data along with it before passing
1816 * to the stack, and to give the guarantee that an ingress eBPF
1817 * program attached as a TC classifier on the same device can pick
1818 * this up for further post-processing. Since TC works with socket
1819 * buffers, it remains possible to set from XDP the **mark** or
1820 * **priority** pointers, or other pointers for the socket buffer.
1821 * Having this scratch space generic and programmable allows for
1822 * more flexibility as the user is free to store whatever meta
1825 * A call to this helper is susceptible to change the underlying
1826 * packet buffer. Therefore, at load time, all checks on pointers
1827 * previously done by the verifier are invalidated and must be
1828 * performed again, if the helper is used in combination with
1829 * direct packet access.
1831 * 0 on success, or a negative error in case of failure.
1833 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1835 * Read the value of a perf event counter, and store it into *buf*
1836 * of size *buf_size*. This helper relies on a *map* of type
1837 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1838 * counter is selected when *map* is updated with perf event file
1839 * descriptors. The *map* is an array whose size is the number of
1840 * available CPUs, and each cell contains a value relative to one
1841 * CPU. The value to retrieve is indicated by *flags*, that
1842 * contains the index of the CPU to look up, masked with
1843 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1844 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1845 * current CPU should be retrieved.
1847 * This helper behaves in a way close to
1848 * **bpf_perf_event_read**\ () helper, save that instead of
1849 * just returning the value observed, it fills the *buf*
1850 * structure. This allows for additional data to be retrieved: in
1851 * particular, the enabled and running times (in *buf*\
1852 * **->enabled** and *buf*\ **->running**, respectively) are
1853 * copied. In general, **bpf_perf_event_read_value**\ () is
1854 * recommended over **bpf_perf_event_read**\ (), which has some
1855 * ABI issues and provides fewer functionalities.
1857 * These values are interesting, because hardware PMU (Performance
1858 * Monitoring Unit) counters are limited resources. When there are
1859 * more PMU based perf events opened than available counters,
1860 * kernel will multiplex these events so each event gets certain
1861 * percentage (but not all) of the PMU time. In case that
1862 * multiplexing happens, the number of samples or counter value
1863 * will not reflect the case compared to when no multiplexing
1864 * occurs. This makes comparison between different runs difficult.
1865 * Typically, the counter value should be normalized before
1866 * comparing to other experiments. The usual normalization is done
1871 * normalized_counter = counter * t_enabled / t_running
1873 * Where t_enabled is the time enabled for event and t_running is
1874 * the time running for event since last normalization. The
1875 * enabled and running times are accumulated since the perf event
1876 * open. To achieve scaling factor between two invocations of an
1877 * eBPF program, users can use CPU id as the key (which is
1878 * typical for perf array usage model) to remember the previous
1879 * value and do the calculation inside the eBPF program.
1881 * 0 on success, or a negative error in case of failure.
1883 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1885 * For en eBPF program attached to a perf event, retrieve the
1886 * value of the event counter associated to *ctx* and store it in
1887 * the structure pointed by *buf* and of size *buf_size*. Enabled
1888 * and running times are also stored in the structure (see
1889 * description of helper **bpf_perf_event_read_value**\ () for
1892 * 0 on success, or a negative error in case of failure.
1894 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1896 * Emulate a call to **getsockopt()** on the socket associated to
1897 * *bpf_socket*, which must be a full socket. The *level* at
1898 * which the option resides and the name *optname* of the option
1899 * must be specified, see **getsockopt(2)** for more information.
1900 * The retrieved value is stored in the structure pointed by
1901 * *opval* and of length *optlen*.
1903 * *bpf_socket* should be one of the following:
1905 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1906 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1907 * and **BPF_CGROUP_INET6_CONNECT**.
1909 * This helper actually implements a subset of **getsockopt()**.
1910 * It supports the following *level*\ s:
1912 * * **IPPROTO_TCP**, which supports *optname*
1913 * **TCP_CONGESTION**.
1914 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1915 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1917 * 0 on success, or a negative error in case of failure.
1919 * long bpf_override_return(struct pt_regs *regs, u64 rc)
1921 * Used for error injection, this helper uses kprobes to override
1922 * the return value of the probed function, and to set it to *rc*.
1923 * The first argument is the context *regs* on which the kprobe
1926 * This helper works by setting the PC (program counter)
1927 * to an override function which is run in place of the original
1928 * probed function. This means the probed function is not run at
1929 * all. The replacement function just returns with the required
1932 * This helper has security implications, and thus is subject to
1933 * restrictions. It is only available if the kernel was compiled
1934 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1935 * option, and in this case it only works on functions tagged with
1936 * **ALLOW_ERROR_INJECTION** in the kernel code.
1938 * Also, the helper is only available for the architectures having
1939 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1940 * x86 architecture is the only one to support this feature.
1944 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1946 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1947 * for the full TCP socket associated to *bpf_sock_ops* to
1950 * The primary use of this field is to determine if there should
1951 * be calls to eBPF programs of type
1952 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1953 * code. A program of the same type can change its value, per
1954 * connection and as necessary, when the connection is
1955 * established. This field is directly accessible for reading, but
1956 * this helper must be used for updates in order to return an
1957 * error if an eBPF program tries to set a callback that is not
1958 * supported in the current kernel.
1960 * *argval* is a flag array which can combine these flags:
1962 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1963 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1964 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1965 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1967 * Therefore, this function can be used to clear a callback flag by
1968 * setting the appropriate bit to zero. e.g. to disable the RTO
1971 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1972 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1974 * Here are some examples of where one could call such eBPF
1978 * * When a packet is retransmitted.
1979 * * When the connection terminates.
1980 * * When a packet is sent.
1981 * * When a packet is received.
1983 * Code **-EINVAL** if the socket is not a full TCP socket;
1984 * otherwise, a positive number containing the bits that could not
1985 * be set is returned (which comes down to 0 if all bits were set
1988 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1990 * This helper is used in programs implementing policies at the
1991 * socket level. If the message *msg* is allowed to pass (i.e. if
1992 * the verdict eBPF program returns **SK_PASS**), redirect it to
1993 * the socket referenced by *map* (of type
1994 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1995 * egress interfaces can be used for redirection. The
1996 * **BPF_F_INGRESS** value in *flags* is used to make the
1997 * distinction (ingress path is selected if the flag is present,
1998 * egress path otherwise). This is the only flag supported for now.
2000 * **SK_PASS** on success, or **SK_DROP** on error.
2002 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2004 * For socket policies, apply the verdict of the eBPF program to
2005 * the next *bytes* (number of bytes) of message *msg*.
2007 * For example, this helper can be used in the following cases:
2009 * * A single **sendmsg**\ () or **sendfile**\ () system call
2010 * contains multiple logical messages that the eBPF program is
2011 * supposed to read and for which it should apply a verdict.
2012 * * An eBPF program only cares to read the first *bytes* of a
2013 * *msg*. If the message has a large payload, then setting up
2014 * and calling the eBPF program repeatedly for all bytes, even
2015 * though the verdict is already known, would create unnecessary
2018 * When called from within an eBPF program, the helper sets a
2019 * counter internal to the BPF infrastructure, that is used to
2020 * apply the last verdict to the next *bytes*. If *bytes* is
2021 * smaller than the current data being processed from a
2022 * **sendmsg**\ () or **sendfile**\ () system call, the first
2023 * *bytes* will be sent and the eBPF program will be re-run with
2024 * the pointer for start of data pointing to byte number *bytes*
2025 * **+ 1**. If *bytes* is larger than the current data being
2026 * processed, then the eBPF verdict will be applied to multiple
2027 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2030 * Note that if a socket closes with the internal counter holding
2031 * a non-zero value, this is not a problem because data is not
2032 * being buffered for *bytes* and is sent as it is received.
2036 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2038 * For socket policies, prevent the execution of the verdict eBPF
2039 * program for message *msg* until *bytes* (byte number) have been
2042 * This can be used when one needs a specific number of bytes
2043 * before a verdict can be assigned, even if the data spans
2044 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2045 * case would be a user calling **sendmsg**\ () repeatedly with
2046 * 1-byte long message segments. Obviously, this is bad for
2047 * performance, but it is still valid. If the eBPF program needs
2048 * *bytes* bytes to validate a header, this helper can be used to
2049 * prevent the eBPF program to be called again until *bytes* have
2054 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2056 * For socket policies, pull in non-linear data from user space
2057 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2058 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2061 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2062 * *msg* it can only parse data that the (**data**, **data_end**)
2063 * pointers have already consumed. For **sendmsg**\ () hooks this
2064 * is likely the first scatterlist element. But for calls relying
2065 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2066 * be the range (**0**, **0**) because the data is shared with
2067 * user space and by default the objective is to avoid allowing
2068 * user space to modify data while (or after) eBPF verdict is
2069 * being decided. This helper can be used to pull in data and to
2070 * set the start and end pointer to given values. Data will be
2071 * copied if necessary (i.e. if data was not linear and if start
2072 * and end pointers do not point to the same chunk).
2074 * A call to this helper is susceptible to change the underlying
2075 * packet buffer. Therefore, at load time, all checks on pointers
2076 * previously done by the verifier are invalidated and must be
2077 * performed again, if the helper is used in combination with
2078 * direct packet access.
2080 * All values for *flags* are reserved for future usage, and must
2083 * 0 on success, or a negative error in case of failure.
2085 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2087 * Bind the socket associated to *ctx* to the address pointed by
2088 * *addr*, of length *addr_len*. This allows for making outgoing
2089 * connection from the desired IP address, which can be useful for
2090 * example when all processes inside a cgroup should use one
2091 * single IP address on a host that has multiple IP configured.
2093 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2094 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2095 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2096 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2097 * behavior and lets the kernel efficiently pick up an unused
2098 * port as long as 4-tuple is unique. Passing non-zero port might
2099 * lead to degraded performance.
2101 * 0 on success, or a negative error in case of failure.
2103 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2105 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2106 * possible to both shrink and grow the packet tail.
2107 * Shrink done via *delta* being a negative integer.
2109 * A call to this helper is susceptible to change the underlying
2110 * packet buffer. Therefore, at load time, all checks on pointers
2111 * previously done by the verifier are invalidated and must be
2112 * performed again, if the helper is used in combination with
2113 * direct packet access.
2115 * 0 on success, or a negative error in case of failure.
2117 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2119 * Retrieve the XFRM state (IP transform framework, see also
2120 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2122 * The retrieved value is stored in the **struct bpf_xfrm_state**
2123 * pointed by *xfrm_state* and of length *size*.
2125 * All values for *flags* are reserved for future usage, and must
2128 * This helper is available only if the kernel was compiled with
2129 * **CONFIG_XFRM** configuration option.
2131 * 0 on success, or a negative error in case of failure.
2133 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2135 * Return a user or a kernel stack in bpf program provided buffer.
2136 * To achieve this, the helper needs *ctx*, which is a pointer
2137 * to the context on which the tracing program is executed.
2138 * To store the stacktrace, the bpf program provides *buf* with
2139 * a nonnegative *size*.
2141 * The last argument, *flags*, holds the number of stack frames to
2142 * skip (from 0 to 255), masked with
2143 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2144 * the following flags:
2146 * **BPF_F_USER_STACK**
2147 * Collect a user space stack instead of a kernel stack.
2148 * **BPF_F_USER_BUILD_ID**
2149 * Collect buildid+offset instead of ips for user stack,
2150 * only valid if **BPF_F_USER_STACK** is also specified.
2152 * **bpf_get_stack**\ () can collect up to
2153 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2154 * to sufficient large buffer size. Note that
2155 * this limit can be controlled with the **sysctl** program, and
2156 * that it should be manually increased in order to profile long
2157 * user stacks (such as stacks for Java programs). To do so, use:
2161 * # sysctl kernel.perf_event_max_stack=<new value>
2163 * A non-negative value equal to or less than *size* on success,
2164 * or a negative error in case of failure.
2166 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2168 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2169 * it provides an easy way to load *len* bytes from *offset*
2170 * from the packet associated to *skb*, into the buffer pointed
2171 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2172 * a fifth argument *start_header* exists in order to select a
2173 * base offset to start from. *start_header* can be one of:
2175 * **BPF_HDR_START_MAC**
2176 * Base offset to load data from is *skb*'s mac header.
2177 * **BPF_HDR_START_NET**
2178 * Base offset to load data from is *skb*'s network header.
2180 * In general, "direct packet access" is the preferred method to
2181 * access packet data, however, this helper is in particular useful
2182 * in socket filters where *skb*\ **->data** does not always point
2183 * to the start of the mac header and where "direct packet access"
2186 * 0 on success, or a negative error in case of failure.
2188 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2190 * Do FIB lookup in kernel tables using parameters in *params*.
2191 * If lookup is successful and result shows packet is to be
2192 * forwarded, the neighbor tables are searched for the nexthop.
2193 * If successful (ie., FIB lookup shows forwarding and nexthop
2194 * is resolved), the nexthop address is returned in ipv4_dst
2195 * or ipv6_dst based on family, smac is set to mac address of
2196 * egress device, dmac is set to nexthop mac address, rt_metric
2197 * is set to metric from route (IPv4/IPv6 only), and ifindex
2198 * is set to the device index of the nexthop from the FIB lookup.
2200 * *plen* argument is the size of the passed in struct.
2201 * *flags* argument can be a combination of one or more of the
2204 * **BPF_FIB_LOOKUP_DIRECT**
2205 * Do a direct table lookup vs full lookup using FIB
2207 * **BPF_FIB_LOOKUP_OUTPUT**
2208 * Perform lookup from an egress perspective (default is
2211 * *ctx* is either **struct xdp_md** for XDP programs or
2212 * **struct sk_buff** tc cls_act programs.
2214 * * < 0 if any input argument is invalid
2215 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2216 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2217 * packet is not forwarded or needs assist from full stack
2219 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2221 * Add an entry to, or update a sockhash *map* referencing sockets.
2222 * The *skops* is used as a new value for the entry associated to
2223 * *key*. *flags* is one of:
2226 * The entry for *key* must not exist in the map.
2228 * The entry for *key* must already exist in the map.
2230 * No condition on the existence of the entry for *key*.
2232 * If the *map* has eBPF programs (parser and verdict), those will
2233 * be inherited by the socket being added. If the socket is
2234 * already attached to eBPF programs, this results in an error.
2236 * 0 on success, or a negative error in case of failure.
2238 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2240 * This helper is used in programs implementing policies at the
2241 * socket level. If the message *msg* is allowed to pass (i.e. if
2242 * the verdict eBPF program returns **SK_PASS**), redirect it to
2243 * the socket referenced by *map* (of type
2244 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2245 * egress interfaces can be used for redirection. The
2246 * **BPF_F_INGRESS** value in *flags* is used to make the
2247 * distinction (ingress path is selected if the flag is present,
2248 * egress path otherwise). This is the only flag supported for now.
2250 * **SK_PASS** on success, or **SK_DROP** on error.
2252 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2254 * This helper is used in programs implementing policies at the
2255 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2256 * if the verdict eBPF program returns **SK_PASS**), redirect it
2257 * to the socket referenced by *map* (of type
2258 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2259 * egress interfaces can be used for redirection. The
2260 * **BPF_F_INGRESS** value in *flags* is used to make the
2261 * distinction (ingress path is selected if the flag is present,
2262 * egress otherwise). This is the only flag supported for now.
2264 * **SK_PASS** on success, or **SK_DROP** on error.
2266 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2268 * Encapsulate the packet associated to *skb* within a Layer 3
2269 * protocol header. This header is provided in the buffer at
2270 * address *hdr*, with *len* its size in bytes. *type* indicates
2271 * the protocol of the header and can be one of:
2273 * **BPF_LWT_ENCAP_SEG6**
2274 * IPv6 encapsulation with Segment Routing Header
2275 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2276 * the IPv6 header is computed by the kernel.
2277 * **BPF_LWT_ENCAP_SEG6_INLINE**
2278 * Only works if *skb* contains an IPv6 packet. Insert a
2279 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2281 * **BPF_LWT_ENCAP_IP**
2282 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2283 * must be IPv4 or IPv6, followed by zero or more
2284 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2285 * total bytes in all prepended headers. Please note that
2286 * if **skb_is_gso**\ (*skb*) is true, no more than two
2287 * headers can be prepended, and the inner header, if
2288 * present, should be either GRE or UDP/GUE.
2290 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2291 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2292 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2293 * **BPF_PROG_TYPE_LWT_XMIT**.
2295 * A call to this helper is susceptible to change the underlying
2296 * packet buffer. Therefore, at load time, all checks on pointers
2297 * previously done by the verifier are invalidated and must be
2298 * performed again, if the helper is used in combination with
2299 * direct packet access.
2301 * 0 on success, or a negative error in case of failure.
2303 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2305 * Store *len* bytes from address *from* into the packet
2306 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2307 * inside the outermost IPv6 Segment Routing Header can be
2308 * modified through this helper.
2310 * A call to this helper is susceptible to change the underlying
2311 * packet buffer. Therefore, at load time, all checks on pointers
2312 * previously done by the verifier are invalidated and must be
2313 * performed again, if the helper is used in combination with
2314 * direct packet access.
2316 * 0 on success, or a negative error in case of failure.
2318 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2320 * Adjust the size allocated to TLVs in the outermost IPv6
2321 * Segment Routing Header contained in the packet associated to
2322 * *skb*, at position *offset* by *delta* bytes. Only offsets
2323 * after the segments are accepted. *delta* can be as well
2324 * positive (growing) as negative (shrinking).
2326 * A call to this helper is susceptible to change the underlying
2327 * packet buffer. Therefore, at load time, all checks on pointers
2328 * previously done by the verifier are invalidated and must be
2329 * performed again, if the helper is used in combination with
2330 * direct packet access.
2332 * 0 on success, or a negative error in case of failure.
2334 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2336 * Apply an IPv6 Segment Routing action of type *action* to the
2337 * packet associated to *skb*. Each action takes a parameter
2338 * contained at address *param*, and of length *param_len* bytes.
2339 * *action* can be one of:
2341 * **SEG6_LOCAL_ACTION_END_X**
2342 * End.X action: Endpoint with Layer-3 cross-connect.
2343 * Type of *param*: **struct in6_addr**.
2344 * **SEG6_LOCAL_ACTION_END_T**
2345 * End.T action: Endpoint with specific IPv6 table lookup.
2346 * Type of *param*: **int**.
2347 * **SEG6_LOCAL_ACTION_END_B6**
2348 * End.B6 action: Endpoint bound to an SRv6 policy.
2349 * Type of *param*: **struct ipv6_sr_hdr**.
2350 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2351 * End.B6.Encap action: Endpoint bound to an SRv6
2352 * encapsulation policy.
2353 * Type of *param*: **struct ipv6_sr_hdr**.
2355 * A call to this helper is susceptible to change the underlying
2356 * packet buffer. Therefore, at load time, all checks on pointers
2357 * previously done by the verifier are invalidated and must be
2358 * performed again, if the helper is used in combination with
2359 * direct packet access.
2361 * 0 on success, or a negative error in case of failure.
2363 * long bpf_rc_repeat(void *ctx)
2365 * This helper is used in programs implementing IR decoding, to
2366 * report a successfully decoded repeat key message. This delays
2367 * the generation of a key up event for previously generated
2370 * Some IR protocols like NEC have a special IR message for
2371 * repeating last button, for when a button is held down.
2373 * The *ctx* should point to the lirc sample as passed into
2376 * This helper is only available is the kernel was compiled with
2377 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2382 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2384 * This helper is used in programs implementing IR decoding, to
2385 * report a successfully decoded key press with *scancode*,
2386 * *toggle* value in the given *protocol*. The scancode will be
2387 * translated to a keycode using the rc keymap, and reported as
2388 * an input key down event. After a period a key up event is
2389 * generated. This period can be extended by calling either
2390 * **bpf_rc_keydown**\ () again with the same values, or calling
2391 * **bpf_rc_repeat**\ ().
2393 * Some protocols include a toggle bit, in case the button was
2394 * released and pressed again between consecutive scancodes.
2396 * The *ctx* should point to the lirc sample as passed into
2399 * The *protocol* is the decoded protocol number (see
2400 * **enum rc_proto** for some predefined values).
2402 * This helper is only available is the kernel was compiled with
2403 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2408 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2410 * Return the cgroup v2 id of the socket associated with the *skb*.
2411 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2412 * helper for cgroup v1 by providing a tag resp. identifier that
2413 * can be matched on or used for map lookups e.g. to implement
2414 * policy. The cgroup v2 id of a given path in the hierarchy is
2415 * exposed in user space through the f_handle API in order to get
2416 * to the same 64-bit id.
2418 * This helper can be used on TC egress path, but not on ingress,
2419 * and is available only if the kernel was compiled with the
2420 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2422 * The id is returned or 0 in case the id could not be retrieved.
2424 * u64 bpf_get_current_cgroup_id(void)
2426 * A 64-bit integer containing the current cgroup id based
2427 * on the cgroup within which the current task is running.
2429 * void *bpf_get_local_storage(void *map, u64 flags)
2431 * Get the pointer to the local storage area.
2432 * The type and the size of the local storage is defined
2433 * by the *map* argument.
2434 * The *flags* meaning is specific for each map type,
2435 * and has to be 0 for cgroup local storage.
2437 * Depending on the BPF program type, a local storage area
2438 * can be shared between multiple instances of the BPF program,
2439 * running simultaneously.
2441 * A user should care about the synchronization by himself.
2442 * For example, by using the **BPF_STX_XADD** instruction to alter
2445 * A pointer to the local storage area.
2447 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2449 * Select a **SO_REUSEPORT** socket from a
2450 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2451 * It checks the selected socket is matching the incoming
2452 * request in the socket buffer.
2454 * 0 on success, or a negative error in case of failure.
2456 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2458 * Return id of cgroup v2 that is ancestor of cgroup associated
2459 * with the *skb* at the *ancestor_level*. The root cgroup is at
2460 * *ancestor_level* zero and each step down the hierarchy
2461 * increments the level. If *ancestor_level* == level of cgroup
2462 * associated with *skb*, then return value will be same as that
2463 * of **bpf_skb_cgroup_id**\ ().
2465 * The helper is useful to implement policies based on cgroups
2466 * that are upper in hierarchy than immediate cgroup associated
2469 * The format of returned id and helper limitations are same as in
2470 * **bpf_skb_cgroup_id**\ ().
2472 * The id is returned or 0 in case the id could not be retrieved.
2474 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2476 * Look for TCP socket matching *tuple*, optionally in a child
2477 * network namespace *netns*. The return value must be checked,
2478 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2480 * The *ctx* should point to the context of the program, such as
2481 * the skb or socket (depending on the hook in use). This is used
2482 * to determine the base network namespace for the lookup.
2484 * *tuple_size* must be one of:
2486 * **sizeof**\ (*tuple*\ **->ipv4**)
2487 * Look for an IPv4 socket.
2488 * **sizeof**\ (*tuple*\ **->ipv6**)
2489 * Look for an IPv6 socket.
2491 * If the *netns* is a negative signed 32-bit integer, then the
2492 * socket lookup table in the netns associated with the *ctx*
2493 * will be used. For the TC hooks, this is the netns of the device
2494 * in the skb. For socket hooks, this is the netns of the socket.
2495 * If *netns* is any other signed 32-bit value greater than or
2496 * equal to zero then it specifies the ID of the netns relative to
2497 * the netns associated with the *ctx*. *netns* values beyond the
2498 * range of 32-bit integers are reserved for future use.
2500 * All values for *flags* are reserved for future usage, and must
2503 * This helper is available only if the kernel was compiled with
2504 * **CONFIG_NET** configuration option.
2506 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2507 * For sockets with reuseport option, the **struct bpf_sock**
2508 * result is from *reuse*\ **->socks**\ [] using the hash of the
2511 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2513 * Look for UDP socket matching *tuple*, optionally in a child
2514 * network namespace *netns*. The return value must be checked,
2515 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2517 * The *ctx* should point to the context of the program, such as
2518 * the skb or socket (depending on the hook in use). This is used
2519 * to determine the base network namespace for the lookup.
2521 * *tuple_size* must be one of:
2523 * **sizeof**\ (*tuple*\ **->ipv4**)
2524 * Look for an IPv4 socket.
2525 * **sizeof**\ (*tuple*\ **->ipv6**)
2526 * Look for an IPv6 socket.
2528 * If the *netns* is a negative signed 32-bit integer, then the
2529 * socket lookup table in the netns associated with the *ctx*
2530 * will be used. For the TC hooks, this is the netns of the device
2531 * in the skb. For socket hooks, this is the netns of the socket.
2532 * If *netns* is any other signed 32-bit value greater than or
2533 * equal to zero then it specifies the ID of the netns relative to
2534 * the netns associated with the *ctx*. *netns* values beyond the
2535 * range of 32-bit integers are reserved for future use.
2537 * All values for *flags* are reserved for future usage, and must
2540 * This helper is available only if the kernel was compiled with
2541 * **CONFIG_NET** configuration option.
2543 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2544 * For sockets with reuseport option, the **struct bpf_sock**
2545 * result is from *reuse*\ **->socks**\ [] using the hash of the
2548 * long bpf_sk_release(void *sock)
2550 * Release the reference held by *sock*. *sock* must be a
2551 * non-**NULL** pointer that was returned from
2552 * **bpf_sk_lookup_xxx**\ ().
2554 * 0 on success, or a negative error in case of failure.
2556 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2558 * Push an element *value* in *map*. *flags* is one of:
2561 * If the queue/stack is full, the oldest element is
2562 * removed to make room for this.
2564 * 0 on success, or a negative error in case of failure.
2566 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
2568 * Pop an element from *map*.
2570 * 0 on success, or a negative error in case of failure.
2572 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
2574 * Get an element from *map* without removing it.
2576 * 0 on success, or a negative error in case of failure.
2578 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2580 * For socket policies, insert *len* bytes into *msg* at offset
2583 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2584 * *msg* it may want to insert metadata or options into the *msg*.
2585 * This can later be read and used by any of the lower layer BPF
2588 * This helper may fail if under memory pressure (a malloc
2589 * fails) in these cases BPF programs will get an appropriate
2590 * error and BPF programs will need to handle them.
2592 * 0 on success, or a negative error in case of failure.
2594 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2596 * Will remove *len* bytes from a *msg* starting at byte *start*.
2597 * This may result in **ENOMEM** errors under certain situations if
2598 * an allocation and copy are required due to a full ring buffer.
2599 * However, the helper will try to avoid doing the allocation
2600 * if possible. Other errors can occur if input parameters are
2601 * invalid either due to *start* byte not being valid part of *msg*
2602 * payload and/or *pop* value being to large.
2604 * 0 on success, or a negative error in case of failure.
2606 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2608 * This helper is used in programs implementing IR decoding, to
2609 * report a successfully decoded pointer movement.
2611 * The *ctx* should point to the lirc sample as passed into
2614 * This helper is only available is the kernel was compiled with
2615 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2620 * long bpf_spin_lock(struct bpf_spin_lock *lock)
2622 * Acquire a spinlock represented by the pointer *lock*, which is
2623 * stored as part of a value of a map. Taking the lock allows to
2624 * safely update the rest of the fields in that value. The
2625 * spinlock can (and must) later be released with a call to
2626 * **bpf_spin_unlock**\ (\ *lock*\ ).
2628 * Spinlocks in BPF programs come with a number of restrictions
2631 * * **bpf_spin_lock** objects are only allowed inside maps of
2632 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2633 * list could be extended in the future).
2634 * * BTF description of the map is mandatory.
2635 * * The BPF program can take ONE lock at a time, since taking two
2636 * or more could cause dead locks.
2637 * * Only one **struct bpf_spin_lock** is allowed per map element.
2638 * * When the lock is taken, calls (either BPF to BPF or helpers)
2640 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2641 * allowed inside a spinlock-ed region.
2642 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2643 * the lock, on all execution paths, before it returns.
2644 * * The BPF program can access **struct bpf_spin_lock** only via
2645 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2646 * helpers. Loading or storing data into the **struct
2647 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2648 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2649 * of the map value must be a struct and have **struct
2650 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2651 * Nested lock inside another struct is not allowed.
2652 * * The **struct bpf_spin_lock** *lock* field in a map value must
2653 * be aligned on a multiple of 4 bytes in that value.
2654 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2655 * the **bpf_spin_lock** field to user space.
2656 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2657 * a BPF program, do not update the **bpf_spin_lock** field.
2658 * * **bpf_spin_lock** cannot be on the stack or inside a
2659 * networking packet (it can only be inside of a map values).
2660 * * **bpf_spin_lock** is available to root only.
2661 * * Tracing programs and socket filter programs cannot use
2662 * **bpf_spin_lock**\ () due to insufficient preemption checks
2663 * (but this may change in the future).
2664 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2668 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
2670 * Release the *lock* previously locked by a call to
2671 * **bpf_spin_lock**\ (\ *lock*\ ).
2675 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2677 * This helper gets a **struct bpf_sock** pointer such
2678 * that all the fields in this **bpf_sock** can be accessed.
2680 * A **struct bpf_sock** pointer on success, or **NULL** in
2683 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2685 * This helper gets a **struct bpf_tcp_sock** pointer from a
2686 * **struct bpf_sock** pointer.
2688 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2691 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
2693 * Set ECN (Explicit Congestion Notification) field of IP header
2694 * to **CE** (Congestion Encountered) if current value is **ECT**
2695 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2698 * 1 if the **CE** flag is set (either by the current helper call
2699 * or because it was already present), 0 if it is not set.
2701 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2703 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2704 * **bpf_sk_release**\ () is unnecessary and not allowed.
2706 * A **struct bpf_sock** pointer on success, or **NULL** in
2709 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2711 * Look for TCP socket matching *tuple*, optionally in a child
2712 * network namespace *netns*. The return value must be checked,
2713 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2715 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2716 * that it also returns timewait or request sockets. Use
2717 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2720 * This helper is available only if the kernel was compiled with
2721 * **CONFIG_NET** configuration option.
2723 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2724 * For sockets with reuseport option, the **struct bpf_sock**
2725 * result is from *reuse*\ **->socks**\ [] using the hash of the
2728 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2730 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2731 * the listening socket in *sk*.
2733 * *iph* points to the start of the IPv4 or IPv6 header, while
2734 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2735 * **sizeof**\ (**struct ip6hdr**).
2737 * *th* points to the start of the TCP header, while *th_len*
2738 * contains **sizeof**\ (**struct tcphdr**).
2740 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2743 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2745 * Get name of sysctl in /proc/sys/ and copy it into provided by
2746 * program buffer *buf* of size *buf_len*.
2748 * The buffer is always NUL terminated, unless it's zero-sized.
2750 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2751 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2752 * only (e.g. "tcp_mem").
2754 * Number of character copied (not including the trailing NUL).
2756 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2757 * truncated name in this case).
2759 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2761 * Get current value of sysctl as it is presented in /proc/sys
2762 * (incl. newline, etc), and copy it as a string into provided
2763 * by program buffer *buf* of size *buf_len*.
2765 * The whole value is copied, no matter what file position user
2766 * space issued e.g. sys_read at.
2768 * The buffer is always NUL terminated, unless it's zero-sized.
2770 * Number of character copied (not including the trailing NUL).
2772 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2773 * truncated name in this case).
2775 * **-EINVAL** if current value was unavailable, e.g. because
2776 * sysctl is uninitialized and read returns -EIO for it.
2778 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2780 * Get new value being written by user space to sysctl (before
2781 * the actual write happens) and copy it as a string into
2782 * provided by program buffer *buf* of size *buf_len*.
2784 * User space may write new value at file position > 0.
2786 * The buffer is always NUL terminated, unless it's zero-sized.
2788 * Number of character copied (not including the trailing NUL).
2790 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2791 * truncated name in this case).
2793 * **-EINVAL** if sysctl is being read.
2795 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2797 * Override new value being written by user space to sysctl with
2798 * value provided by program in buffer *buf* of size *buf_len*.
2800 * *buf* should contain a string in same form as provided by user
2801 * space on sysctl write.
2803 * User space may write new value at file position > 0. To override
2804 * the whole sysctl value file position should be set to zero.
2808 * **-E2BIG** if the *buf_len* is too big.
2810 * **-EINVAL** if sysctl is being read.
2812 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2814 * Convert the initial part of the string from buffer *buf* of
2815 * size *buf_len* to a long integer according to the given base
2816 * and save the result in *res*.
2818 * The string may begin with an arbitrary amount of white space
2819 * (as determined by **isspace**\ (3)) followed by a single
2820 * optional '**-**' sign.
2822 * Five least significant bits of *flags* encode base, other bits
2823 * are currently unused.
2825 * Base must be either 8, 10, 16 or 0 to detect it automatically
2826 * similar to user space **strtol**\ (3).
2828 * Number of characters consumed on success. Must be positive but
2829 * no more than *buf_len*.
2831 * **-EINVAL** if no valid digits were found or unsupported base
2834 * **-ERANGE** if resulting value was out of range.
2836 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2838 * Convert the initial part of the string from buffer *buf* of
2839 * size *buf_len* to an unsigned long integer according to the
2840 * given base and save the result in *res*.
2842 * The string may begin with an arbitrary amount of white space
2843 * (as determined by **isspace**\ (3)).
2845 * Five least significant bits of *flags* encode base, other bits
2846 * are currently unused.
2848 * Base must be either 8, 10, 16 or 0 to detect it automatically
2849 * similar to user space **strtoul**\ (3).
2851 * Number of characters consumed on success. Must be positive but
2852 * no more than *buf_len*.
2854 * **-EINVAL** if no valid digits were found or unsupported base
2857 * **-ERANGE** if resulting value was out of range.
2859 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
2861 * Get a bpf-local-storage from a *sk*.
2863 * Logically, it could be thought of getting the value from
2864 * a *map* with *sk* as the **key**. From this
2865 * perspective, the usage is not much different from
2866 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2867 * helper enforces the key must be a full socket and the map must
2868 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2870 * Underneath, the value is stored locally at *sk* instead of
2871 * the *map*. The *map* is used as the bpf-local-storage
2872 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2873 * searched against all bpf-local-storages residing at *sk*.
2875 * *sk* is a kernel **struct sock** pointer for LSM program.
2876 * *sk* is a **struct bpf_sock** pointer for other program types.
2878 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2879 * used such that a new bpf-local-storage will be
2880 * created if one does not exist. *value* can be used
2881 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2882 * the initial value of a bpf-local-storage. If *value* is
2883 * **NULL**, the new bpf-local-storage will be zero initialized.
2885 * A bpf-local-storage pointer is returned on success.
2887 * **NULL** if not found or there was an error in adding
2888 * a new bpf-local-storage.
2890 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
2892 * Delete a bpf-local-storage from a *sk*.
2896 * **-ENOENT** if the bpf-local-storage cannot be found.
2897 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
2899 * long bpf_send_signal(u32 sig)
2901 * Send signal *sig* to the process of the current task.
2902 * The signal may be delivered to any of this process's threads.
2904 * 0 on success or successfully queued.
2906 * **-EBUSY** if work queue under nmi is full.
2908 * **-EINVAL** if *sig* is invalid.
2910 * **-EPERM** if no permission to send the *sig*.
2912 * **-EAGAIN** if bpf program can try again.
2914 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2916 * Try to issue a SYN cookie for the packet with corresponding
2917 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2919 * *iph* points to the start of the IPv4 or IPv6 header, while
2920 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2921 * **sizeof**\ (**struct ip6hdr**).
2923 * *th* points to the start of the TCP header, while *th_len*
2924 * contains the length of the TCP header.
2926 * On success, lower 32 bits hold the generated SYN cookie in
2927 * followed by 16 bits which hold the MSS value for that cookie,
2928 * and the top 16 bits are unused.
2930 * On failure, the returned value is one of the following:
2932 * **-EINVAL** SYN cookie cannot be issued due to error
2934 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2936 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2938 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2940 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2942 * Write raw *data* blob into a special BPF perf event held by
2943 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2944 * event must have the following attributes: **PERF_SAMPLE_RAW**
2945 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2946 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2948 * The *flags* are used to indicate the index in *map* for which
2949 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2950 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2951 * to indicate that the index of the current CPU core should be
2954 * The value to write, of *size*, is passed through eBPF stack and
2955 * pointed by *data*.
2957 * *ctx* is a pointer to in-kernel struct sk_buff.
2959 * This helper is similar to **bpf_perf_event_output**\ () but
2960 * restricted to raw_tracepoint bpf programs.
2962 * 0 on success, or a negative error in case of failure.
2964 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2966 * Safely attempt to read *size* bytes from user space address
2967 * *unsafe_ptr* and store the data in *dst*.
2969 * 0 on success, or a negative error in case of failure.
2971 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2973 * Safely attempt to read *size* bytes from kernel space address
2974 * *unsafe_ptr* and store the data in *dst*.
2976 * 0 on success, or a negative error in case of failure.
2978 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2980 * Copy a NUL terminated string from an unsafe user address
2981 * *unsafe_ptr* to *dst*. The *size* should include the
2982 * terminating NUL byte. In case the string length is smaller than
2983 * *size*, the target is not padded with further NUL bytes. If the
2984 * string length is larger than *size*, just *size*-1 bytes are
2985 * copied and the last byte is set to NUL.
2987 * On success, the length of the copied string is returned. This
2988 * makes this helper useful in tracing programs for reading
2989 * strings, and more importantly to get its length at runtime. See
2990 * the following snippet:
2994 * SEC("kprobe/sys_open")
2995 * void bpf_sys_open(struct pt_regs *ctx)
2997 * char buf[PATHLEN]; // PATHLEN is defined to 256
2998 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
3001 * // Consume buf, for example push it to
3002 * // userspace via bpf_perf_event_output(); we
3003 * // can use res (the string length) as event
3004 * // size, after checking its boundaries.
3007 * In comparison, using **bpf_probe_read_user**\ () helper here
3008 * instead to read the string would require to estimate the length
3009 * at compile time, and would often result in copying more memory
3012 * Another useful use case is when parsing individual process
3013 * arguments or individual environment variables navigating
3014 * *current*\ **->mm->arg_start** and *current*\
3015 * **->mm->env_start**: using this helper and the return value,
3016 * one can quickly iterate at the right offset of the memory area.
3018 * On success, the strictly positive length of the string,
3019 * including the trailing NUL character. On error, a negative
3022 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3024 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3025 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3027 * On success, the strictly positive length of the string, including
3028 * the trailing NUL character. On error, a negative value.
3030 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3032 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3033 * *rcv_nxt* is the ack_seq to be sent out.
3035 * 0 on success, or a negative error in case of failure.
3037 * long bpf_send_signal_thread(u32 sig)
3039 * Send signal *sig* to the thread corresponding to the current task.
3041 * 0 on success or successfully queued.
3043 * **-EBUSY** if work queue under nmi is full.
3045 * **-EINVAL** if *sig* is invalid.
3047 * **-EPERM** if no permission to send the *sig*.
3049 * **-EAGAIN** if bpf program can try again.
3051 * u64 bpf_jiffies64(void)
3053 * Obtain the 64bit jiffies
3055 * The 64 bit jiffies
3057 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3059 * For an eBPF program attached to a perf event, retrieve the
3060 * branch records (**struct perf_branch_entry**) associated to *ctx*
3061 * and store it in the buffer pointed by *buf* up to size
3064 * On success, number of bytes written to *buf*. On error, a
3067 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3068 * instead return the number of bytes required to store all the
3069 * branch entries. If this flag is set, *buf* may be NULL.
3071 * **-EINVAL** if arguments invalid or **size** not a multiple
3072 * of **sizeof**\ (**struct perf_branch_entry**\ ).
3074 * **-ENOENT** if architecture does not support branch records.
3076 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3078 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3079 * *namespace* will be returned in *nsdata*.
3081 * 0 on success, or one of the following in case of failure:
3083 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3084 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3086 * **-ENOENT** if pidns does not exists for the current task.
3088 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3090 * Write raw *data* blob into a special BPF perf event held by
3091 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3092 * event must have the following attributes: **PERF_SAMPLE_RAW**
3093 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3094 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3096 * The *flags* are used to indicate the index in *map* for which
3097 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3098 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3099 * to indicate that the index of the current CPU core should be
3102 * The value to write, of *size*, is passed through eBPF stack and
3103 * pointed by *data*.
3105 * *ctx* is a pointer to in-kernel struct xdp_buff.
3107 * This helper is similar to **bpf_perf_eventoutput**\ () but
3108 * restricted to raw_tracepoint bpf programs.
3110 * 0 on success, or a negative error in case of failure.
3112 * u64 bpf_get_netns_cookie(void *ctx)
3114 * Retrieve the cookie (generated by the kernel) of the network
3115 * namespace the input *ctx* is associated with. The network
3116 * namespace cookie remains stable for its lifetime and provides
3117 * a global identifier that can be assumed unique. If *ctx* is
3118 * NULL, then the helper returns the cookie for the initial
3119 * network namespace. The cookie itself is very similar to that
3120 * of **bpf_get_socket_cookie**\ () helper, but for network
3121 * namespaces instead of sockets.
3123 * A 8-byte long opaque number.
3125 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3127 * Return id of cgroup v2 that is ancestor of the cgroup associated
3128 * with the current task at the *ancestor_level*. The root cgroup
3129 * is at *ancestor_level* zero and each step down the hierarchy
3130 * increments the level. If *ancestor_level* == level of cgroup
3131 * associated with the current task, then return value will be the
3132 * same as that of **bpf_get_current_cgroup_id**\ ().
3134 * The helper is useful to implement policies based on cgroups
3135 * that are upper in hierarchy than immediate cgroup associated
3136 * with the current task.
3138 * The format of returned id and helper limitations are same as in
3139 * **bpf_get_current_cgroup_id**\ ().
3141 * The id is returned or 0 in case the id could not be retrieved.
3143 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
3145 * Helper is overloaded depending on BPF program type. This
3146 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3147 * **BPF_PROG_TYPE_SCHED_ACT** programs.
3149 * Assign the *sk* to the *skb*. When combined with appropriate
3150 * routing configuration to receive the packet towards the socket,
3151 * will cause *skb* to be delivered to the specified socket.
3152 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3153 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3154 * interfere with successful delivery to the socket.
3156 * This operation is only valid from TC ingress path.
3158 * The *flags* argument must be zero.
3160 * 0 on success, or a negative error in case of failure:
3162 * **-EINVAL** if specified *flags* are not supported.
3164 * **-ENOENT** if the socket is unavailable for assignment.
3166 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3168 * **-EOPNOTSUPP** if the operation is not supported, for example
3169 * a call from outside of TC ingress.
3171 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3174 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3176 * Helper is overloaded depending on BPF program type. This
3177 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
3179 * Select the *sk* as a result of a socket lookup.
3181 * For the operation to succeed passed socket must be compatible
3182 * with the packet description provided by the *ctx* object.
3184 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
3185 * be an exact match. While IP family (**AF_INET** or
3186 * **AF_INET6**) must be compatible, that is IPv6 sockets
3187 * that are not v6-only can be selected for IPv4 packets.
3189 * Only TCP listeners and UDP unconnected sockets can be
3190 * selected. *sk* can also be NULL to reset any previous
3193 * *flags* argument can combination of following values:
3195 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
3196 * socket selection, potentially done by a BPF program
3197 * that ran before us.
3199 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
3200 * load-balancing within reuseport group for the socket
3203 * On success *ctx->sk* will point to the selected socket.
3206 * 0 on success, or a negative errno in case of failure.
3208 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
3209 * not compatible with packet family (*ctx->family*).
3211 * * **-EEXIST** if socket has been already selected,
3212 * potentially by another program, and
3213 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
3215 * * **-EINVAL** if unsupported flags were specified.
3217 * * **-EPROTOTYPE** if socket L4 protocol
3218 * (*sk->protocol*) doesn't match packet protocol
3219 * (*ctx->protocol*).
3221 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
3222 * state (TCP listening or UDP unconnected).
3224 * u64 bpf_ktime_get_boot_ns(void)
3226 * Return the time elapsed since system boot, in nanoseconds.
3227 * Does include the time the system was suspended.
3228 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
3232 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3234 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3235 * out the format string.
3236 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3237 * the format string itself. The *data* and *data_len* are format string
3238 * arguments. The *data* are a **u64** array and corresponding format string
3239 * values are stored in the array. For strings and pointers where pointees
3240 * are accessed, only the pointer values are stored in the *data* array.
3241 * The *data_len* is the size of *data* in bytes.
3243 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3244 * Reading kernel memory may fail due to either invalid address or
3245 * valid address but requiring a major memory fault. If reading kernel memory
3246 * fails, the string for **%s** will be an empty string, and the ip
3247 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
3248 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
3250 * 0 on success, or a negative error in case of failure:
3252 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3253 * by returning 1 from bpf program.
3255 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3257 * **-E2BIG** if *fmt* contains too many format specifiers.
3259 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3261 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
3263 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
3264 * The *m* represents the seq_file. The *data* and *len* represent the
3265 * data to write in bytes.
3267 * 0 on success, or a negative error in case of failure:
3269 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3271 * u64 bpf_sk_cgroup_id(void *sk)
3273 * Return the cgroup v2 id of the socket *sk*.
3275 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
3276 * returned from **bpf_sk_lookup_xxx**\ (),
3277 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3278 * same as in **bpf_skb_cgroup_id**\ ().
3280 * This helper is available only if the kernel was compiled with
3281 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3283 * The id is returned or 0 in case the id could not be retrieved.
3285 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
3287 * Return id of cgroup v2 that is ancestor of cgroup associated
3288 * with the *sk* at the *ancestor_level*. The root cgroup is at
3289 * *ancestor_level* zero and each step down the hierarchy
3290 * increments the level. If *ancestor_level* == level of cgroup
3291 * associated with *sk*, then return value will be same as that
3292 * of **bpf_sk_cgroup_id**\ ().
3294 * The helper is useful to implement policies based on cgroups
3295 * that are upper in hierarchy than immediate cgroup associated
3298 * The format of returned id and helper limitations are same as in
3299 * **bpf_sk_cgroup_id**\ ().
3301 * The id is returned or 0 in case the id could not be retrieved.
3303 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
3305 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
3306 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3307 * of new data availability is sent.
3308 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3309 * of new data availability is sent unconditionally.
3311 * 0 on success, or a negative error in case of failure.
3313 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3315 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3317 * Valid pointer with *size* bytes of memory available; NULL,
3320 * void bpf_ringbuf_submit(void *data, u64 flags)
3322 * Submit reserved ring buffer sample, pointed to by *data*.
3323 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3324 * of new data availability is sent.
3325 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3326 * of new data availability is sent unconditionally.
3328 * Nothing. Always succeeds.
3330 * void bpf_ringbuf_discard(void *data, u64 flags)
3332 * Discard reserved ring buffer sample, pointed to by *data*.
3333 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3334 * of new data availability is sent.
3335 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3336 * of new data availability is sent unconditionally.
3338 * Nothing. Always succeeds.
3340 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3342 * Query various characteristics of provided ring buffer. What
3343 * exactly is queries is determined by *flags*:
3345 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
3346 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
3347 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
3348 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
3350 * Data returned is just a momentary snapshot of actual values
3351 * and could be inaccurate, so this facility should be used to
3352 * power heuristics and for reporting, not to make 100% correct
3355 * Requested value, or 0, if *flags* are not recognized.
3357 * long bpf_csum_level(struct sk_buff *skb, u64 level)
3359 * Change the skbs checksum level by one layer up or down, or
3360 * reset it entirely to none in order to have the stack perform
3361 * checksum validation. The level is applicable to the following
3362 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
3363 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
3364 * through **bpf_skb_adjust_room**\ () helper with passing in
3365 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
3366 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
3367 * the UDP header is removed. Similarly, an encap of the latter
3368 * into the former could be accompanied by a helper call to
3369 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
3370 * skb is still intended to be processed in higher layers of the
3371 * stack instead of just egressing at tc.
3373 * There are three supported level settings at this time:
3375 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
3376 * with CHECKSUM_UNNECESSARY.
3377 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
3378 * with CHECKSUM_UNNECESSARY.
3379 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
3380 * sets CHECKSUM_NONE to force checksum validation by the stack.
3381 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
3384 * 0 on success, or a negative error in case of failure. In the
3385 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
3386 * is returned or the error code -EACCES in case the skb is not
3387 * subject to CHECKSUM_UNNECESSARY.
3389 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
3391 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
3393 * *sk* if casting is valid, or **NULL** otherwise.
3395 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
3397 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
3399 * *sk* if casting is valid, or **NULL** otherwise.
3401 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
3403 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
3405 * *sk* if casting is valid, or **NULL** otherwise.
3407 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
3409 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
3411 * *sk* if casting is valid, or **NULL** otherwise.
3413 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
3415 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
3417 * *sk* if casting is valid, or **NULL** otherwise.
3419 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
3421 * Return a user or a kernel stack in bpf program provided buffer.
3422 * To achieve this, the helper needs *task*, which is a valid
3423 * pointer to **struct task_struct**. To store the stacktrace, the
3424 * bpf program provides *buf* with a nonnegative *size*.
3426 * The last argument, *flags*, holds the number of stack frames to
3427 * skip (from 0 to 255), masked with
3428 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3429 * the following flags:
3431 * **BPF_F_USER_STACK**
3432 * Collect a user space stack instead of a kernel stack.
3433 * **BPF_F_USER_BUILD_ID**
3434 * Collect buildid+offset instead of ips for user stack,
3435 * only valid if **BPF_F_USER_STACK** is also specified.
3437 * **bpf_get_task_stack**\ () can collect up to
3438 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3439 * to sufficient large buffer size. Note that
3440 * this limit can be controlled with the **sysctl** program, and
3441 * that it should be manually increased in order to profile long
3442 * user stacks (such as stacks for Java programs). To do so, use:
3446 * # sysctl kernel.perf_event_max_stack=<new value>
3448 * A non-negative value equal to or less than *size* on success,
3449 * or a negative error in case of failure.
3451 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
3453 * Load header option. Support reading a particular TCP header
3454 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
3456 * If *flags* is 0, it will search the option from the
3457 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
3458 * has details on what skb_data contains under different
3459 * *skops*\ **->op**.
3461 * The first byte of the *searchby_res* specifies the
3462 * kind that it wants to search.
3464 * If the searching kind is an experimental kind
3465 * (i.e. 253 or 254 according to RFC6994). It also
3466 * needs to specify the "magic" which is either
3467 * 2 bytes or 4 bytes. It then also needs to
3468 * specify the size of the magic by using
3469 * the 2nd byte which is "kind-length" of a TCP
3470 * header option and the "kind-length" also
3471 * includes the first 2 bytes "kind" and "kind-length"
3472 * itself as a normal TCP header option also does.
3474 * For example, to search experimental kind 254 with
3475 * 2 byte magic 0xeB9F, the searchby_res should be
3476 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
3478 * To search for the standard window scale option (3),
3479 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
3480 * Note, kind-length must be 0 for regular option.
3482 * Searching for No-Op (0) and End-of-Option-List (1) are
3485 * *len* must be at least 2 bytes which is the minimal size
3486 * of a header option.
3490 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
3491 * saved_syn packet or the just-received syn packet.
3494 * > 0 when found, the header option is copied to *searchby_res*.
3495 * The return value is the total length copied. On failure, a
3496 * negative error code is returned:
3498 * **-EINVAL** if a parameter is invalid.
3500 * **-ENOMSG** if the option is not found.
3502 * **-ENOENT** if no syn packet is available when
3503 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
3505 * **-ENOSPC** if there is not enough space. Only *len* number of
3508 * **-EFAULT** on failure to parse the header options in the
3511 * **-EPERM** if the helper cannot be used under the current
3512 * *skops*\ **->op**.
3514 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
3516 * Store header option. The data will be copied
3517 * from buffer *from* with length *len* to the TCP header.
3519 * The buffer *from* should have the whole option that
3520 * includes the kind, kind-length, and the actual
3521 * option data. The *len* must be at least kind-length
3522 * long. The kind-length does not have to be 4 byte
3523 * aligned. The kernel will take care of the padding
3524 * and setting the 4 bytes aligned value to th->doff.
3526 * This helper will check for duplicated option
3527 * by searching the same option in the outgoing skb.
3529 * This helper can only be called during
3530 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
3533 * 0 on success, or negative error in case of failure:
3535 * **-EINVAL** If param is invalid.
3537 * **-ENOSPC** if there is not enough space in the header.
3538 * Nothing has been written
3540 * **-EEXIST** if the option already exists.
3542 * **-EFAULT** on failrue to parse the existing header options.
3544 * **-EPERM** if the helper cannot be used under the current
3545 * *skops*\ **->op**.
3547 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
3549 * Reserve *len* bytes for the bpf header option. The
3550 * space will be used by **bpf_store_hdr_opt**\ () later in
3551 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
3553 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
3554 * the total number of bytes will be reserved.
3556 * This helper can only be called during
3557 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
3560 * 0 on success, or negative error in case of failure:
3562 * **-EINVAL** if a parameter is invalid.
3564 * **-ENOSPC** if there is not enough space in the header.
3566 * **-EPERM** if the helper cannot be used under the current
3567 * *skops*\ **->op**.
3569 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
3571 * Get a bpf_local_storage from an *inode*.
3573 * Logically, it could be thought of as getting the value from
3574 * a *map* with *inode* as the **key**. From this
3575 * perspective, the usage is not much different from
3576 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
3577 * helper enforces the key must be an inode and the map must also
3578 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
3580 * Underneath, the value is stored locally at *inode* instead of
3581 * the *map*. The *map* is used as the bpf-local-storage
3582 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3583 * searched against all bpf_local_storage residing at *inode*.
3585 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
3586 * used such that a new bpf_local_storage will be
3587 * created if one does not exist. *value* can be used
3588 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
3589 * the initial value of a bpf_local_storage. If *value* is
3590 * **NULL**, the new bpf_local_storage will be zero initialized.
3592 * A bpf_local_storage pointer is returned on success.
3594 * **NULL** if not found or there was an error in adding
3595 * a new bpf_local_storage.
3597 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
3599 * Delete a bpf_local_storage from an *inode*.
3603 * **-ENOENT** if the bpf_local_storage cannot be found.
3605 * long bpf_d_path(struct path *path, char *buf, u32 sz)
3607 * Return full path for given **struct path** object, which
3608 * needs to be the kernel BTF *path* object. The path is
3609 * returned in the provided buffer *buf* of size *sz* and
3610 * is zero terminated.
3613 * On success, the strictly positive length of the string,
3614 * including the trailing NUL character. On error, a negative
3617 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
3619 * Read *size* bytes from user space address *user_ptr* and store
3620 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
3622 * 0 on success, or a negative error in case of failure.
3624 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
3626 * Use BTF to store a string representation of *ptr*->ptr in *str*,
3627 * using *ptr*->type_id. This value should specify the type
3628 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
3629 * can be used to look up vmlinux BTF type ids. Traversing the
3630 * data structure using BTF, the type information and values are
3631 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
3632 * the pointer data is carried out to avoid kernel crashes during
3633 * operation. Smaller types can use string space on the stack;
3634 * larger programs can use map data to store the string
3637 * The string can be subsequently shared with userspace via
3638 * bpf_perf_event_output() or ring buffer interfaces.
3639 * bpf_trace_printk() is to be avoided as it places too small
3640 * a limit on string size to be useful.
3642 * *flags* is a combination of
3645 * no formatting around type information
3647 * no struct/union member names/types
3649 * show raw (unobfuscated) pointer values;
3650 * equivalent to printk specifier %px.
3652 * show zero-valued struct/union members; they
3653 * are not displayed by default
3656 * The number of bytes that were written (or would have been
3657 * written if output had to be truncated due to string size),
3658 * or a negative error in cases of failure.
3660 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
3662 * Use BTF to write to seq_write a string representation of
3663 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
3664 * *flags* are identical to those used for bpf_snprintf_btf.
3666 * 0 on success or a negative error in case of failure.
3668 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
3670 * See **bpf_get_cgroup_classid**\ () for the main description.
3671 * This helper differs from **bpf_get_cgroup_classid**\ () in that
3672 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
3673 * associated socket instead of the current process.
3675 * The id is returned or 0 in case the id could not be retrieved.
3677 * long bpf_redirect_neigh(u32 ifindex, u64 flags)
3679 * Redirect the packet to another net device of index *ifindex*
3680 * and fill in L2 addresses from neighboring subsystem. This helper
3681 * is somewhat similar to **bpf_redirect**\ (), except that it
3682 * fills in e.g. MAC addresses based on the L3 information from
3683 * the packet. This helper is supported for IPv4 and IPv6 protocols.
3684 * The *flags* argument is reserved and must be 0. The helper is
3685 * currently only supported for tc BPF program types.
3687 * The helper returns **TC_ACT_REDIRECT** on success or
3688 * **TC_ACT_SHOT** on error.
3690 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
3692 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
3693 * pointer to the percpu kernel variable on *cpu*. A ksym is an
3694 * extern variable decorated with '__ksym'. For ksym, there is a
3695 * global var (either static or global) defined of the same name
3696 * in the kernel. The ksym is percpu if the global var is percpu.
3697 * The returned pointer points to the global percpu var on *cpu*.
3699 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
3700 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
3701 * happens if *cpu* is larger than nr_cpu_ids. The caller of
3702 * bpf_per_cpu_ptr() must check the returned value.
3704 * A pointer pointing to the kernel percpu variable on *cpu*, or
3705 * NULL, if *cpu* is invalid.
3707 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
3709 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
3710 * pointer to the percpu kernel variable on this cpu. See the
3711 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
3713 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
3714 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
3715 * never return NULL.
3717 * A pointer pointing to the kernel percpu variable on this cpu.
3719 #define __BPF_FUNC_MAPPER(FN) \
3721 FN(map_lookup_elem), \
3722 FN(map_update_elem), \
3723 FN(map_delete_elem), \
3727 FN(get_prandom_u32), \
3728 FN(get_smp_processor_id), \
3729 FN(skb_store_bytes), \
3730 FN(l3_csum_replace), \
3731 FN(l4_csum_replace), \
3733 FN(clone_redirect), \
3734 FN(get_current_pid_tgid), \
3735 FN(get_current_uid_gid), \
3736 FN(get_current_comm), \
3737 FN(get_cgroup_classid), \
3738 FN(skb_vlan_push), \
3740 FN(skb_get_tunnel_key), \
3741 FN(skb_set_tunnel_key), \
3742 FN(perf_event_read), \
3744 FN(get_route_realm), \
3745 FN(perf_event_output), \
3746 FN(skb_load_bytes), \
3749 FN(skb_get_tunnel_opt), \
3750 FN(skb_set_tunnel_opt), \
3751 FN(skb_change_proto), \
3752 FN(skb_change_type), \
3753 FN(skb_under_cgroup), \
3754 FN(get_hash_recalc), \
3755 FN(get_current_task), \
3756 FN(probe_write_user), \
3757 FN(current_task_under_cgroup), \
3758 FN(skb_change_tail), \
3759 FN(skb_pull_data), \
3761 FN(set_hash_invalid), \
3762 FN(get_numa_node_id), \
3763 FN(skb_change_head), \
3764 FN(xdp_adjust_head), \
3765 FN(probe_read_str), \
3766 FN(get_socket_cookie), \
3767 FN(get_socket_uid), \
3770 FN(skb_adjust_room), \
3772 FN(sk_redirect_map), \
3773 FN(sock_map_update), \
3774 FN(xdp_adjust_meta), \
3775 FN(perf_event_read_value), \
3776 FN(perf_prog_read_value), \
3778 FN(override_return), \
3779 FN(sock_ops_cb_flags_set), \
3780 FN(msg_redirect_map), \
3781 FN(msg_apply_bytes), \
3782 FN(msg_cork_bytes), \
3783 FN(msg_pull_data), \
3785 FN(xdp_adjust_tail), \
3786 FN(skb_get_xfrm_state), \
3788 FN(skb_load_bytes_relative), \
3790 FN(sock_hash_update), \
3791 FN(msg_redirect_hash), \
3792 FN(sk_redirect_hash), \
3793 FN(lwt_push_encap), \
3794 FN(lwt_seg6_store_bytes), \
3795 FN(lwt_seg6_adjust_srh), \
3796 FN(lwt_seg6_action), \
3799 FN(skb_cgroup_id), \
3800 FN(get_current_cgroup_id), \
3801 FN(get_local_storage), \
3802 FN(sk_select_reuseport), \
3803 FN(skb_ancestor_cgroup_id), \
3804 FN(sk_lookup_tcp), \
3805 FN(sk_lookup_udp), \
3807 FN(map_push_elem), \
3809 FN(map_peek_elem), \
3810 FN(msg_push_data), \
3812 FN(rc_pointer_rel), \
3817 FN(skb_ecn_set_ce), \
3818 FN(get_listener_sock), \
3819 FN(skc_lookup_tcp), \
3820 FN(tcp_check_syncookie), \
3821 FN(sysctl_get_name), \
3822 FN(sysctl_get_current_value), \
3823 FN(sysctl_get_new_value), \
3824 FN(sysctl_set_new_value), \
3827 FN(sk_storage_get), \
3828 FN(sk_storage_delete), \
3830 FN(tcp_gen_syncookie), \
3832 FN(probe_read_user), \
3833 FN(probe_read_kernel), \
3834 FN(probe_read_user_str), \
3835 FN(probe_read_kernel_str), \
3837 FN(send_signal_thread), \
3839 FN(read_branch_records), \
3840 FN(get_ns_current_pid_tgid), \
3842 FN(get_netns_cookie), \
3843 FN(get_current_ancestor_cgroup_id), \
3845 FN(ktime_get_boot_ns), \
3849 FN(sk_ancestor_cgroup_id), \
3850 FN(ringbuf_output), \
3851 FN(ringbuf_reserve), \
3852 FN(ringbuf_submit), \
3853 FN(ringbuf_discard), \
3854 FN(ringbuf_query), \
3856 FN(skc_to_tcp6_sock), \
3857 FN(skc_to_tcp_sock), \
3858 FN(skc_to_tcp_timewait_sock), \
3859 FN(skc_to_tcp_request_sock), \
3860 FN(skc_to_udp6_sock), \
3861 FN(get_task_stack), \
3863 FN(store_hdr_opt), \
3864 FN(reserve_hdr_opt), \
3865 FN(inode_storage_get), \
3866 FN(inode_storage_delete), \
3868 FN(copy_from_user), \
3870 FN(seq_printf_btf), \
3871 FN(skb_cgroup_classid), \
3872 FN(redirect_neigh), \
3873 FN(bpf_per_cpu_ptr), \
3874 FN(bpf_this_cpu_ptr), \
3877 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
3878 * function eBPF program intends to call
3880 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
3882 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
3885 #undef __BPF_ENUM_FN
3887 /* All flags used by eBPF helper functions, placed here. */
3889 /* BPF_FUNC_skb_store_bytes flags. */
3891 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3892 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3895 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3896 * First 4 bits are for passing the header field size.
3899 BPF_F_HDR_FIELD_MASK = 0xfULL,
3902 /* BPF_FUNC_l4_csum_replace flags. */
3904 BPF_F_PSEUDO_HDR = (1ULL << 4),
3905 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3906 BPF_F_MARK_ENFORCE = (1ULL << 6),
3909 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
3911 BPF_F_INGRESS = (1ULL << 0),
3914 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
3916 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3919 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
3921 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3922 BPF_F_USER_STACK = (1ULL << 8),
3923 /* flags used by BPF_FUNC_get_stackid only. */
3924 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3925 BPF_F_REUSE_STACKID = (1ULL << 10),
3926 /* flags used by BPF_FUNC_get_stack only. */
3927 BPF_F_USER_BUILD_ID = (1ULL << 11),
3930 /* BPF_FUNC_skb_set_tunnel_key flags. */
3932 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3933 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3934 BPF_F_SEQ_NUMBER = (1ULL << 3),
3937 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3938 * BPF_FUNC_perf_event_read_value flags.
3941 BPF_F_INDEX_MASK = 0xffffffffULL,
3942 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
3943 /* BPF_FUNC_perf_event_output for sk_buff input context. */
3944 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3947 /* Current network namespace */
3949 BPF_F_CURRENT_NETNS = (-1L),
3952 /* BPF_FUNC_csum_level level values. */
3954 BPF_CSUM_LEVEL_QUERY,
3957 BPF_CSUM_LEVEL_RESET,
3960 /* BPF_FUNC_skb_adjust_room flags. */
3962 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3963 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3964 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3965 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3966 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
3967 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
3971 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3972 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3975 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
3976 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3977 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
3979 /* BPF_FUNC_sysctl_get_name flags. */
3981 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3984 /* BPF_FUNC_<kernel_obj>_storage_get flags */
3986 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
3987 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
3988 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
3990 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
3993 /* BPF_FUNC_read_branch_records flags. */
3995 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3998 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
3999 * BPF_FUNC_bpf_ringbuf_output flags.
4002 BPF_RB_NO_WAKEUP = (1ULL << 0),
4003 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
4006 /* BPF_FUNC_bpf_ringbuf_query flags */
4008 BPF_RB_AVAIL_DATA = 0,
4009 BPF_RB_RING_SIZE = 1,
4010 BPF_RB_CONS_POS = 2,
4011 BPF_RB_PROD_POS = 3,
4014 /* BPF ring buffer constants */
4016 BPF_RINGBUF_BUSY_BIT = (1U << 31),
4017 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
4018 BPF_RINGBUF_HDR_SZ = 8,
4021 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
4023 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
4024 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
4027 /* Mode for BPF_FUNC_skb_adjust_room helper. */
4028 enum bpf_adj_room_mode {
4033 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
4034 enum bpf_hdr_start_off {
4039 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
4040 enum bpf_lwt_encap_mode {
4042 BPF_LWT_ENCAP_SEG6_INLINE,
4046 #define __bpf_md_ptr(type, name) \
4050 } __attribute__((aligned(8)))
4052 /* user accessible mirror of in-kernel sk_buff.
4053 * new fields can only be added to the end of this structure
4059 __u32 queue_mapping;
4065 __u32 ingress_ifindex;
4075 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
4077 __u32 remote_ip4; /* Stored in network byte order */
4078 __u32 local_ip4; /* Stored in network byte order */
4079 __u32 remote_ip6[4]; /* Stored in network byte order */
4080 __u32 local_ip6[4]; /* Stored in network byte order */
4081 __u32 remote_port; /* Stored in network byte order */
4082 __u32 local_port; /* stored in host byte order */
4086 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
4090 __bpf_md_ptr(struct bpf_sock *, sk);
4094 struct bpf_tunnel_key {
4098 __u32 remote_ipv6[4];
4102 __u16 tunnel_ext; /* Padding, future use. */
4106 /* user accessible mirror of in-kernel xfrm_state.
4107 * new fields can only be added to the end of this structure
4109 struct bpf_xfrm_state {
4111 __u32 spi; /* Stored in network byte order */
4113 __u16 ext; /* Padding, future use. */
4115 __u32 remote_ipv4; /* Stored in network byte order */
4116 __u32 remote_ipv6[4]; /* Stored in network byte order */
4120 /* Generic BPF return codes which all BPF program types may support.
4121 * The values are binary compatible with their TC_ACT_* counter-part to
4122 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
4125 * XDP is handled seprately, see XDP_*.
4133 /* >127 are reserved for prog type specific return codes.
4135 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
4136 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
4137 * changed and should be routed based on its new L3 header.
4138 * (This is an L3 redirect, as opposed to L2 redirect
4139 * represented by BPF_REDIRECT above).
4141 BPF_LWT_REROUTE = 128,
4151 /* IP address also allows 1 and 2 bytes access */
4154 __u32 src_port; /* host byte order */
4155 __u32 dst_port; /* network byte order */
4159 __s32 rx_queue_mapping;
4162 struct bpf_tcp_sock {
4163 __u32 snd_cwnd; /* Sending congestion window */
4164 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
4166 __u32 snd_ssthresh; /* Slow start size threshold */
4167 __u32 rcv_nxt; /* What we want to receive next */
4168 __u32 snd_nxt; /* Next sequence we send */
4169 __u32 snd_una; /* First byte we want an ack for */
4170 __u32 mss_cache; /* Cached effective mss, not including SACKS */
4171 __u32 ecn_flags; /* ECN status bits. */
4172 __u32 rate_delivered; /* saved rate sample: packets delivered */
4173 __u32 rate_interval_us; /* saved rate sample: time elapsed */
4174 __u32 packets_out; /* Packets which are "in flight" */
4175 __u32 retrans_out; /* Retransmitted packets out */
4176 __u32 total_retrans; /* Total retransmits for entire connection */
4177 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
4178 * total number of segments in.
4180 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
4181 * total number of data segments in.
4183 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
4184 * The total number of segments sent.
4186 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
4187 * total number of data segments sent.
4189 __u32 lost_out; /* Lost packets */
4190 __u32 sacked_out; /* SACK'd packets */
4191 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
4192 * sum(delta(rcv_nxt)), or how many bytes
4195 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
4196 * sum(delta(snd_una)), or how many bytes
4199 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
4200 * total number of DSACK blocks received
4202 __u32 delivered; /* Total data packets delivered incl. rexmits */
4203 __u32 delivered_ce; /* Like the above but only ECE marked packets */
4204 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
4207 struct bpf_sock_tuple {
4224 struct bpf_xdp_sock {
4228 #define XDP_PACKET_HEADROOM 256
4230 /* User return codes for XDP prog type.
4231 * A valid XDP program must return one of these defined values. All other
4232 * return codes are reserved for future use. Unknown return codes will
4233 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
4243 /* user accessible metadata for XDP packet hook
4244 * new fields must be added to the end of this structure
4250 /* Below access go through struct xdp_rxq_info */
4251 __u32 ingress_ifindex; /* rxq->dev->ifindex */
4252 __u32 rx_queue_index; /* rxq->queue_index */
4254 __u32 egress_ifindex; /* txq->dev->ifindex */
4257 /* DEVMAP map-value layout
4259 * The struct data-layout of map-value is a configuration interface.
4260 * New members can only be added to the end of this structure.
4262 struct bpf_devmap_val {
4263 __u32 ifindex; /* device index */
4265 int fd; /* prog fd on map write */
4266 __u32 id; /* prog id on map read */
4270 /* CPUMAP map-value layout
4272 * The struct data-layout of map-value is a configuration interface.
4273 * New members can only be added to the end of this structure.
4275 struct bpf_cpumap_val {
4276 __u32 qsize; /* queue size to remote target CPU */
4278 int fd; /* prog fd on map write */
4279 __u32 id; /* prog id on map read */
4288 /* user accessible metadata for SK_MSG packet hook, new fields must
4289 * be added to the end of this structure
4292 __bpf_md_ptr(void *, data);
4293 __bpf_md_ptr(void *, data_end);
4296 __u32 remote_ip4; /* Stored in network byte order */
4297 __u32 local_ip4; /* Stored in network byte order */
4298 __u32 remote_ip6[4]; /* Stored in network byte order */
4299 __u32 local_ip6[4]; /* Stored in network byte order */
4300 __u32 remote_port; /* Stored in network byte order */
4301 __u32 local_port; /* stored in host byte order */
4302 __u32 size; /* Total size of sk_msg */
4304 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
4307 struct sk_reuseport_md {
4309 * Start of directly accessible data. It begins from
4310 * the tcp/udp header.
4312 __bpf_md_ptr(void *, data);
4313 /* End of directly accessible data */
4314 __bpf_md_ptr(void *, data_end);
4316 * Total length of packet (starting from the tcp/udp header).
4317 * Note that the directly accessible bytes (data_end - data)
4318 * could be less than this "len". Those bytes could be
4319 * indirectly read by a helper "bpf_skb_load_bytes()".
4323 * Eth protocol in the mac header (network byte order). e.g.
4324 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
4327 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
4328 __u32 bind_inany; /* Is sock bound to an INANY address? */
4329 __u32 hash; /* A hash of the packet 4 tuples */
4332 #define BPF_TAG_SIZE 8
4334 struct bpf_prog_info {
4337 __u8 tag[BPF_TAG_SIZE];
4338 __u32 jited_prog_len;
4339 __u32 xlated_prog_len;
4340 __aligned_u64 jited_prog_insns;
4341 __aligned_u64 xlated_prog_insns;
4342 __u64 load_time; /* ns since boottime */
4343 __u32 created_by_uid;
4345 __aligned_u64 map_ids;
4346 char name[BPF_OBJ_NAME_LEN];
4348 __u32 gpl_compatible:1;
4349 __u32 :31; /* alignment pad */
4352 __u32 nr_jited_ksyms;
4353 __u32 nr_jited_func_lens;
4354 __aligned_u64 jited_ksyms;
4355 __aligned_u64 jited_func_lens;
4357 __u32 func_info_rec_size;
4358 __aligned_u64 func_info;
4361 __aligned_u64 line_info;
4362 __aligned_u64 jited_line_info;
4363 __u32 nr_jited_line_info;
4364 __u32 line_info_rec_size;
4365 __u32 jited_line_info_rec_size;
4367 __aligned_u64 prog_tags;
4370 } __attribute__((aligned(8)));
4372 struct bpf_map_info {
4379 char name[BPF_OBJ_NAME_LEN];
4381 __u32 btf_vmlinux_value_type_id;
4385 __u32 btf_key_type_id;
4386 __u32 btf_value_type_id;
4387 } __attribute__((aligned(8)));
4389 struct bpf_btf_info {
4393 } __attribute__((aligned(8)));
4395 struct bpf_link_info {
4401 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
4402 __u32 tp_name_len; /* in/out: tp_name buffer len */
4412 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
4413 __u32 target_name_len; /* in/out: target_name buffer len */
4428 } __attribute__((aligned(8)));
4430 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
4431 * by user and intended to be used by socket (e.g. to bind to, depends on
4434 struct bpf_sock_addr {
4435 __u32 user_family; /* Allows 4-byte read, but no write. */
4436 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4437 * Stored in network byte order.
4439 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4440 * Stored in network byte order.
4442 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
4443 * Stored in network byte order
4445 __u32 family; /* Allows 4-byte read, but no write */
4446 __u32 type; /* Allows 4-byte read, but no write */
4447 __u32 protocol; /* Allows 4-byte read, but no write */
4448 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4449 * Stored in network byte order.
4451 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4452 * Stored in network byte order.
4454 __bpf_md_ptr(struct bpf_sock *, sk);
4457 /* User bpf_sock_ops struct to access socket values and specify request ops
4458 * and their replies.
4459 * Some of this fields are in network (bigendian) byte order and may need
4460 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
4461 * New fields can only be added at the end of this structure
4463 struct bpf_sock_ops {
4466 __u32 args[4]; /* Optionally passed to bpf program */
4467 __u32 reply; /* Returned by bpf program */
4468 __u32 replylong[4]; /* Optionally returned by bpf prog */
4471 __u32 remote_ip4; /* Stored in network byte order */
4472 __u32 local_ip4; /* Stored in network byte order */
4473 __u32 remote_ip6[4]; /* Stored in network byte order */
4474 __u32 local_ip6[4]; /* Stored in network byte order */
4475 __u32 remote_port; /* Stored in network byte order */
4476 __u32 local_port; /* stored in host byte order */
4477 __u32 is_fullsock; /* Some TCP fields are only valid if
4478 * there is a full socket. If not, the
4479 * fields read as zero.
4482 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
4483 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
4492 __u32 rate_delivered;
4493 __u32 rate_interval_us;
4496 __u32 total_retrans;
4500 __u32 data_segs_out;
4504 __u64 bytes_received;
4506 __bpf_md_ptr(struct bpf_sock *, sk);
4507 /* [skb_data, skb_data_end) covers the whole TCP header.
4509 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
4510 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
4511 * header has not been written.
4512 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
4513 * been written so far.
4514 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
4516 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
4519 * bpf_load_hdr_opt() can also be used to read a particular option.
4521 __bpf_md_ptr(void *, skb_data);
4522 __bpf_md_ptr(void *, skb_data_end);
4523 __u32 skb_len; /* The total length of a packet.
4524 * It includes the header, options,
4527 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
4528 * an easy way to check for tcp_flags
4529 * without parsing skb_data.
4531 * In particular, the skb_tcp_flags
4532 * will still be available in
4533 * BPF_SOCK_OPS_HDR_OPT_LEN even though
4534 * the outgoing header has not
4539 /* Definitions for bpf_sock_ops_cb_flags */
4541 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
4542 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
4543 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
4544 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
4545 /* Call bpf for all received TCP headers. The bpf prog will be
4546 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4548 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4549 * for the header option related helpers that will be useful
4550 * to the bpf programs.
4552 * It could be used at the client/active side (i.e. connect() side)
4553 * when the server told it that the server was in syncookie
4554 * mode and required the active side to resend the bpf-written
4555 * options. The active side can keep writing the bpf-options until
4556 * it received a valid packet from the server side to confirm
4557 * the earlier packet (and options) has been received. The later
4558 * example patch is using it like this at the active side when the
4559 * server is in syncookie mode.
4561 * The bpf prog will usually turn this off in the common cases.
4563 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
4564 /* Call bpf when kernel has received a header option that
4565 * the kernel cannot handle. The bpf prog will be called under
4566 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
4568 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4569 * for the header option related helpers that will be useful
4570 * to the bpf programs.
4572 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
4573 /* Call bpf when the kernel is writing header options for the
4574 * outgoing packet. The bpf prog will first be called
4575 * to reserve space in a skb under
4576 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
4577 * the bpf prog will be called to write the header option(s)
4578 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4580 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
4581 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
4582 * related helpers that will be useful to the bpf programs.
4584 * The kernel gets its chance to reserve space and write
4585 * options first before the BPF program does.
4587 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
4588 /* Mask of all currently supported cb flags */
4589 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
4592 /* List of known BPF sock_ops operators.
4593 * New entries can only be added at the end
4597 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
4598 * -1 if default value should be used
4600 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
4601 * window (in packets) or -1 if default
4602 * value should be used
4604 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
4605 * active connection is initialized
4607 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
4608 * active connection is
4611 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
4612 * passive connection is
4615 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
4618 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
4619 * based on the path and may be
4620 * dependent on the congestion control
4621 * algorithm. In general it indicates
4622 * a congestion threshold. RTTs above
4623 * this indicate congestion
4625 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
4626 * Arg1: value of icsk_retransmits
4627 * Arg2: value of icsk_rto
4628 * Arg3: whether RTO has expired
4630 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
4631 * Arg1: sequence number of 1st byte
4633 * Arg3: return value of
4634 * tcp_transmit_skb (0 => success)
4636 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
4640 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
4641 * socket transition to LISTEN state.
4643 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
4645 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
4646 * It will be called to handle
4647 * the packets received at
4648 * an already established
4651 * sock_ops->skb_data:
4652 * Referring to the received skb.
4653 * It covers the TCP header only.
4655 * bpf_load_hdr_opt() can also
4656 * be used to search for a
4657 * particular option.
4659 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
4660 * header option later in
4661 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4662 * Arg1: bool want_cookie. (in
4663 * writing SYNACK only)
4665 * sock_ops->skb_data:
4666 * Not available because no header has
4669 * sock_ops->skb_tcp_flags:
4670 * The tcp_flags of the
4671 * outgoing skb. (e.g. SYN, ACK, FIN).
4673 * bpf_reserve_hdr_opt() should
4674 * be used to reserve space.
4676 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
4677 * Arg1: bool want_cookie. (in
4678 * writing SYNACK only)
4680 * sock_ops->skb_data:
4681 * Referring to the outgoing skb.
4682 * It covers the TCP header
4683 * that has already been written
4684 * by the kernel and the
4685 * earlier bpf-progs.
4687 * sock_ops->skb_tcp_flags:
4688 * The tcp_flags of the outgoing
4689 * skb. (e.g. SYN, ACK, FIN).
4691 * bpf_store_hdr_opt() should
4692 * be used to write the
4695 * bpf_load_hdr_opt() can also
4696 * be used to search for a
4697 * particular option that
4698 * has already been written
4699 * by the kernel or the
4700 * earlier bpf-progs.
4704 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
4705 * changes between the TCP and BPF versions. Ideally this should never happen.
4706 * If it does, we need to add code to convert them before calling
4707 * the BPF sock_ops function.
4710 BPF_TCP_ESTABLISHED = 1,
4720 BPF_TCP_CLOSING, /* Now a valid state */
4721 BPF_TCP_NEW_SYN_RECV,
4723 BPF_TCP_MAX_STATES /* Leave at the end! */
4727 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4728 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
4729 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
4730 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
4731 /* Copy the SYN pkt to optval
4733 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
4734 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
4735 * to only getting from the saved_syn. It can either get the
4738 * 1. the just-received SYN packet (only available when writing the
4739 * SYNACK). It will be useful when it is not necessary to
4740 * save the SYN packet for latter use. It is also the only way
4741 * to get the SYN during syncookie mode because the syn
4742 * packet cannot be saved during syncookie.
4746 * 2. the earlier saved syn which was done by
4747 * bpf_setsockopt(TCP_SAVE_SYN).
4749 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
4750 * SYN packet is obtained.
4752 * If the bpf-prog does not need the IP[46] header, the
4753 * bpf-prog can avoid parsing the IP header by using
4754 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
4755 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
4757 * >0: Total number of bytes copied
4758 * -ENOSPC: Not enough space in optval. Only optlen number of
4760 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
4761 * is not saved by setsockopt(TCP_SAVE_SYN).
4763 TCP_BPF_SYN = 1005, /* Copy the TCP header */
4764 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
4765 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
4769 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
4772 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
4773 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4776 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
4777 * total option spaces
4778 * required for an established
4779 * sk in order to calculate the
4780 * MSS. No skb is actually
4783 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
4784 * when sending a SYN.
4788 struct bpf_perf_event_value {
4795 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
4796 BPF_DEVCG_ACC_READ = (1ULL << 1),
4797 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
4801 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
4802 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
4805 struct bpf_cgroup_dev_ctx {
4806 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
4812 struct bpf_raw_tracepoint_args {
4816 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
4817 * OUTPUT: Do lookup from egress perspective; default is ingress
4820 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
4821 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
4825 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
4826 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
4827 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
4828 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
4829 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
4830 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
4831 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
4832 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
4833 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
4836 struct bpf_fib_lookup {
4837 /* input: network family for lookup (AF_INET, AF_INET6)
4838 * output: network family of egress nexthop
4842 /* set if lookup is to consider L4 data - e.g., FIB rules */
4847 /* total length of packet from network header - used for MTU check */
4850 /* input: L3 device index for lookup
4851 * output: device index from FIB lookup
4856 /* inputs to lookup */
4857 __u8 tos; /* AF_INET */
4858 __be32 flowinfo; /* AF_INET6, flow_label + priority */
4860 /* output: metric of fib result (IPv4/IPv6 only) */
4866 __u32 ipv6_src[4]; /* in6_addr; network order */
4869 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
4870 * network header. output: bpf_fib_lookup sets to gateway address
4871 * if FIB lookup returns gateway route
4875 __u32 ipv6_dst[4]; /* in6_addr; network order */
4879 __be16 h_vlan_proto;
4881 __u8 smac[6]; /* ETH_ALEN */
4882 __u8 dmac[6]; /* ETH_ALEN */
4885 enum bpf_task_fd_type {
4886 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
4887 BPF_FD_TYPE_TRACEPOINT, /* tp name */
4888 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
4889 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
4890 BPF_FD_TYPE_UPROBE, /* filename + offset */
4891 BPF_FD_TYPE_URETPROBE, /* filename + offset */
4895 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
4896 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
4897 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
4900 struct bpf_flow_keys {
4903 __u16 addr_proto; /* ETH_P_* of valid addrs */
4917 __u32 ipv6_src[4]; /* in6_addr; network order */
4918 __u32 ipv6_dst[4]; /* in6_addr; network order */
4925 struct bpf_func_info {
4930 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
4931 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
4933 struct bpf_line_info {
4935 __u32 file_name_off;
4940 struct bpf_spin_lock {
4945 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
4946 * Allows 1,2,4-byte read, but no write.
4948 __u32 file_pos; /* Sysctl file position to read from, write to.
4949 * Allows 1,2,4-byte read an 4-byte write.
4953 struct bpf_sockopt {
4954 __bpf_md_ptr(struct bpf_sock *, sk);
4955 __bpf_md_ptr(void *, optval);
4956 __bpf_md_ptr(void *, optval_end);
4964 struct bpf_pidns_info {
4969 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
4970 struct bpf_sk_lookup {
4971 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
4973 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
4974 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
4975 __u32 remote_ip4; /* Network byte order */
4976 __u32 remote_ip6[4]; /* Network byte order */
4977 __u32 remote_port; /* Network byte order */
4978 __u32 local_ip4; /* Network byte order */
4979 __u32 local_ip6[4]; /* Network byte order */
4980 __u32 local_port; /* Host byte order */
4984 * struct btf_ptr is used for typed pointer representation; the
4985 * type id is used to render the pointer data as the appropriate type
4986 * via the bpf_snprintf_btf() helper described above. A flags field -
4987 * potentially to specify additional details about the BTF pointer
4988 * (rather than its mode of display) - is included for future use.
4989 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
4994 __u32 flags; /* BTF ptr flags; unused at present. */
4998 * Flags to control bpf_snprintf_btf() behaviour.
4999 * - BTF_F_COMPACT: no formatting around type information
5000 * - BTF_F_NONAME: no struct/union member names/types
5001 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
5002 * equivalent to %px.
5003 * - BTF_F_ZERO: show zero-valued struct/union members; they
5004 * are not displayed by default
5007 BTF_F_COMPACT = (1ULL << 0),
5008 BTF_F_NONAME = (1ULL << 1),
5009 BTF_F_PTR_RAW = (1ULL << 2),
5010 BTF_F_ZERO = (1ULL << 3),
5013 #endif /* _UAPI__LINUX_BPF_H__ */