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,
160 BPF_MAP_TYPE_TASK_STORAGE,
163 /* Note that tracing related programs such as
164 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
165 * are not subject to a stable API since kernel internal data
166 * structures can change from release to release and may
167 * therefore break existing tracing BPF programs. Tracing BPF
168 * programs correspond to /a/ specific kernel which is to be
169 * analyzed, and not /a/ specific kernel /and/ all future ones.
172 BPF_PROG_TYPE_UNSPEC,
173 BPF_PROG_TYPE_SOCKET_FILTER,
174 BPF_PROG_TYPE_KPROBE,
175 BPF_PROG_TYPE_SCHED_CLS,
176 BPF_PROG_TYPE_SCHED_ACT,
177 BPF_PROG_TYPE_TRACEPOINT,
179 BPF_PROG_TYPE_PERF_EVENT,
180 BPF_PROG_TYPE_CGROUP_SKB,
181 BPF_PROG_TYPE_CGROUP_SOCK,
182 BPF_PROG_TYPE_LWT_IN,
183 BPF_PROG_TYPE_LWT_OUT,
184 BPF_PROG_TYPE_LWT_XMIT,
185 BPF_PROG_TYPE_SOCK_OPS,
186 BPF_PROG_TYPE_SK_SKB,
187 BPF_PROG_TYPE_CGROUP_DEVICE,
188 BPF_PROG_TYPE_SK_MSG,
189 BPF_PROG_TYPE_RAW_TRACEPOINT,
190 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
191 BPF_PROG_TYPE_LWT_SEG6LOCAL,
192 BPF_PROG_TYPE_LIRC_MODE2,
193 BPF_PROG_TYPE_SK_REUSEPORT,
194 BPF_PROG_TYPE_FLOW_DISSECTOR,
195 BPF_PROG_TYPE_CGROUP_SYSCTL,
196 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
197 BPF_PROG_TYPE_CGROUP_SOCKOPT,
198 BPF_PROG_TYPE_TRACING,
199 BPF_PROG_TYPE_STRUCT_OPS,
202 BPF_PROG_TYPE_SK_LOOKUP,
205 enum bpf_attach_type {
206 BPF_CGROUP_INET_INGRESS,
207 BPF_CGROUP_INET_EGRESS,
208 BPF_CGROUP_INET_SOCK_CREATE,
210 BPF_SK_SKB_STREAM_PARSER,
211 BPF_SK_SKB_STREAM_VERDICT,
214 BPF_CGROUP_INET4_BIND,
215 BPF_CGROUP_INET6_BIND,
216 BPF_CGROUP_INET4_CONNECT,
217 BPF_CGROUP_INET6_CONNECT,
218 BPF_CGROUP_INET4_POST_BIND,
219 BPF_CGROUP_INET6_POST_BIND,
220 BPF_CGROUP_UDP4_SENDMSG,
221 BPF_CGROUP_UDP6_SENDMSG,
225 BPF_CGROUP_UDP4_RECVMSG,
226 BPF_CGROUP_UDP6_RECVMSG,
227 BPF_CGROUP_GETSOCKOPT,
228 BPF_CGROUP_SETSOCKOPT,
235 BPF_CGROUP_INET4_GETPEERNAME,
236 BPF_CGROUP_INET6_GETPEERNAME,
237 BPF_CGROUP_INET4_GETSOCKNAME,
238 BPF_CGROUP_INET6_GETSOCKNAME,
240 BPF_CGROUP_INET_SOCK_RELEASE,
244 __MAX_BPF_ATTACH_TYPE
247 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
250 BPF_LINK_TYPE_UNSPEC = 0,
251 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
252 BPF_LINK_TYPE_TRACING = 2,
253 BPF_LINK_TYPE_CGROUP = 3,
254 BPF_LINK_TYPE_ITER = 4,
255 BPF_LINK_TYPE_NETNS = 5,
256 BPF_LINK_TYPE_XDP = 6,
261 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
263 * NONE(default): No further bpf programs allowed in the subtree.
265 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
266 * the program in this cgroup yields to sub-cgroup program.
268 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
269 * that cgroup program gets run in addition to the program in this cgroup.
271 * Only one program is allowed to be attached to a cgroup with
272 * NONE or BPF_F_ALLOW_OVERRIDE flag.
273 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
274 * release old program and attach the new one. Attach flags has to match.
276 * Multiple programs are allowed to be attached to a cgroup with
277 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
278 * (those that were attached first, run first)
279 * The programs of sub-cgroup are executed first, then programs of
280 * this cgroup and then programs of parent cgroup.
281 * When children program makes decision (like picking TCP CA or sock bind)
282 * parent program has a chance to override it.
284 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
285 * programs for a cgroup. Though it's possible to replace an old program at
286 * any position by also specifying BPF_F_REPLACE flag and position itself in
287 * replace_bpf_fd attribute. Old program at this position will be released.
289 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
290 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
292 * cgrp1 (MULTI progs A, B) ->
293 * cgrp2 (OVERRIDE prog C) ->
294 * cgrp3 (MULTI prog D) ->
295 * cgrp4 (OVERRIDE prog E) ->
296 * cgrp5 (NONE prog F)
297 * the event in cgrp5 triggers execution of F,D,A,B in that order.
298 * if prog F is detached, the execution is E,D,A,B
299 * if prog F and D are detached, the execution is E,A,B
300 * if prog F, E and D are detached, the execution is C,A,B
302 * All eligible programs are executed regardless of return code from
305 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
306 #define BPF_F_ALLOW_MULTI (1U << 1)
307 #define BPF_F_REPLACE (1U << 2)
309 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
310 * verifier will perform strict alignment checking as if the kernel
311 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
312 * and NET_IP_ALIGN defined to 2.
314 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
316 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
317 * verifier will allow any alignment whatsoever. On platforms
318 * with strict alignment requirements for loads ands stores (such
319 * as sparc and mips) the verifier validates that all loads and
320 * stores provably follow this requirement. This flag turns that
321 * checking and enforcement off.
323 * It is mostly used for testing when we want to validate the
324 * context and memory access aspects of the verifier, but because
325 * of an unaligned access the alignment check would trigger before
326 * the one we are interested in.
328 #define BPF_F_ANY_ALIGNMENT (1U << 1)
330 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
331 * Verifier does sub-register def/use analysis and identifies instructions whose
332 * def only matters for low 32-bit, high 32-bit is never referenced later
333 * through implicit zero extension. Therefore verifier notifies JIT back-ends
334 * that it is safe to ignore clearing high 32-bit for these instructions. This
335 * saves some back-ends a lot of code-gen. However such optimization is not
336 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
337 * hence hasn't used verifier's analysis result. But, we really want to have a
338 * way to be able to verify the correctness of the described optimization on
339 * x86_64 on which testsuites are frequently exercised.
341 * So, this flag is introduced. Once it is set, verifier will randomize high
342 * 32-bit for those instructions who has been identified as safe to ignore them.
343 * Then, if verifier is not doing correct analysis, such randomization will
344 * regress tests to expose bugs.
346 #define BPF_F_TEST_RND_HI32 (1U << 2)
348 /* The verifier internal test flag. Behavior is undefined */
349 #define BPF_F_TEST_STATE_FREQ (1U << 3)
351 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
352 * restrict map and helper usage for such programs. Sleepable BPF programs can
353 * only be attached to hooks where kernel execution context allows sleeping.
354 * Such programs are allowed to use helpers that may sleep like
355 * bpf_copy_from_user().
357 #define BPF_F_SLEEPABLE (1U << 4)
359 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
360 * the following extensions:
362 * insn[0].src_reg: BPF_PSEUDO_MAP_FD
363 * insn[0].imm: map fd
367 * ldimm64 rewrite: address of map
368 * verifier type: CONST_PTR_TO_MAP
370 #define BPF_PSEUDO_MAP_FD 1
371 /* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE
372 * insn[0].imm: map fd
373 * insn[1].imm: offset into value
376 * ldimm64 rewrite: address of map[0]+offset
377 * verifier type: PTR_TO_MAP_VALUE
379 #define BPF_PSEUDO_MAP_VALUE 2
380 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID
381 * insn[0].imm: kernel btd id of VAR
385 * ldimm64 rewrite: address of the kernel variable
386 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
389 #define BPF_PSEUDO_BTF_ID 3
391 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
392 * offset to another bpf function
394 #define BPF_PSEUDO_CALL 1
396 /* flags for BPF_MAP_UPDATE_ELEM command */
398 BPF_ANY = 0, /* create new element or update existing */
399 BPF_NOEXIST = 1, /* create new element if it didn't exist */
400 BPF_EXIST = 2, /* update existing element */
401 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
404 /* flags for BPF_MAP_CREATE command */
406 BPF_F_NO_PREALLOC = (1U << 0),
407 /* Instead of having one common LRU list in the
408 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
409 * which can scale and perform better.
410 * Note, the LRU nodes (including free nodes) cannot be moved
411 * across different LRU lists.
413 BPF_F_NO_COMMON_LRU = (1U << 1),
414 /* Specify numa node during map creation */
415 BPF_F_NUMA_NODE = (1U << 2),
417 /* Flags for accessing BPF object from syscall side. */
418 BPF_F_RDONLY = (1U << 3),
419 BPF_F_WRONLY = (1U << 4),
421 /* Flag for stack_map, store build_id+offset instead of pointer */
422 BPF_F_STACK_BUILD_ID = (1U << 5),
424 /* Zero-initialize hash function seed. This should only be used for testing. */
425 BPF_F_ZERO_SEED = (1U << 6),
427 /* Flags for accessing BPF object from program side. */
428 BPF_F_RDONLY_PROG = (1U << 7),
429 BPF_F_WRONLY_PROG = (1U << 8),
431 /* Clone map from listener for newly accepted socket */
432 BPF_F_CLONE = (1U << 9),
434 /* Enable memory-mapping BPF map */
435 BPF_F_MMAPABLE = (1U << 10),
437 /* Share perf_event among processes */
438 BPF_F_PRESERVE_ELEMS = (1U << 11),
440 /* Create a map that is suitable to be an inner map with dynamic max entries */
441 BPF_F_INNER_MAP = (1U << 12),
444 /* Flags for BPF_PROG_QUERY. */
446 /* Query effective (directly attached + inherited from ancestor cgroups)
447 * programs that will be executed for events within a cgroup.
448 * attach_flags with this flag are returned only for directly attached programs.
450 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
452 /* Flags for BPF_PROG_TEST_RUN */
454 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
455 #define BPF_F_TEST_RUN_ON_CPU (1U << 0)
457 /* type for BPF_ENABLE_STATS */
458 enum bpf_stats_type {
459 /* enabled run_time_ns and run_cnt */
460 BPF_STATS_RUN_TIME = 0,
463 enum bpf_stack_build_id_status {
464 /* user space need an empty entry to identify end of a trace */
465 BPF_STACK_BUILD_ID_EMPTY = 0,
466 /* with valid build_id and offset */
467 BPF_STACK_BUILD_ID_VALID = 1,
468 /* couldn't get build_id, fallback to ip */
469 BPF_STACK_BUILD_ID_IP = 2,
472 #define BPF_BUILD_ID_SIZE 20
473 struct bpf_stack_build_id {
475 unsigned char build_id[BPF_BUILD_ID_SIZE];
482 #define BPF_OBJ_NAME_LEN 16U
485 struct { /* anonymous struct used by BPF_MAP_CREATE command */
486 __u32 map_type; /* one of enum bpf_map_type */
487 __u32 key_size; /* size of key in bytes */
488 __u32 value_size; /* size of value in bytes */
489 __u32 max_entries; /* max number of entries in a map */
490 __u32 map_flags; /* BPF_MAP_CREATE related
491 * flags defined above.
493 __u32 inner_map_fd; /* fd pointing to the inner map */
494 __u32 numa_node; /* numa node (effective only if
495 * BPF_F_NUMA_NODE is set).
497 char map_name[BPF_OBJ_NAME_LEN];
498 __u32 map_ifindex; /* ifindex of netdev to create on */
499 __u32 btf_fd; /* fd pointing to a BTF type data */
500 __u32 btf_key_type_id; /* BTF type_id of the key */
501 __u32 btf_value_type_id; /* BTF type_id of the value */
502 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
503 * struct stored as the
508 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
513 __aligned_u64 next_key;
518 struct { /* struct used by BPF_MAP_*_BATCH commands */
519 __aligned_u64 in_batch; /* start batch,
520 * NULL to start from beginning
522 __aligned_u64 out_batch; /* output: next start batch */
524 __aligned_u64 values;
525 __u32 count; /* input/output:
526 * input: # of key/value
528 * output: # of filled elements
535 struct { /* anonymous struct used by BPF_PROG_LOAD command */
536 __u32 prog_type; /* one of enum bpf_prog_type */
539 __aligned_u64 license;
540 __u32 log_level; /* verbosity level of verifier */
541 __u32 log_size; /* size of user buffer */
542 __aligned_u64 log_buf; /* user supplied buffer */
543 __u32 kern_version; /* not used */
545 char prog_name[BPF_OBJ_NAME_LEN];
546 __u32 prog_ifindex; /* ifindex of netdev to prep for */
547 /* For some prog types expected attach type must be known at
548 * load time to verify attach type specific parts of prog
549 * (context accesses, allowed helpers, etc).
551 __u32 expected_attach_type;
552 __u32 prog_btf_fd; /* fd pointing to BTF type data */
553 __u32 func_info_rec_size; /* userspace bpf_func_info size */
554 __aligned_u64 func_info; /* func info */
555 __u32 func_info_cnt; /* number of bpf_func_info records */
556 __u32 line_info_rec_size; /* userspace bpf_line_info size */
557 __aligned_u64 line_info; /* line info */
558 __u32 line_info_cnt; /* number of bpf_line_info records */
559 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
560 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
563 struct { /* anonymous struct used by BPF_OBJ_* commands */
564 __aligned_u64 pathname;
569 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
570 __u32 target_fd; /* container object to attach to */
571 __u32 attach_bpf_fd; /* eBPF program to attach */
574 __u32 replace_bpf_fd; /* previously attached eBPF
575 * program to replace if
576 * BPF_F_REPLACE is used
580 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
583 __u32 data_size_in; /* input: len of data_in */
584 __u32 data_size_out; /* input/output: len of data_out
585 * returns ENOSPC if data_out
588 __aligned_u64 data_in;
589 __aligned_u64 data_out;
592 __u32 ctx_size_in; /* input: len of ctx_in */
593 __u32 ctx_size_out; /* input/output: len of ctx_out
594 * returns ENOSPC if ctx_out
597 __aligned_u64 ctx_in;
598 __aligned_u64 ctx_out;
603 struct { /* anonymous struct used by BPF_*_GET_*_ID */
615 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
621 struct { /* anonymous struct used by BPF_PROG_QUERY command */
622 __u32 target_fd; /* container object to query */
626 __aligned_u64 prog_ids;
630 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
635 struct { /* anonymous struct for BPF_BTF_LOAD */
637 __aligned_u64 btf_log_buf;
644 __u32 pid; /* input: pid */
645 __u32 fd; /* input: fd */
646 __u32 flags; /* input: flags */
647 __u32 buf_len; /* input/output: buf len */
648 __aligned_u64 buf; /* input/output:
649 * tp_name for tracepoint
651 * filename for uprobe
653 __u32 prog_id; /* output: prod_id */
654 __u32 fd_type; /* output: BPF_FD_TYPE_* */
655 __u64 probe_offset; /* output: probe_offset */
656 __u64 probe_addr; /* output: probe_addr */
659 struct { /* struct used by BPF_LINK_CREATE command */
660 __u32 prog_fd; /* eBPF program to attach */
662 __u32 target_fd; /* object to attach to */
663 __u32 target_ifindex; /* target ifindex */
665 __u32 attach_type; /* attach type */
666 __u32 flags; /* extra flags */
668 __u32 target_btf_id; /* btf_id of target to attach to */
670 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
671 __u32 iter_info_len; /* iter_info length */
676 struct { /* struct used by BPF_LINK_UPDATE command */
677 __u32 link_fd; /* link fd */
678 /* new program fd to update link with */
680 __u32 flags; /* extra flags */
681 /* expected link's program fd; is specified only if
682 * BPF_F_REPLACE flag is set in flags */
690 struct { /* struct used by BPF_ENABLE_STATS command */
694 struct { /* struct used by BPF_ITER_CREATE command */
699 struct { /* struct used by BPF_PROG_BIND_MAP command */
702 __u32 flags; /* extra flags */
705 } __attribute__((aligned(8)));
707 /* The description below is an attempt at providing documentation to eBPF
708 * developers about the multiple available eBPF helper functions. It can be
709 * parsed and used to produce a manual page. The workflow is the following,
710 * and requires the rst2man utility:
712 * $ ./scripts/bpf_helpers_doc.py \
713 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
714 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
715 * $ man /tmp/bpf-helpers.7
717 * Note that in order to produce this external documentation, some RST
718 * formatting is used in the descriptions to get "bold" and "italics" in
719 * manual pages. Also note that the few trailing white spaces are
720 * intentional, removing them would break paragraphs for rst2man.
722 * Start of BPF helper function descriptions:
724 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
726 * Perform a lookup in *map* for an entry associated to *key*.
728 * Map value associated to *key*, or **NULL** if no entry was
731 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
733 * Add or update the value of the entry associated to *key* in
734 * *map* with *value*. *flags* is one of:
737 * The entry for *key* must not exist in the map.
739 * The entry for *key* must already exist in the map.
741 * No condition on the existence of the entry for *key*.
743 * Flag value **BPF_NOEXIST** cannot be used for maps of types
744 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
745 * elements always exist), the helper would return an error.
747 * 0 on success, or a negative error in case of failure.
749 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
751 * Delete entry with *key* from *map*.
753 * 0 on success, or a negative error in case of failure.
755 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
757 * For tracing programs, safely attempt to read *size* bytes from
758 * kernel space address *unsafe_ptr* and store the data in *dst*.
760 * Generally, use **bpf_probe_read_user**\ () or
761 * **bpf_probe_read_kernel**\ () instead.
763 * 0 on success, or a negative error in case of failure.
765 * u64 bpf_ktime_get_ns(void)
767 * Return the time elapsed since system boot, in nanoseconds.
768 * Does not include time the system was suspended.
769 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
773 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
775 * This helper is a "printk()-like" facility for debugging. It
776 * prints a message defined by format *fmt* (of size *fmt_size*)
777 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
778 * available. It can take up to three additional **u64**
779 * arguments (as an eBPF helpers, the total number of arguments is
782 * Each time the helper is called, it appends a line to the trace.
783 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
784 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
785 * The format of the trace is customizable, and the exact output
786 * one will get depends on the options set in
787 * *\/sys/kernel/debug/tracing/trace_options* (see also the
788 * *README* file under the same directory). However, it usually
789 * defaults to something like:
793 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
797 * * ``telnet`` is the name of the current task.
798 * * ``470`` is the PID of the current task.
799 * * ``001`` is the CPU number on which the task is
801 * * In ``.N..``, each character refers to a set of
802 * options (whether irqs are enabled, scheduling
803 * options, whether hard/softirqs are running, level of
804 * preempt_disabled respectively). **N** means that
805 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
807 * * ``419421.045894`` is a timestamp.
808 * * ``0x00000001`` is a fake value used by BPF for the
809 * instruction pointer register.
810 * * ``<formatted msg>`` is the message formatted with
813 * The conversion specifiers supported by *fmt* are similar, but
814 * more limited than for printk(). They are **%d**, **%i**,
815 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
816 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
817 * of field, padding with zeroes, etc.) is available, and the
818 * helper will return **-EINVAL** (but print nothing) if it
819 * encounters an unknown specifier.
821 * Also, note that **bpf_trace_printk**\ () is slow, and should
822 * only be used for debugging purposes. For this reason, a notice
823 * block (spanning several lines) is printed to kernel logs and
824 * states that the helper should not be used "for production use"
825 * the first time this helper is used (or more precisely, when
826 * **trace_printk**\ () buffers are allocated). For passing values
827 * to user space, perf events should be preferred.
829 * The number of bytes written to the buffer, or a negative error
830 * in case of failure.
832 * u32 bpf_get_prandom_u32(void)
834 * Get a pseudo-random number.
836 * From a security point of view, this helper uses its own
837 * pseudo-random internal state, and cannot be used to infer the
838 * seed of other random functions in the kernel. However, it is
839 * essential to note that the generator used by the helper is not
840 * cryptographically secure.
842 * A random 32-bit unsigned value.
844 * u32 bpf_get_smp_processor_id(void)
846 * Get the SMP (symmetric multiprocessing) processor id. Note that
847 * all programs run with preemption disabled, which means that the
848 * SMP processor id is stable during all the execution of the
851 * The SMP id of the processor running the program.
853 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
855 * Store *len* bytes from address *from* into the packet
856 * associated to *skb*, at *offset*. *flags* are a combination of
857 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
858 * checksum for the packet after storing the bytes) and
859 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
860 * **->swhash** and *skb*\ **->l4hash** to 0).
862 * A call to this helper is susceptible to change the underlying
863 * packet buffer. Therefore, at load time, all checks on pointers
864 * previously done by the verifier are invalidated and must be
865 * performed again, if the helper is used in combination with
866 * direct packet access.
868 * 0 on success, or a negative error in case of failure.
870 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
872 * Recompute the layer 3 (e.g. IP) checksum for the packet
873 * associated to *skb*. Computation is incremental, so the helper
874 * must know the former value of the header field that was
875 * modified (*from*), the new value of this field (*to*), and the
876 * number of bytes (2 or 4) for this field, stored in *size*.
877 * Alternatively, it is possible to store the difference between
878 * the previous and the new values of the header field in *to*, by
879 * setting *from* and *size* to 0. For both methods, *offset*
880 * indicates the location of the IP checksum within the packet.
882 * This helper works in combination with **bpf_csum_diff**\ (),
883 * which does not update the checksum in-place, but offers more
884 * flexibility and can handle sizes larger than 2 or 4 for the
885 * checksum to update.
887 * A call to this helper is susceptible to change the underlying
888 * packet buffer. Therefore, at load time, all checks on pointers
889 * previously done by the verifier are invalidated and must be
890 * performed again, if the helper is used in combination with
891 * direct packet access.
893 * 0 on success, or a negative error in case of failure.
895 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
897 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
898 * packet associated to *skb*. Computation is incremental, so the
899 * helper must know the former value of the header field that was
900 * modified (*from*), the new value of this field (*to*), and the
901 * number of bytes (2 or 4) for this field, stored on the lowest
902 * four bits of *flags*. Alternatively, it is possible to store
903 * the difference between the previous and the new values of the
904 * header field in *to*, by setting *from* and the four lowest
905 * bits of *flags* to 0. For both methods, *offset* indicates the
906 * location of the IP checksum within the packet. In addition to
907 * the size of the field, *flags* can be added (bitwise OR) actual
908 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
909 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
910 * for updates resulting in a null checksum the value is set to
911 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
912 * the checksum is to be computed against a pseudo-header.
914 * This helper works in combination with **bpf_csum_diff**\ (),
915 * which does not update the checksum in-place, but offers more
916 * flexibility and can handle sizes larger than 2 or 4 for the
917 * checksum to update.
919 * A call to this helper is susceptible to change the underlying
920 * packet buffer. Therefore, at load time, all checks on pointers
921 * previously done by the verifier are invalidated and must be
922 * performed again, if the helper is used in combination with
923 * direct packet access.
925 * 0 on success, or a negative error in case of failure.
927 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
929 * This special helper is used to trigger a "tail call", or in
930 * other words, to jump into another eBPF program. The same stack
931 * frame is used (but values on stack and in registers for the
932 * caller are not accessible to the callee). This mechanism allows
933 * for program chaining, either for raising the maximum number of
934 * available eBPF instructions, or to execute given programs in
935 * conditional blocks. For security reasons, there is an upper
936 * limit to the number of successive tail calls that can be
939 * Upon call of this helper, the program attempts to jump into a
940 * program referenced at index *index* in *prog_array_map*, a
941 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
942 * *ctx*, a pointer to the context.
944 * If the call succeeds, the kernel immediately runs the first
945 * instruction of the new program. This is not a function call,
946 * and it never returns to the previous program. If the call
947 * fails, then the helper has no effect, and the caller continues
948 * to run its subsequent instructions. A call can fail if the
949 * destination program for the jump does not exist (i.e. *index*
950 * is superior to the number of entries in *prog_array_map*), or
951 * if the maximum number of tail calls has been reached for this
952 * chain of programs. This limit is defined in the kernel by the
953 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
954 * which is currently set to 32.
956 * 0 on success, or a negative error in case of failure.
958 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
960 * Clone and redirect the packet associated to *skb* to another
961 * net device of index *ifindex*. Both ingress and egress
962 * interfaces can be used for redirection. The **BPF_F_INGRESS**
963 * value in *flags* is used to make the distinction (ingress path
964 * is selected if the flag is present, egress path otherwise).
965 * This is the only flag supported for now.
967 * In comparison with **bpf_redirect**\ () helper,
968 * **bpf_clone_redirect**\ () has the associated cost of
969 * duplicating the packet buffer, but this can be executed out of
970 * the eBPF program. Conversely, **bpf_redirect**\ () is more
971 * efficient, but it is handled through an action code where the
972 * redirection happens only after the eBPF program has returned.
974 * A call to this helper is susceptible to change the underlying
975 * packet buffer. Therefore, at load time, all checks on pointers
976 * previously done by the verifier are invalidated and must be
977 * performed again, if the helper is used in combination with
978 * direct packet access.
980 * 0 on success, or a negative error in case of failure.
982 * u64 bpf_get_current_pid_tgid(void)
984 * A 64-bit integer containing the current tgid and pid, and
986 * *current_task*\ **->tgid << 32 \|**
987 * *current_task*\ **->pid**.
989 * u64 bpf_get_current_uid_gid(void)
991 * A 64-bit integer containing the current GID and UID, and
992 * created as such: *current_gid* **<< 32 \|** *current_uid*.
994 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
996 * Copy the **comm** attribute of the current task into *buf* of
997 * *size_of_buf*. The **comm** attribute contains the name of
998 * the executable (excluding the path) for the current task. The
999 * *size_of_buf* must be strictly positive. On success, the
1000 * helper makes sure that the *buf* is NUL-terminated. On failure,
1001 * it is filled with zeroes.
1003 * 0 on success, or a negative error in case of failure.
1005 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1007 * Retrieve the classid for the current task, i.e. for the net_cls
1008 * cgroup to which *skb* belongs.
1010 * This helper can be used on TC egress path, but not on ingress.
1012 * The net_cls cgroup provides an interface to tag network packets
1013 * based on a user-provided identifier for all traffic coming from
1014 * the tasks belonging to the related cgroup. See also the related
1015 * kernel documentation, available from the Linux sources in file
1016 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1018 * The Linux kernel has two versions for cgroups: there are
1019 * cgroups v1 and cgroups v2. Both are available to users, who can
1020 * use a mixture of them, but note that the net_cls cgroup is for
1021 * cgroup v1 only. This makes it incompatible with BPF programs
1022 * run on cgroups, which is a cgroup-v2-only feature (a socket can
1023 * only hold data for one version of cgroups at a time).
1025 * This helper is only available is the kernel was compiled with
1026 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1027 * "**y**" or to "**m**".
1029 * The classid, or 0 for the default unconfigured classid.
1031 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1033 * Push a *vlan_tci* (VLAN tag control information) of protocol
1034 * *vlan_proto* to the packet associated to *skb*, then update
1035 * the checksum. Note that if *vlan_proto* is different from
1036 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1037 * be **ETH_P_8021Q**.
1039 * A call to this helper is susceptible to change the underlying
1040 * packet buffer. Therefore, at load time, all checks on pointers
1041 * previously done by the verifier are invalidated and must be
1042 * performed again, if the helper is used in combination with
1043 * direct packet access.
1045 * 0 on success, or a negative error in case of failure.
1047 * long bpf_skb_vlan_pop(struct sk_buff *skb)
1049 * Pop a VLAN header from the packet associated to *skb*.
1051 * A call to this helper is susceptible to change the underlying
1052 * packet buffer. Therefore, at load time, all checks on pointers
1053 * previously done by the verifier are invalidated and must be
1054 * performed again, if the helper is used in combination with
1055 * direct packet access.
1057 * 0 on success, or a negative error in case of failure.
1059 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1061 * Get tunnel metadata. This helper takes a pointer *key* to an
1062 * empty **struct bpf_tunnel_key** of **size**, that will be
1063 * filled with tunnel metadata for the packet associated to *skb*.
1064 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1065 * indicates that the tunnel is based on IPv6 protocol instead of
1068 * The **struct bpf_tunnel_key** is an object that generalizes the
1069 * principal parameters used by various tunneling protocols into a
1070 * single struct. This way, it can be used to easily make a
1071 * decision based on the contents of the encapsulation header,
1072 * "summarized" in this struct. In particular, it holds the IP
1073 * address of the remote end (IPv4 or IPv6, depending on the case)
1074 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1075 * this struct exposes the *key*\ **->tunnel_id**, which is
1076 * generally mapped to a VNI (Virtual Network Identifier), making
1077 * it programmable together with the **bpf_skb_set_tunnel_key**\
1080 * Let's imagine that the following code is part of a program
1081 * attached to the TC ingress interface, on one end of a GRE
1082 * tunnel, and is supposed to filter out all messages coming from
1083 * remote ends with IPv4 address other than 10.0.0.1:
1088 * struct bpf_tunnel_key key = {};
1090 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1092 * return TC_ACT_SHOT; // drop packet
1094 * if (key.remote_ipv4 != 0x0a000001)
1095 * return TC_ACT_SHOT; // drop packet
1097 * return TC_ACT_OK; // accept packet
1099 * This interface can also be used with all encapsulation devices
1100 * that can operate in "collect metadata" mode: instead of having
1101 * one network device per specific configuration, the "collect
1102 * metadata" mode only requires a single device where the
1103 * configuration can be extracted from this helper.
1105 * This can be used together with various tunnels such as VXLan,
1106 * Geneve, GRE or IP in IP (IPIP).
1108 * 0 on success, or a negative error in case of failure.
1110 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1112 * Populate tunnel metadata for packet associated to *skb.* The
1113 * tunnel metadata is set to the contents of *key*, of *size*. The
1114 * *flags* can be set to a combination of the following values:
1116 * **BPF_F_TUNINFO_IPV6**
1117 * Indicate that the tunnel is based on IPv6 protocol
1119 * **BPF_F_ZERO_CSUM_TX**
1120 * For IPv4 packets, add a flag to tunnel metadata
1121 * indicating that checksum computation should be skipped
1122 * and checksum set to zeroes.
1123 * **BPF_F_DONT_FRAGMENT**
1124 * Add a flag to tunnel metadata indicating that the
1125 * packet should not be fragmented.
1126 * **BPF_F_SEQ_NUMBER**
1127 * Add a flag to tunnel metadata indicating that a
1128 * sequence number should be added to tunnel header before
1129 * sending the packet. This flag was added for GRE
1130 * encapsulation, but might be used with other protocols
1131 * as well in the future.
1133 * Here is a typical usage on the transmit path:
1137 * struct bpf_tunnel_key key;
1139 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1140 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1142 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1143 * helper for additional information.
1145 * 0 on success, or a negative error in case of failure.
1147 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1149 * Read the value of a perf event counter. This helper relies on a
1150 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1151 * the perf event counter is selected when *map* is updated with
1152 * perf event file descriptors. The *map* is an array whose size
1153 * is the number of available CPUs, and each cell contains a value
1154 * relative to one CPU. The value to retrieve is indicated by
1155 * *flags*, that contains the index of the CPU to look up, masked
1156 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1157 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1158 * current CPU should be retrieved.
1160 * Note that before Linux 4.13, only hardware perf event can be
1163 * Also, be aware that the newer helper
1164 * **bpf_perf_event_read_value**\ () is recommended over
1165 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1166 * quirks where error and counter value are used as a return code
1167 * (which is wrong to do since ranges may overlap). This issue is
1168 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1169 * time provides more features over the **bpf_perf_event_read**\
1170 * () interface. Please refer to the description of
1171 * **bpf_perf_event_read_value**\ () for details.
1173 * The value of the perf event counter read from the map, or a
1174 * negative error code in case of failure.
1176 * long bpf_redirect(u32 ifindex, u64 flags)
1178 * Redirect the packet to another net device of index *ifindex*.
1179 * This helper is somewhat similar to **bpf_clone_redirect**\
1180 * (), except that the packet is not cloned, which provides
1181 * increased performance.
1183 * Except for XDP, both ingress and egress interfaces can be used
1184 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1185 * to make the distinction (ingress path is selected if the flag
1186 * is present, egress path otherwise). Currently, XDP only
1187 * supports redirection to the egress interface, and accepts no
1190 * The same effect can also be attained with the more generic
1191 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1192 * redirect target instead of providing it directly to the helper.
1194 * For XDP, the helper returns **XDP_REDIRECT** on success or
1195 * **XDP_ABORTED** on error. For other program types, the values
1196 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1199 * u32 bpf_get_route_realm(struct sk_buff *skb)
1201 * Retrieve the realm or the route, that is to say the
1202 * **tclassid** field of the destination for the *skb*. The
1203 * identifier retrieved is a user-provided tag, similar to the
1204 * one used with the net_cls cgroup (see description for
1205 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1206 * held by a route (a destination entry), not by a task.
1208 * Retrieving this identifier works with the clsact TC egress hook
1209 * (see also **tc-bpf(8)**), or alternatively on conventional
1210 * classful egress qdiscs, but not on TC ingress path. In case of
1211 * clsact TC egress hook, this has the advantage that, internally,
1212 * the destination entry has not been dropped yet in the transmit
1213 * path. Therefore, the destination entry does not need to be
1214 * artificially held via **netif_keep_dst**\ () for a classful
1215 * qdisc until the *skb* is freed.
1217 * This helper is available only if the kernel was compiled with
1218 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1220 * The realm of the route for the packet associated to *skb*, or 0
1221 * if none was found.
1223 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1225 * Write raw *data* blob into a special BPF perf event held by
1226 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1227 * event must have the following attributes: **PERF_SAMPLE_RAW**
1228 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1229 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1231 * The *flags* are used to indicate the index in *map* for which
1232 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1233 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1234 * to indicate that the index of the current CPU core should be
1237 * The value to write, of *size*, is passed through eBPF stack and
1238 * pointed by *data*.
1240 * The context of the program *ctx* needs also be passed to the
1243 * On user space, a program willing to read the values needs to
1244 * call **perf_event_open**\ () on the perf event (either for
1245 * one or for all CPUs) and to store the file descriptor into the
1246 * *map*. This must be done before the eBPF program can send data
1247 * into it. An example is available in file
1248 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1249 * tree (the eBPF program counterpart is in
1250 * *samples/bpf/trace_output_kern.c*).
1252 * **bpf_perf_event_output**\ () achieves better performance
1253 * than **bpf_trace_printk**\ () for sharing data with user
1254 * space, and is much better suitable for streaming data from eBPF
1257 * Note that this helper is not restricted to tracing use cases
1258 * and can be used with programs attached to TC or XDP as well,
1259 * where it allows for passing data to user space listeners. Data
1262 * * Only custom structs,
1263 * * Only the packet payload, or
1264 * * A combination of both.
1266 * 0 on success, or a negative error in case of failure.
1268 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1270 * This helper was provided as an easy way to load data from a
1271 * packet. It can be used to load *len* bytes from *offset* from
1272 * the packet associated to *skb*, into the buffer pointed by
1275 * Since Linux 4.7, usage of this helper has mostly been replaced
1276 * by "direct packet access", enabling packet data to be
1277 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1278 * pointing respectively to the first byte of packet data and to
1279 * the byte after the last byte of packet data. However, it
1280 * remains useful if one wishes to read large quantities of data
1281 * at once from a packet into the eBPF stack.
1283 * 0 on success, or a negative error in case of failure.
1285 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1287 * Walk a user or a kernel stack and return its id. To achieve
1288 * this, the helper needs *ctx*, which is a pointer to the context
1289 * on which the tracing program is executed, and a pointer to a
1290 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1292 * The last argument, *flags*, holds the number of stack frames to
1293 * skip (from 0 to 255), masked with
1294 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1295 * a combination of the following flags:
1297 * **BPF_F_USER_STACK**
1298 * Collect a user space stack instead of a kernel stack.
1299 * **BPF_F_FAST_STACK_CMP**
1300 * Compare stacks by hash only.
1301 * **BPF_F_REUSE_STACKID**
1302 * If two different stacks hash into the same *stackid*,
1303 * discard the old one.
1305 * The stack id retrieved is a 32 bit long integer handle which
1306 * can be further combined with other data (including other stack
1307 * ids) and used as a key into maps. This can be useful for
1308 * generating a variety of graphs (such as flame graphs or off-cpu
1311 * For walking a stack, this helper is an improvement over
1312 * **bpf_probe_read**\ (), which can be used with unrolled loops
1313 * but is not efficient and consumes a lot of eBPF instructions.
1314 * Instead, **bpf_get_stackid**\ () can collect up to
1315 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1316 * this limit can be controlled with the **sysctl** program, and
1317 * that it should be manually increased in order to profile long
1318 * user stacks (such as stacks for Java programs). To do so, use:
1322 * # sysctl kernel.perf_event_max_stack=<new value>
1324 * The positive or null stack id on success, or a negative error
1325 * in case of failure.
1327 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1329 * Compute a checksum difference, from the raw buffer pointed by
1330 * *from*, of length *from_size* (that must be a multiple of 4),
1331 * towards the raw buffer pointed by *to*, of size *to_size*
1332 * (same remark). An optional *seed* can be added to the value
1333 * (this can be cascaded, the seed may come from a previous call
1336 * This is flexible enough to be used in several ways:
1338 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1339 * checksum, it can be used when pushing new data.
1340 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1341 * checksum, it can be used when removing data from a packet.
1342 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1343 * can be used to compute a diff. Note that *from_size* and
1344 * *to_size* do not need to be equal.
1346 * This helper can be used in combination with
1347 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1348 * which one can feed in the difference computed with
1349 * **bpf_csum_diff**\ ().
1351 * The checksum result, or a negative error code in case of
1354 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1356 * Retrieve tunnel options metadata for the packet associated to
1357 * *skb*, and store the raw tunnel option data to the buffer *opt*
1360 * This helper can be used with encapsulation devices that can
1361 * operate in "collect metadata" mode (please refer to the related
1362 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1363 * more details). A particular example where this can be used is
1364 * in combination with the Geneve encapsulation protocol, where it
1365 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1366 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1367 * the eBPF program. This allows for full customization of these
1370 * The size of the option data retrieved.
1372 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1374 * Set tunnel options metadata for the packet associated to *skb*
1375 * to the option data contained in the raw buffer *opt* of *size*.
1377 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1378 * helper for additional information.
1380 * 0 on success, or a negative error in case of failure.
1382 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1384 * Change the protocol of the *skb* to *proto*. Currently
1385 * supported are transition from IPv4 to IPv6, and from IPv6 to
1386 * IPv4. The helper takes care of the groundwork for the
1387 * transition, including resizing the socket buffer. The eBPF
1388 * program is expected to fill the new headers, if any, via
1389 * **skb_store_bytes**\ () and to recompute the checksums with
1390 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1391 * (). The main case for this helper is to perform NAT64
1392 * operations out of an eBPF program.
1394 * Internally, the GSO type is marked as dodgy so that headers are
1395 * checked and segments are recalculated by the GSO/GRO engine.
1396 * The size for GSO target is adapted as well.
1398 * All values for *flags* are reserved for future usage, and must
1401 * A call to this helper is susceptible to change the underlying
1402 * packet buffer. Therefore, at load time, all checks on pointers
1403 * previously done by the verifier are invalidated and must be
1404 * performed again, if the helper is used in combination with
1405 * direct packet access.
1407 * 0 on success, or a negative error in case of failure.
1409 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
1411 * Change the packet type for the packet associated to *skb*. This
1412 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1413 * the eBPF program does not have a write access to *skb*\
1414 * **->pkt_type** beside this helper. Using a helper here allows
1415 * for graceful handling of errors.
1417 * The major use case is to change incoming *skb*s to
1418 * **PACKET_HOST** in a programmatic way instead of having to
1419 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1422 * Note that *type* only allows certain values. At this time, they
1427 * **PACKET_BROADCAST**
1428 * Send packet to all.
1429 * **PACKET_MULTICAST**
1430 * Send packet to group.
1431 * **PACKET_OTHERHOST**
1432 * Send packet to someone else.
1434 * 0 on success, or a negative error in case of failure.
1436 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1438 * Check whether *skb* is a descendant of the cgroup2 held by
1439 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1441 * The return value depends on the result of the test, and can be:
1443 * * 0, if the *skb* failed the cgroup2 descendant test.
1444 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1445 * * A negative error code, if an error occurred.
1447 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1449 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1450 * not set, in particular if the hash was cleared due to mangling,
1451 * recompute this hash. Later accesses to the hash can be done
1452 * directly with *skb*\ **->hash**.
1454 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1455 * prototype with **bpf_skb_change_proto**\ (), or calling
1456 * **bpf_skb_store_bytes**\ () with the
1457 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1458 * the hash and to trigger a new computation for the next call to
1459 * **bpf_get_hash_recalc**\ ().
1463 * u64 bpf_get_current_task(void)
1465 * A pointer to the current task struct.
1467 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
1469 * Attempt in a safe way to write *len* bytes from the buffer
1470 * *src* to *dst* in memory. It only works for threads that are in
1471 * user context, and *dst* must be a valid user space address.
1473 * This helper should not be used to implement any kind of
1474 * security mechanism because of TOC-TOU attacks, but rather to
1475 * debug, divert, and manipulate execution of semi-cooperative
1478 * Keep in mind that this feature is meant for experiments, and it
1479 * has a risk of crashing the system and running programs.
1480 * Therefore, when an eBPF program using this helper is attached,
1481 * a warning including PID and process name is printed to kernel
1484 * 0 on success, or a negative error in case of failure.
1486 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1488 * Check whether the probe is being run is the context of a given
1489 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1490 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1492 * The return value depends on the result of the test, and can be:
1494 * * 0, if current task belongs to the cgroup2.
1495 * * 1, if current task does not belong to the cgroup2.
1496 * * A negative error code, if an error occurred.
1498 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1500 * Resize (trim or grow) the packet associated to *skb* to the
1501 * new *len*. The *flags* are reserved for future usage, and must
1504 * The basic idea is that the helper performs the needed work to
1505 * change the size of the packet, then the eBPF program rewrites
1506 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1507 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1508 * and others. This helper is a slow path utility intended for
1509 * replies with control messages. And because it is targeted for
1510 * slow path, the helper itself can afford to be slow: it
1511 * implicitly linearizes, unclones and drops offloads from the
1514 * A call to this helper is susceptible to change the underlying
1515 * packet buffer. Therefore, at load time, all checks on pointers
1516 * previously done by the verifier are invalidated and must be
1517 * performed again, if the helper is used in combination with
1518 * direct packet access.
1520 * 0 on success, or a negative error in case of failure.
1522 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1524 * Pull in non-linear data in case the *skb* is non-linear and not
1525 * all of *len* are part of the linear section. Make *len* bytes
1526 * from *skb* readable and writable. If a zero value is passed for
1527 * *len*, then the whole length of the *skb* is pulled.
1529 * This helper is only needed for reading and writing with direct
1532 * For direct packet access, testing that offsets to access
1533 * are within packet boundaries (test on *skb*\ **->data_end**) is
1534 * susceptible to fail if offsets are invalid, or if the requested
1535 * data is in non-linear parts of the *skb*. On failure the
1536 * program can just bail out, or in the case of a non-linear
1537 * buffer, use a helper to make the data available. The
1538 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1539 * the data. Another one consists in using **bpf_skb_pull_data**
1540 * to pull in once the non-linear parts, then retesting and
1541 * eventually access the data.
1543 * At the same time, this also makes sure the *skb* is uncloned,
1544 * which is a necessary condition for direct write. As this needs
1545 * to be an invariant for the write part only, the verifier
1546 * detects writes and adds a prologue that is calling
1547 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1548 * the very beginning in case it is indeed cloned.
1550 * A call to this helper is susceptible to change the underlying
1551 * packet buffer. Therefore, at load time, all checks on pointers
1552 * previously done by the verifier are invalidated and must be
1553 * performed again, if the helper is used in combination with
1554 * direct packet access.
1556 * 0 on success, or a negative error in case of failure.
1558 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1560 * Add the checksum *csum* into *skb*\ **->csum** in case the
1561 * driver has supplied a checksum for the entire packet into that
1562 * field. Return an error otherwise. This helper is intended to be
1563 * used in combination with **bpf_csum_diff**\ (), in particular
1564 * when the checksum needs to be updated after data has been
1565 * written into the packet through direct packet access.
1567 * The checksum on success, or a negative error code in case of
1570 * void bpf_set_hash_invalid(struct sk_buff *skb)
1572 * Invalidate the current *skb*\ **->hash**. It can be used after
1573 * mangling on headers through direct packet access, in order to
1574 * indicate that the hash is outdated and to trigger a
1575 * recalculation the next time the kernel tries to access this
1576 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1578 * long bpf_get_numa_node_id(void)
1580 * Return the id of the current NUMA node. The primary use case
1581 * for this helper is the selection of sockets for the local NUMA
1582 * node, when the program is attached to sockets using the
1583 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1584 * but the helper is also available to other eBPF program types,
1585 * similarly to **bpf_get_smp_processor_id**\ ().
1587 * The id of current NUMA node.
1589 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1591 * Grows headroom of packet associated to *skb* and adjusts the
1592 * offset of the MAC header accordingly, adding *len* bytes of
1593 * space. It automatically extends and reallocates memory as
1596 * This helper can be used on a layer 3 *skb* to push a MAC header
1597 * for redirection into a layer 2 device.
1599 * All values for *flags* are reserved for future usage, and must
1602 * A call to this helper is susceptible to change the underlying
1603 * packet buffer. Therefore, at load time, all checks on pointers
1604 * previously done by the verifier are invalidated and must be
1605 * performed again, if the helper is used in combination with
1606 * direct packet access.
1608 * 0 on success, or a negative error in case of failure.
1610 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1612 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1613 * it is possible to use a negative value for *delta*. This helper
1614 * can be used to prepare the packet for pushing or popping
1617 * A call to this helper is susceptible to change the underlying
1618 * packet buffer. Therefore, at load time, all checks on pointers
1619 * previously done by the verifier are invalidated and must be
1620 * performed again, if the helper is used in combination with
1621 * direct packet access.
1623 * 0 on success, or a negative error in case of failure.
1625 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1627 * Copy a NUL terminated string from an unsafe kernel address
1628 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
1631 * Generally, use **bpf_probe_read_user_str**\ () or
1632 * **bpf_probe_read_kernel_str**\ () instead.
1634 * On success, the strictly positive length of the string,
1635 * including the trailing NUL character. On error, a negative
1638 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1640 * If the **struct sk_buff** pointed by *skb* has a known socket,
1641 * retrieve the cookie (generated by the kernel) of this socket.
1642 * If no cookie has been set yet, generate a new cookie. Once
1643 * generated, the socket cookie remains stable for the life of the
1644 * socket. This helper can be useful for monitoring per socket
1645 * networking traffic statistics as it provides a global socket
1646 * identifier that can be assumed unique.
1648 * A 8-byte long non-decreasing number on success, or 0 if the
1649 * socket field is missing inside *skb*.
1651 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1653 * Equivalent to bpf_get_socket_cookie() helper that accepts
1654 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1656 * A 8-byte long non-decreasing number.
1658 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1660 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
1661 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1663 * A 8-byte long non-decreasing number.
1665 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1667 * The owner UID of the socket associated to *skb*. If the socket
1668 * is **NULL**, or if it is not a full socket (i.e. if it is a
1669 * time-wait or a request socket instead), **overflowuid** value
1670 * is returned (note that **overflowuid** might also be the actual
1671 * UID value for the socket).
1673 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
1675 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1680 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1682 * Emulate a call to **setsockopt()** on the socket associated to
1683 * *bpf_socket*, which must be a full socket. The *level* at
1684 * which the option resides and the name *optname* of the option
1685 * must be specified, see **setsockopt(2)** for more information.
1686 * The option value of length *optlen* is pointed by *optval*.
1688 * *bpf_socket* should be one of the following:
1690 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1691 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1692 * and **BPF_CGROUP_INET6_CONNECT**.
1694 * This helper actually implements a subset of **setsockopt()**.
1695 * It supports the following *level*\ s:
1697 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1698 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1699 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
1700 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
1701 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1702 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1703 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
1704 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
1705 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**.
1706 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1707 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1709 * 0 on success, or a negative error in case of failure.
1711 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1713 * Grow or shrink the room for data in the packet associated to
1714 * *skb* by *len_diff*, and according to the selected *mode*.
1716 * By default, the helper will reset any offloaded checksum
1717 * indicator of the skb to CHECKSUM_NONE. This can be avoided
1718 * by the following flag:
1720 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
1721 * checksum data of the skb to CHECKSUM_NONE.
1723 * There are two supported modes at this time:
1725 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1726 * (room space is added or removed below the layer 2 header).
1728 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1729 * (room space is added or removed below the layer 3 header).
1731 * The following flags are supported at this time:
1733 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1734 * Adjusting mss in this way is not allowed for datagrams.
1736 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1737 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1738 * Any new space is reserved to hold a tunnel header.
1739 * Configure skb offsets and other fields accordingly.
1741 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1742 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1743 * Use with ENCAP_L3 flags to further specify the tunnel type.
1745 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1746 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1747 * type; *len* is the length of the inner MAC header.
1749 * A call to this helper is susceptible to change the underlying
1750 * packet buffer. Therefore, at load time, all checks on pointers
1751 * previously done by the verifier are invalidated and must be
1752 * performed again, if the helper is used in combination with
1753 * direct packet access.
1755 * 0 on success, or a negative error in case of failure.
1757 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1759 * Redirect the packet to the endpoint referenced by *map* at
1760 * index *key*. Depending on its type, this *map* can contain
1761 * references to net devices (for forwarding packets through other
1762 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1763 * but this is only implemented for native XDP (with driver
1764 * support) as of this writing).
1766 * The lower two bits of *flags* are used as the return code if
1767 * the map lookup fails. This is so that the return value can be
1768 * one of the XDP program return codes up to **XDP_TX**, as chosen
1769 * by the caller. Any higher bits in the *flags* argument must be
1772 * See also **bpf_redirect**\ (), which only supports redirecting
1773 * to an ifindex, but doesn't require a map to do so.
1775 * **XDP_REDIRECT** on success, or the value of the two lower bits
1776 * of the *flags* argument on error.
1778 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1780 * Redirect the packet to the socket referenced by *map* (of type
1781 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1782 * egress interfaces can be used for redirection. The
1783 * **BPF_F_INGRESS** value in *flags* is used to make the
1784 * distinction (ingress path is selected if the flag is present,
1785 * egress path otherwise). This is the only flag supported for now.
1787 * **SK_PASS** on success, or **SK_DROP** on error.
1789 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1791 * Add an entry to, or update a *map* referencing sockets. The
1792 * *skops* is used as a new value for the entry associated to
1793 * *key*. *flags* is one of:
1796 * The entry for *key* must not exist in the map.
1798 * The entry for *key* must already exist in the map.
1800 * No condition on the existence of the entry for *key*.
1802 * If the *map* has eBPF programs (parser and verdict), those will
1803 * be inherited by the socket being added. If the socket is
1804 * already attached to eBPF programs, this results in an error.
1806 * 0 on success, or a negative error in case of failure.
1808 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1810 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1811 * *delta* (which can be positive or negative). Note that this
1812 * operation modifies the address stored in *xdp_md*\ **->data**,
1813 * so the latter must be loaded only after the helper has been
1816 * The use of *xdp_md*\ **->data_meta** is optional and programs
1817 * are not required to use it. The rationale is that when the
1818 * packet is processed with XDP (e.g. as DoS filter), it is
1819 * possible to push further meta data along with it before passing
1820 * to the stack, and to give the guarantee that an ingress eBPF
1821 * program attached as a TC classifier on the same device can pick
1822 * this up for further post-processing. Since TC works with socket
1823 * buffers, it remains possible to set from XDP the **mark** or
1824 * **priority** pointers, or other pointers for the socket buffer.
1825 * Having this scratch space generic and programmable allows for
1826 * more flexibility as the user is free to store whatever meta
1829 * A call to this helper is susceptible to change the underlying
1830 * packet buffer. Therefore, at load time, all checks on pointers
1831 * previously done by the verifier are invalidated and must be
1832 * performed again, if the helper is used in combination with
1833 * direct packet access.
1835 * 0 on success, or a negative error in case of failure.
1837 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1839 * Read the value of a perf event counter, and store it into *buf*
1840 * of size *buf_size*. This helper relies on a *map* of type
1841 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1842 * counter is selected when *map* is updated with perf event file
1843 * descriptors. The *map* is an array whose size is the number of
1844 * available CPUs, and each cell contains a value relative to one
1845 * CPU. The value to retrieve is indicated by *flags*, that
1846 * contains the index of the CPU to look up, masked with
1847 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1848 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1849 * current CPU should be retrieved.
1851 * This helper behaves in a way close to
1852 * **bpf_perf_event_read**\ () helper, save that instead of
1853 * just returning the value observed, it fills the *buf*
1854 * structure. This allows for additional data to be retrieved: in
1855 * particular, the enabled and running times (in *buf*\
1856 * **->enabled** and *buf*\ **->running**, respectively) are
1857 * copied. In general, **bpf_perf_event_read_value**\ () is
1858 * recommended over **bpf_perf_event_read**\ (), which has some
1859 * ABI issues and provides fewer functionalities.
1861 * These values are interesting, because hardware PMU (Performance
1862 * Monitoring Unit) counters are limited resources. When there are
1863 * more PMU based perf events opened than available counters,
1864 * kernel will multiplex these events so each event gets certain
1865 * percentage (but not all) of the PMU time. In case that
1866 * multiplexing happens, the number of samples or counter value
1867 * will not reflect the case compared to when no multiplexing
1868 * occurs. This makes comparison between different runs difficult.
1869 * Typically, the counter value should be normalized before
1870 * comparing to other experiments. The usual normalization is done
1875 * normalized_counter = counter * t_enabled / t_running
1877 * Where t_enabled is the time enabled for event and t_running is
1878 * the time running for event since last normalization. The
1879 * enabled and running times are accumulated since the perf event
1880 * open. To achieve scaling factor between two invocations of an
1881 * eBPF program, users can use CPU id as the key (which is
1882 * typical for perf array usage model) to remember the previous
1883 * value and do the calculation inside the eBPF program.
1885 * 0 on success, or a negative error in case of failure.
1887 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1889 * For en eBPF program attached to a perf event, retrieve the
1890 * value of the event counter associated to *ctx* and store it in
1891 * the structure pointed by *buf* and of size *buf_size*. Enabled
1892 * and running times are also stored in the structure (see
1893 * description of helper **bpf_perf_event_read_value**\ () for
1896 * 0 on success, or a negative error in case of failure.
1898 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1900 * Emulate a call to **getsockopt()** on the socket associated to
1901 * *bpf_socket*, which must be a full socket. The *level* at
1902 * which the option resides and the name *optname* of the option
1903 * must be specified, see **getsockopt(2)** for more information.
1904 * The retrieved value is stored in the structure pointed by
1905 * *opval* and of length *optlen*.
1907 * *bpf_socket* should be one of the following:
1909 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1910 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1911 * and **BPF_CGROUP_INET6_CONNECT**.
1913 * This helper actually implements a subset of **getsockopt()**.
1914 * It supports the following *level*\ s:
1916 * * **IPPROTO_TCP**, which supports *optname*
1917 * **TCP_CONGESTION**.
1918 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1919 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1921 * 0 on success, or a negative error in case of failure.
1923 * long bpf_override_return(struct pt_regs *regs, u64 rc)
1925 * Used for error injection, this helper uses kprobes to override
1926 * the return value of the probed function, and to set it to *rc*.
1927 * The first argument is the context *regs* on which the kprobe
1930 * This helper works by setting the PC (program counter)
1931 * to an override function which is run in place of the original
1932 * probed function. This means the probed function is not run at
1933 * all. The replacement function just returns with the required
1936 * This helper has security implications, and thus is subject to
1937 * restrictions. It is only available if the kernel was compiled
1938 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1939 * option, and in this case it only works on functions tagged with
1940 * **ALLOW_ERROR_INJECTION** in the kernel code.
1942 * Also, the helper is only available for the architectures having
1943 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1944 * x86 architecture is the only one to support this feature.
1948 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1950 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1951 * for the full TCP socket associated to *bpf_sock_ops* to
1954 * The primary use of this field is to determine if there should
1955 * be calls to eBPF programs of type
1956 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1957 * code. A program of the same type can change its value, per
1958 * connection and as necessary, when the connection is
1959 * established. This field is directly accessible for reading, but
1960 * this helper must be used for updates in order to return an
1961 * error if an eBPF program tries to set a callback that is not
1962 * supported in the current kernel.
1964 * *argval* is a flag array which can combine these flags:
1966 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1967 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1968 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1969 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1971 * Therefore, this function can be used to clear a callback flag by
1972 * setting the appropriate bit to zero. e.g. to disable the RTO
1975 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1976 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1978 * Here are some examples of where one could call such eBPF
1982 * * When a packet is retransmitted.
1983 * * When the connection terminates.
1984 * * When a packet is sent.
1985 * * When a packet is received.
1987 * Code **-EINVAL** if the socket is not a full TCP socket;
1988 * otherwise, a positive number containing the bits that could not
1989 * be set is returned (which comes down to 0 if all bits were set
1992 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1994 * This helper is used in programs implementing policies at the
1995 * socket level. If the message *msg* is allowed to pass (i.e. if
1996 * the verdict eBPF program returns **SK_PASS**), redirect it to
1997 * the socket referenced by *map* (of type
1998 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1999 * egress interfaces can be used for redirection. The
2000 * **BPF_F_INGRESS** value in *flags* is used to make the
2001 * distinction (ingress path is selected if the flag is present,
2002 * egress path otherwise). This is the only flag supported for now.
2004 * **SK_PASS** on success, or **SK_DROP** on error.
2006 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2008 * For socket policies, apply the verdict of the eBPF program to
2009 * the next *bytes* (number of bytes) of message *msg*.
2011 * For example, this helper can be used in the following cases:
2013 * * A single **sendmsg**\ () or **sendfile**\ () system call
2014 * contains multiple logical messages that the eBPF program is
2015 * supposed to read and for which it should apply a verdict.
2016 * * An eBPF program only cares to read the first *bytes* of a
2017 * *msg*. If the message has a large payload, then setting up
2018 * and calling the eBPF program repeatedly for all bytes, even
2019 * though the verdict is already known, would create unnecessary
2022 * When called from within an eBPF program, the helper sets a
2023 * counter internal to the BPF infrastructure, that is used to
2024 * apply the last verdict to the next *bytes*. If *bytes* is
2025 * smaller than the current data being processed from a
2026 * **sendmsg**\ () or **sendfile**\ () system call, the first
2027 * *bytes* will be sent and the eBPF program will be re-run with
2028 * the pointer for start of data pointing to byte number *bytes*
2029 * **+ 1**. If *bytes* is larger than the current data being
2030 * processed, then the eBPF verdict will be applied to multiple
2031 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2034 * Note that if a socket closes with the internal counter holding
2035 * a non-zero value, this is not a problem because data is not
2036 * being buffered for *bytes* and is sent as it is received.
2040 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2042 * For socket policies, prevent the execution of the verdict eBPF
2043 * program for message *msg* until *bytes* (byte number) have been
2046 * This can be used when one needs a specific number of bytes
2047 * before a verdict can be assigned, even if the data spans
2048 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2049 * case would be a user calling **sendmsg**\ () repeatedly with
2050 * 1-byte long message segments. Obviously, this is bad for
2051 * performance, but it is still valid. If the eBPF program needs
2052 * *bytes* bytes to validate a header, this helper can be used to
2053 * prevent the eBPF program to be called again until *bytes* have
2058 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2060 * For socket policies, pull in non-linear data from user space
2061 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2062 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2065 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2066 * *msg* it can only parse data that the (**data**, **data_end**)
2067 * pointers have already consumed. For **sendmsg**\ () hooks this
2068 * is likely the first scatterlist element. But for calls relying
2069 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2070 * be the range (**0**, **0**) because the data is shared with
2071 * user space and by default the objective is to avoid allowing
2072 * user space to modify data while (or after) eBPF verdict is
2073 * being decided. This helper can be used to pull in data and to
2074 * set the start and end pointer to given values. Data will be
2075 * copied if necessary (i.e. if data was not linear and if start
2076 * and end pointers do not point to the same chunk).
2078 * A call to this helper is susceptible to change the underlying
2079 * packet buffer. Therefore, at load time, all checks on pointers
2080 * previously done by the verifier are invalidated and must be
2081 * performed again, if the helper is used in combination with
2082 * direct packet access.
2084 * All values for *flags* are reserved for future usage, and must
2087 * 0 on success, or a negative error in case of failure.
2089 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2091 * Bind the socket associated to *ctx* to the address pointed by
2092 * *addr*, of length *addr_len*. This allows for making outgoing
2093 * connection from the desired IP address, which can be useful for
2094 * example when all processes inside a cgroup should use one
2095 * single IP address on a host that has multiple IP configured.
2097 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2098 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2099 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2100 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2101 * behavior and lets the kernel efficiently pick up an unused
2102 * port as long as 4-tuple is unique. Passing non-zero port might
2103 * lead to degraded performance.
2105 * 0 on success, or a negative error in case of failure.
2107 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2109 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2110 * possible to both shrink and grow the packet tail.
2111 * Shrink done via *delta* being a negative integer.
2113 * A call to this helper is susceptible to change the underlying
2114 * packet buffer. Therefore, at load time, all checks on pointers
2115 * previously done by the verifier are invalidated and must be
2116 * performed again, if the helper is used in combination with
2117 * direct packet access.
2119 * 0 on success, or a negative error in case of failure.
2121 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2123 * Retrieve the XFRM state (IP transform framework, see also
2124 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2126 * The retrieved value is stored in the **struct bpf_xfrm_state**
2127 * pointed by *xfrm_state* and of length *size*.
2129 * All values for *flags* are reserved for future usage, and must
2132 * This helper is available only if the kernel was compiled with
2133 * **CONFIG_XFRM** configuration option.
2135 * 0 on success, or a negative error in case of failure.
2137 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2139 * Return a user or a kernel stack in bpf program provided buffer.
2140 * To achieve this, the helper needs *ctx*, which is a pointer
2141 * to the context on which the tracing program is executed.
2142 * To store the stacktrace, the bpf program provides *buf* with
2143 * a nonnegative *size*.
2145 * The last argument, *flags*, holds the number of stack frames to
2146 * skip (from 0 to 255), masked with
2147 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2148 * the following flags:
2150 * **BPF_F_USER_STACK**
2151 * Collect a user space stack instead of a kernel stack.
2152 * **BPF_F_USER_BUILD_ID**
2153 * Collect buildid+offset instead of ips for user stack,
2154 * only valid if **BPF_F_USER_STACK** is also specified.
2156 * **bpf_get_stack**\ () can collect up to
2157 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2158 * to sufficient large buffer size. Note that
2159 * this limit can be controlled with the **sysctl** program, and
2160 * that it should be manually increased in order to profile long
2161 * user stacks (such as stacks for Java programs). To do so, use:
2165 * # sysctl kernel.perf_event_max_stack=<new value>
2167 * A non-negative value equal to or less than *size* on success,
2168 * or a negative error in case of failure.
2170 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2172 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2173 * it provides an easy way to load *len* bytes from *offset*
2174 * from the packet associated to *skb*, into the buffer pointed
2175 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2176 * a fifth argument *start_header* exists in order to select a
2177 * base offset to start from. *start_header* can be one of:
2179 * **BPF_HDR_START_MAC**
2180 * Base offset to load data from is *skb*'s mac header.
2181 * **BPF_HDR_START_NET**
2182 * Base offset to load data from is *skb*'s network header.
2184 * In general, "direct packet access" is the preferred method to
2185 * access packet data, however, this helper is in particular useful
2186 * in socket filters where *skb*\ **->data** does not always point
2187 * to the start of the mac header and where "direct packet access"
2190 * 0 on success, or a negative error in case of failure.
2192 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2194 * Do FIB lookup in kernel tables using parameters in *params*.
2195 * If lookup is successful and result shows packet is to be
2196 * forwarded, the neighbor tables are searched for the nexthop.
2197 * If successful (ie., FIB lookup shows forwarding and nexthop
2198 * is resolved), the nexthop address is returned in ipv4_dst
2199 * or ipv6_dst based on family, smac is set to mac address of
2200 * egress device, dmac is set to nexthop mac address, rt_metric
2201 * is set to metric from route (IPv4/IPv6 only), and ifindex
2202 * is set to the device index of the nexthop from the FIB lookup.
2204 * *plen* argument is the size of the passed in struct.
2205 * *flags* argument can be a combination of one or more of the
2208 * **BPF_FIB_LOOKUP_DIRECT**
2209 * Do a direct table lookup vs full lookup using FIB
2211 * **BPF_FIB_LOOKUP_OUTPUT**
2212 * Perform lookup from an egress perspective (default is
2215 * *ctx* is either **struct xdp_md** for XDP programs or
2216 * **struct sk_buff** tc cls_act programs.
2218 * * < 0 if any input argument is invalid
2219 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2220 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2221 * packet is not forwarded or needs assist from full stack
2223 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2225 * Add an entry to, or update a sockhash *map* referencing sockets.
2226 * The *skops* is used as a new value for the entry associated to
2227 * *key*. *flags* is one of:
2230 * The entry for *key* must not exist in the map.
2232 * The entry for *key* must already exist in the map.
2234 * No condition on the existence of the entry for *key*.
2236 * If the *map* has eBPF programs (parser and verdict), those will
2237 * be inherited by the socket being added. If the socket is
2238 * already attached to eBPF programs, this results in an error.
2240 * 0 on success, or a negative error in case of failure.
2242 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2244 * This helper is used in programs implementing policies at the
2245 * socket level. If the message *msg* is allowed to pass (i.e. if
2246 * the verdict eBPF program returns **SK_PASS**), redirect it to
2247 * the socket referenced by *map* (of type
2248 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2249 * egress interfaces can be used for redirection. The
2250 * **BPF_F_INGRESS** value in *flags* is used to make the
2251 * distinction (ingress path is selected if the flag is present,
2252 * egress path otherwise). This is the only flag supported for now.
2254 * **SK_PASS** on success, or **SK_DROP** on error.
2256 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2258 * This helper is used in programs implementing policies at the
2259 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2260 * if the verdict eBPF program returns **SK_PASS**), redirect it
2261 * to the socket referenced by *map* (of type
2262 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2263 * egress interfaces can be used for redirection. The
2264 * **BPF_F_INGRESS** value in *flags* is used to make the
2265 * distinction (ingress path is selected if the flag is present,
2266 * egress otherwise). This is the only flag supported for now.
2268 * **SK_PASS** on success, or **SK_DROP** on error.
2270 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2272 * Encapsulate the packet associated to *skb* within a Layer 3
2273 * protocol header. This header is provided in the buffer at
2274 * address *hdr*, with *len* its size in bytes. *type* indicates
2275 * the protocol of the header and can be one of:
2277 * **BPF_LWT_ENCAP_SEG6**
2278 * IPv6 encapsulation with Segment Routing Header
2279 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2280 * the IPv6 header is computed by the kernel.
2281 * **BPF_LWT_ENCAP_SEG6_INLINE**
2282 * Only works if *skb* contains an IPv6 packet. Insert a
2283 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2285 * **BPF_LWT_ENCAP_IP**
2286 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2287 * must be IPv4 or IPv6, followed by zero or more
2288 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2289 * total bytes in all prepended headers. Please note that
2290 * if **skb_is_gso**\ (*skb*) is true, no more than two
2291 * headers can be prepended, and the inner header, if
2292 * present, should be either GRE or UDP/GUE.
2294 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2295 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2296 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2297 * **BPF_PROG_TYPE_LWT_XMIT**.
2299 * A call to this helper is susceptible to change the underlying
2300 * packet buffer. Therefore, at load time, all checks on pointers
2301 * previously done by the verifier are invalidated and must be
2302 * performed again, if the helper is used in combination with
2303 * direct packet access.
2305 * 0 on success, or a negative error in case of failure.
2307 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2309 * Store *len* bytes from address *from* into the packet
2310 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2311 * inside the outermost IPv6 Segment Routing Header can be
2312 * modified through this helper.
2314 * A call to this helper is susceptible to change the underlying
2315 * packet buffer. Therefore, at load time, all checks on pointers
2316 * previously done by the verifier are invalidated and must be
2317 * performed again, if the helper is used in combination with
2318 * direct packet access.
2320 * 0 on success, or a negative error in case of failure.
2322 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2324 * Adjust the size allocated to TLVs in the outermost IPv6
2325 * Segment Routing Header contained in the packet associated to
2326 * *skb*, at position *offset* by *delta* bytes. Only offsets
2327 * after the segments are accepted. *delta* can be as well
2328 * positive (growing) as negative (shrinking).
2330 * A call to this helper is susceptible to change the underlying
2331 * packet buffer. Therefore, at load time, all checks on pointers
2332 * previously done by the verifier are invalidated and must be
2333 * performed again, if the helper is used in combination with
2334 * direct packet access.
2336 * 0 on success, or a negative error in case of failure.
2338 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2340 * Apply an IPv6 Segment Routing action of type *action* to the
2341 * packet associated to *skb*. Each action takes a parameter
2342 * contained at address *param*, and of length *param_len* bytes.
2343 * *action* can be one of:
2345 * **SEG6_LOCAL_ACTION_END_X**
2346 * End.X action: Endpoint with Layer-3 cross-connect.
2347 * Type of *param*: **struct in6_addr**.
2348 * **SEG6_LOCAL_ACTION_END_T**
2349 * End.T action: Endpoint with specific IPv6 table lookup.
2350 * Type of *param*: **int**.
2351 * **SEG6_LOCAL_ACTION_END_B6**
2352 * End.B6 action: Endpoint bound to an SRv6 policy.
2353 * Type of *param*: **struct ipv6_sr_hdr**.
2354 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2355 * End.B6.Encap action: Endpoint bound to an SRv6
2356 * encapsulation policy.
2357 * Type of *param*: **struct ipv6_sr_hdr**.
2359 * A call to this helper is susceptible to change the underlying
2360 * packet buffer. Therefore, at load time, all checks on pointers
2361 * previously done by the verifier are invalidated and must be
2362 * performed again, if the helper is used in combination with
2363 * direct packet access.
2365 * 0 on success, or a negative error in case of failure.
2367 * long bpf_rc_repeat(void *ctx)
2369 * This helper is used in programs implementing IR decoding, to
2370 * report a successfully decoded repeat key message. This delays
2371 * the generation of a key up event for previously generated
2374 * Some IR protocols like NEC have a special IR message for
2375 * repeating last button, for when a button is held down.
2377 * The *ctx* should point to the lirc sample as passed into
2380 * This helper is only available is the kernel was compiled with
2381 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2386 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2388 * This helper is used in programs implementing IR decoding, to
2389 * report a successfully decoded key press with *scancode*,
2390 * *toggle* value in the given *protocol*. The scancode will be
2391 * translated to a keycode using the rc keymap, and reported as
2392 * an input key down event. After a period a key up event is
2393 * generated. This period can be extended by calling either
2394 * **bpf_rc_keydown**\ () again with the same values, or calling
2395 * **bpf_rc_repeat**\ ().
2397 * Some protocols include a toggle bit, in case the button was
2398 * released and pressed again between consecutive scancodes.
2400 * The *ctx* should point to the lirc sample as passed into
2403 * The *protocol* is the decoded protocol number (see
2404 * **enum rc_proto** for some predefined values).
2406 * This helper is only available is the kernel was compiled with
2407 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2412 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2414 * Return the cgroup v2 id of the socket associated with the *skb*.
2415 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2416 * helper for cgroup v1 by providing a tag resp. identifier that
2417 * can be matched on or used for map lookups e.g. to implement
2418 * policy. The cgroup v2 id of a given path in the hierarchy is
2419 * exposed in user space through the f_handle API in order to get
2420 * to the same 64-bit id.
2422 * This helper can be used on TC egress path, but not on ingress,
2423 * and is available only if the kernel was compiled with the
2424 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2426 * The id is returned or 0 in case the id could not be retrieved.
2428 * u64 bpf_get_current_cgroup_id(void)
2430 * A 64-bit integer containing the current cgroup id based
2431 * on the cgroup within which the current task is running.
2433 * void *bpf_get_local_storage(void *map, u64 flags)
2435 * Get the pointer to the local storage area.
2436 * The type and the size of the local storage is defined
2437 * by the *map* argument.
2438 * The *flags* meaning is specific for each map type,
2439 * and has to be 0 for cgroup local storage.
2441 * Depending on the BPF program type, a local storage area
2442 * can be shared between multiple instances of the BPF program,
2443 * running simultaneously.
2445 * A user should care about the synchronization by himself.
2446 * For example, by using the **BPF_STX_XADD** instruction to alter
2449 * A pointer to the local storage area.
2451 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2453 * Select a **SO_REUSEPORT** socket from a
2454 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2455 * It checks the selected socket is matching the incoming
2456 * request in the socket buffer.
2458 * 0 on success, or a negative error in case of failure.
2460 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2462 * Return id of cgroup v2 that is ancestor of cgroup associated
2463 * with the *skb* at the *ancestor_level*. The root cgroup is at
2464 * *ancestor_level* zero and each step down the hierarchy
2465 * increments the level. If *ancestor_level* == level of cgroup
2466 * associated with *skb*, then return value will be same as that
2467 * of **bpf_skb_cgroup_id**\ ().
2469 * The helper is useful to implement policies based on cgroups
2470 * that are upper in hierarchy than immediate cgroup associated
2473 * The format of returned id and helper limitations are same as in
2474 * **bpf_skb_cgroup_id**\ ().
2476 * The id is returned or 0 in case the id could not be retrieved.
2478 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2480 * Look for TCP socket matching *tuple*, optionally in a child
2481 * network namespace *netns*. The return value must be checked,
2482 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2484 * The *ctx* should point to the context of the program, such as
2485 * the skb or socket (depending on the hook in use). This is used
2486 * to determine the base network namespace for the lookup.
2488 * *tuple_size* must be one of:
2490 * **sizeof**\ (*tuple*\ **->ipv4**)
2491 * Look for an IPv4 socket.
2492 * **sizeof**\ (*tuple*\ **->ipv6**)
2493 * Look for an IPv6 socket.
2495 * If the *netns* is a negative signed 32-bit integer, then the
2496 * socket lookup table in the netns associated with the *ctx*
2497 * will be used. For the TC hooks, this is the netns of the device
2498 * in the skb. For socket hooks, this is the netns of the socket.
2499 * If *netns* is any other signed 32-bit value greater than or
2500 * equal to zero then it specifies the ID of the netns relative to
2501 * the netns associated with the *ctx*. *netns* values beyond the
2502 * range of 32-bit integers are reserved for future use.
2504 * All values for *flags* are reserved for future usage, and must
2507 * This helper is available only if the kernel was compiled with
2508 * **CONFIG_NET** configuration option.
2510 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2511 * For sockets with reuseport option, the **struct bpf_sock**
2512 * result is from *reuse*\ **->socks**\ [] using the hash of the
2515 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2517 * Look for UDP socket matching *tuple*, optionally in a child
2518 * network namespace *netns*. The return value must be checked,
2519 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2521 * The *ctx* should point to the context of the program, such as
2522 * the skb or socket (depending on the hook in use). This is used
2523 * to determine the base network namespace for the lookup.
2525 * *tuple_size* must be one of:
2527 * **sizeof**\ (*tuple*\ **->ipv4**)
2528 * Look for an IPv4 socket.
2529 * **sizeof**\ (*tuple*\ **->ipv6**)
2530 * Look for an IPv6 socket.
2532 * If the *netns* is a negative signed 32-bit integer, then the
2533 * socket lookup table in the netns associated with the *ctx*
2534 * will be used. For the TC hooks, this is the netns of the device
2535 * in the skb. For socket hooks, this is the netns of the socket.
2536 * If *netns* is any other signed 32-bit value greater than or
2537 * equal to zero then it specifies the ID of the netns relative to
2538 * the netns associated with the *ctx*. *netns* values beyond the
2539 * range of 32-bit integers are reserved for future use.
2541 * All values for *flags* are reserved for future usage, and must
2544 * This helper is available only if the kernel was compiled with
2545 * **CONFIG_NET** configuration option.
2547 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2548 * For sockets with reuseport option, the **struct bpf_sock**
2549 * result is from *reuse*\ **->socks**\ [] using the hash of the
2552 * long bpf_sk_release(void *sock)
2554 * Release the reference held by *sock*. *sock* must be a
2555 * non-**NULL** pointer that was returned from
2556 * **bpf_sk_lookup_xxx**\ ().
2558 * 0 on success, or a negative error in case of failure.
2560 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2562 * Push an element *value* in *map*. *flags* is one of:
2565 * If the queue/stack is full, the oldest element is
2566 * removed to make room for this.
2568 * 0 on success, or a negative error in case of failure.
2570 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
2572 * Pop an element from *map*.
2574 * 0 on success, or a negative error in case of failure.
2576 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
2578 * Get an element from *map* without removing it.
2580 * 0 on success, or a negative error in case of failure.
2582 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2584 * For socket policies, insert *len* bytes into *msg* at offset
2587 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2588 * *msg* it may want to insert metadata or options into the *msg*.
2589 * This can later be read and used by any of the lower layer BPF
2592 * This helper may fail if under memory pressure (a malloc
2593 * fails) in these cases BPF programs will get an appropriate
2594 * error and BPF programs will need to handle them.
2596 * 0 on success, or a negative error in case of failure.
2598 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2600 * Will remove *len* bytes from a *msg* starting at byte *start*.
2601 * This may result in **ENOMEM** errors under certain situations if
2602 * an allocation and copy are required due to a full ring buffer.
2603 * However, the helper will try to avoid doing the allocation
2604 * if possible. Other errors can occur if input parameters are
2605 * invalid either due to *start* byte not being valid part of *msg*
2606 * payload and/or *pop* value being to large.
2608 * 0 on success, or a negative error in case of failure.
2610 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2612 * This helper is used in programs implementing IR decoding, to
2613 * report a successfully decoded pointer movement.
2615 * The *ctx* should point to the lirc sample as passed into
2618 * This helper is only available is the kernel was compiled with
2619 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2624 * long bpf_spin_lock(struct bpf_spin_lock *lock)
2626 * Acquire a spinlock represented by the pointer *lock*, which is
2627 * stored as part of a value of a map. Taking the lock allows to
2628 * safely update the rest of the fields in that value. The
2629 * spinlock can (and must) later be released with a call to
2630 * **bpf_spin_unlock**\ (\ *lock*\ ).
2632 * Spinlocks in BPF programs come with a number of restrictions
2635 * * **bpf_spin_lock** objects are only allowed inside maps of
2636 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2637 * list could be extended in the future).
2638 * * BTF description of the map is mandatory.
2639 * * The BPF program can take ONE lock at a time, since taking two
2640 * or more could cause dead locks.
2641 * * Only one **struct bpf_spin_lock** is allowed per map element.
2642 * * When the lock is taken, calls (either BPF to BPF or helpers)
2644 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2645 * allowed inside a spinlock-ed region.
2646 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2647 * the lock, on all execution paths, before it returns.
2648 * * The BPF program can access **struct bpf_spin_lock** only via
2649 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2650 * helpers. Loading or storing data into the **struct
2651 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2652 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2653 * of the map value must be a struct and have **struct
2654 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2655 * Nested lock inside another struct is not allowed.
2656 * * The **struct bpf_spin_lock** *lock* field in a map value must
2657 * be aligned on a multiple of 4 bytes in that value.
2658 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2659 * the **bpf_spin_lock** field to user space.
2660 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2661 * a BPF program, do not update the **bpf_spin_lock** field.
2662 * * **bpf_spin_lock** cannot be on the stack or inside a
2663 * networking packet (it can only be inside of a map values).
2664 * * **bpf_spin_lock** is available to root only.
2665 * * Tracing programs and socket filter programs cannot use
2666 * **bpf_spin_lock**\ () due to insufficient preemption checks
2667 * (but this may change in the future).
2668 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2672 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
2674 * Release the *lock* previously locked by a call to
2675 * **bpf_spin_lock**\ (\ *lock*\ ).
2679 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2681 * This helper gets a **struct bpf_sock** pointer such
2682 * that all the fields in this **bpf_sock** can be accessed.
2684 * A **struct bpf_sock** pointer on success, or **NULL** in
2687 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2689 * This helper gets a **struct bpf_tcp_sock** pointer from a
2690 * **struct bpf_sock** pointer.
2692 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2695 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
2697 * Set ECN (Explicit Congestion Notification) field of IP header
2698 * to **CE** (Congestion Encountered) if current value is **ECT**
2699 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2702 * 1 if the **CE** flag is set (either by the current helper call
2703 * or because it was already present), 0 if it is not set.
2705 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2707 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2708 * **bpf_sk_release**\ () is unnecessary and not allowed.
2710 * A **struct bpf_sock** pointer on success, or **NULL** in
2713 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2715 * Look for TCP socket matching *tuple*, optionally in a child
2716 * network namespace *netns*. The return value must be checked,
2717 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2719 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2720 * that it also returns timewait or request sockets. Use
2721 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2724 * This helper is available only if the kernel was compiled with
2725 * **CONFIG_NET** configuration option.
2727 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2728 * For sockets with reuseport option, the **struct bpf_sock**
2729 * result is from *reuse*\ **->socks**\ [] using the hash of the
2732 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2734 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2735 * the listening socket in *sk*.
2737 * *iph* points to the start of the IPv4 or IPv6 header, while
2738 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2739 * **sizeof**\ (**struct ip6hdr**).
2741 * *th* points to the start of the TCP header, while *th_len*
2742 * contains **sizeof**\ (**struct tcphdr**).
2744 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2747 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2749 * Get name of sysctl in /proc/sys/ and copy it into provided by
2750 * program buffer *buf* of size *buf_len*.
2752 * The buffer is always NUL terminated, unless it's zero-sized.
2754 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2755 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2756 * only (e.g. "tcp_mem").
2758 * Number of character copied (not including the trailing NUL).
2760 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2761 * truncated name in this case).
2763 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2765 * Get current value of sysctl as it is presented in /proc/sys
2766 * (incl. newline, etc), and copy it as a string into provided
2767 * by program buffer *buf* of size *buf_len*.
2769 * The whole value is copied, no matter what file position user
2770 * space issued e.g. sys_read at.
2772 * The buffer is always NUL terminated, unless it's zero-sized.
2774 * Number of character copied (not including the trailing NUL).
2776 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2777 * truncated name in this case).
2779 * **-EINVAL** if current value was unavailable, e.g. because
2780 * sysctl is uninitialized and read returns -EIO for it.
2782 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2784 * Get new value being written by user space to sysctl (before
2785 * the actual write happens) and copy it as a string into
2786 * provided by program buffer *buf* of size *buf_len*.
2788 * User space may write new value at file position > 0.
2790 * The buffer is always NUL terminated, unless it's zero-sized.
2792 * Number of character copied (not including the trailing NUL).
2794 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2795 * truncated name in this case).
2797 * **-EINVAL** if sysctl is being read.
2799 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2801 * Override new value being written by user space to sysctl with
2802 * value provided by program in buffer *buf* of size *buf_len*.
2804 * *buf* should contain a string in same form as provided by user
2805 * space on sysctl write.
2807 * User space may write new value at file position > 0. To override
2808 * the whole sysctl value file position should be set to zero.
2812 * **-E2BIG** if the *buf_len* is too big.
2814 * **-EINVAL** if sysctl is being read.
2816 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2818 * Convert the initial part of the string from buffer *buf* of
2819 * size *buf_len* to a long integer according to the given base
2820 * and save the result in *res*.
2822 * The string may begin with an arbitrary amount of white space
2823 * (as determined by **isspace**\ (3)) followed by a single
2824 * optional '**-**' sign.
2826 * Five least significant bits of *flags* encode base, other bits
2827 * are currently unused.
2829 * Base must be either 8, 10, 16 or 0 to detect it automatically
2830 * similar to user space **strtol**\ (3).
2832 * Number of characters consumed on success. Must be positive but
2833 * no more than *buf_len*.
2835 * **-EINVAL** if no valid digits were found or unsupported base
2838 * **-ERANGE** if resulting value was out of range.
2840 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2842 * Convert the initial part of the string from buffer *buf* of
2843 * size *buf_len* to an unsigned long integer according to the
2844 * given base and save the result in *res*.
2846 * The string may begin with an arbitrary amount of white space
2847 * (as determined by **isspace**\ (3)).
2849 * Five least significant bits of *flags* encode base, other bits
2850 * are currently unused.
2852 * Base must be either 8, 10, 16 or 0 to detect it automatically
2853 * similar to user space **strtoul**\ (3).
2855 * Number of characters consumed on success. Must be positive but
2856 * no more than *buf_len*.
2858 * **-EINVAL** if no valid digits were found or unsupported base
2861 * **-ERANGE** if resulting value was out of range.
2863 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
2865 * Get a bpf-local-storage from a *sk*.
2867 * Logically, it could be thought of getting the value from
2868 * a *map* with *sk* as the **key**. From this
2869 * perspective, the usage is not much different from
2870 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2871 * helper enforces the key must be a full socket and the map must
2872 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2874 * Underneath, the value is stored locally at *sk* instead of
2875 * the *map*. The *map* is used as the bpf-local-storage
2876 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2877 * searched against all bpf-local-storages residing at *sk*.
2879 * *sk* is a kernel **struct sock** pointer for LSM program.
2880 * *sk* is a **struct bpf_sock** pointer for other program types.
2882 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2883 * used such that a new bpf-local-storage will be
2884 * created if one does not exist. *value* can be used
2885 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2886 * the initial value of a bpf-local-storage. If *value* is
2887 * **NULL**, the new bpf-local-storage will be zero initialized.
2889 * A bpf-local-storage pointer is returned on success.
2891 * **NULL** if not found or there was an error in adding
2892 * a new bpf-local-storage.
2894 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
2896 * Delete a bpf-local-storage from a *sk*.
2900 * **-ENOENT** if the bpf-local-storage cannot be found.
2901 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
2903 * long bpf_send_signal(u32 sig)
2905 * Send signal *sig* to the process of the current task.
2906 * The signal may be delivered to any of this process's threads.
2908 * 0 on success or successfully queued.
2910 * **-EBUSY** if work queue under nmi is full.
2912 * **-EINVAL** if *sig* is invalid.
2914 * **-EPERM** if no permission to send the *sig*.
2916 * **-EAGAIN** if bpf program can try again.
2918 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2920 * Try to issue a SYN cookie for the packet with corresponding
2921 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2923 * *iph* points to the start of the IPv4 or IPv6 header, while
2924 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2925 * **sizeof**\ (**struct ip6hdr**).
2927 * *th* points to the start of the TCP header, while *th_len*
2928 * contains the length of the TCP header.
2930 * On success, lower 32 bits hold the generated SYN cookie in
2931 * followed by 16 bits which hold the MSS value for that cookie,
2932 * and the top 16 bits are unused.
2934 * On failure, the returned value is one of the following:
2936 * **-EINVAL** SYN cookie cannot be issued due to error
2938 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2940 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2942 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2944 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2946 * Write raw *data* blob into a special BPF perf event held by
2947 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2948 * event must have the following attributes: **PERF_SAMPLE_RAW**
2949 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2950 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2952 * The *flags* are used to indicate the index in *map* for which
2953 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2954 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2955 * to indicate that the index of the current CPU core should be
2958 * The value to write, of *size*, is passed through eBPF stack and
2959 * pointed by *data*.
2961 * *ctx* is a pointer to in-kernel struct sk_buff.
2963 * This helper is similar to **bpf_perf_event_output**\ () but
2964 * restricted to raw_tracepoint bpf programs.
2966 * 0 on success, or a negative error in case of failure.
2968 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2970 * Safely attempt to read *size* bytes from user space address
2971 * *unsafe_ptr* and store the data in *dst*.
2973 * 0 on success, or a negative error in case of failure.
2975 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2977 * Safely attempt to read *size* bytes from kernel space address
2978 * *unsafe_ptr* and store the data in *dst*.
2980 * 0 on success, or a negative error in case of failure.
2982 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2984 * Copy a NUL terminated string from an unsafe user address
2985 * *unsafe_ptr* to *dst*. The *size* should include the
2986 * terminating NUL byte. In case the string length is smaller than
2987 * *size*, the target is not padded with further NUL bytes. If the
2988 * string length is larger than *size*, just *size*-1 bytes are
2989 * copied and the last byte is set to NUL.
2991 * On success, the length of the copied string is returned. This
2992 * makes this helper useful in tracing programs for reading
2993 * strings, and more importantly to get its length at runtime. See
2994 * the following snippet:
2998 * SEC("kprobe/sys_open")
2999 * void bpf_sys_open(struct pt_regs *ctx)
3001 * char buf[PATHLEN]; // PATHLEN is defined to 256
3002 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
3005 * // Consume buf, for example push it to
3006 * // userspace via bpf_perf_event_output(); we
3007 * // can use res (the string length) as event
3008 * // size, after checking its boundaries.
3011 * In comparison, using **bpf_probe_read_user**\ () helper here
3012 * instead to read the string would require to estimate the length
3013 * at compile time, and would often result in copying more memory
3016 * Another useful use case is when parsing individual process
3017 * arguments or individual environment variables navigating
3018 * *current*\ **->mm->arg_start** and *current*\
3019 * **->mm->env_start**: using this helper and the return value,
3020 * one can quickly iterate at the right offset of the memory area.
3022 * On success, the strictly positive length of the string,
3023 * including the trailing NUL character. On error, a negative
3026 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3028 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3029 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3031 * On success, the strictly positive length of the string, including
3032 * the trailing NUL character. On error, a negative value.
3034 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3036 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3037 * *rcv_nxt* is the ack_seq to be sent out.
3039 * 0 on success, or a negative error in case of failure.
3041 * long bpf_send_signal_thread(u32 sig)
3043 * Send signal *sig* to the thread corresponding to the current task.
3045 * 0 on success or successfully queued.
3047 * **-EBUSY** if work queue under nmi is full.
3049 * **-EINVAL** if *sig* is invalid.
3051 * **-EPERM** if no permission to send the *sig*.
3053 * **-EAGAIN** if bpf program can try again.
3055 * u64 bpf_jiffies64(void)
3057 * Obtain the 64bit jiffies
3059 * The 64 bit jiffies
3061 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3063 * For an eBPF program attached to a perf event, retrieve the
3064 * branch records (**struct perf_branch_entry**) associated to *ctx*
3065 * and store it in the buffer pointed by *buf* up to size
3068 * On success, number of bytes written to *buf*. On error, a
3071 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3072 * instead return the number of bytes required to store all the
3073 * branch entries. If this flag is set, *buf* may be NULL.
3075 * **-EINVAL** if arguments invalid or **size** not a multiple
3076 * of **sizeof**\ (**struct perf_branch_entry**\ ).
3078 * **-ENOENT** if architecture does not support branch records.
3080 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3082 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3083 * *namespace* will be returned in *nsdata*.
3085 * 0 on success, or one of the following in case of failure:
3087 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3088 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3090 * **-ENOENT** if pidns does not exists for the current task.
3092 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3094 * Write raw *data* blob into a special BPF perf event held by
3095 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3096 * event must have the following attributes: **PERF_SAMPLE_RAW**
3097 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3098 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3100 * The *flags* are used to indicate the index in *map* for which
3101 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3102 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3103 * to indicate that the index of the current CPU core should be
3106 * The value to write, of *size*, is passed through eBPF stack and
3107 * pointed by *data*.
3109 * *ctx* is a pointer to in-kernel struct xdp_buff.
3111 * This helper is similar to **bpf_perf_eventoutput**\ () but
3112 * restricted to raw_tracepoint bpf programs.
3114 * 0 on success, or a negative error in case of failure.
3116 * u64 bpf_get_netns_cookie(void *ctx)
3118 * Retrieve the cookie (generated by the kernel) of the network
3119 * namespace the input *ctx* is associated with. The network
3120 * namespace cookie remains stable for its lifetime and provides
3121 * a global identifier that can be assumed unique. If *ctx* is
3122 * NULL, then the helper returns the cookie for the initial
3123 * network namespace. The cookie itself is very similar to that
3124 * of **bpf_get_socket_cookie**\ () helper, but for network
3125 * namespaces instead of sockets.
3127 * A 8-byte long opaque number.
3129 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3131 * Return id of cgroup v2 that is ancestor of the cgroup associated
3132 * with the current task at the *ancestor_level*. The root cgroup
3133 * is at *ancestor_level* zero and each step down the hierarchy
3134 * increments the level. If *ancestor_level* == level of cgroup
3135 * associated with the current task, then return value will be the
3136 * same as that of **bpf_get_current_cgroup_id**\ ().
3138 * The helper is useful to implement policies based on cgroups
3139 * that are upper in hierarchy than immediate cgroup associated
3140 * with the current task.
3142 * The format of returned id and helper limitations are same as in
3143 * **bpf_get_current_cgroup_id**\ ().
3145 * The id is returned or 0 in case the id could not be retrieved.
3147 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
3149 * Helper is overloaded depending on BPF program type. This
3150 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3151 * **BPF_PROG_TYPE_SCHED_ACT** programs.
3153 * Assign the *sk* to the *skb*. When combined with appropriate
3154 * routing configuration to receive the packet towards the socket,
3155 * will cause *skb* to be delivered to the specified socket.
3156 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3157 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3158 * interfere with successful delivery to the socket.
3160 * This operation is only valid from TC ingress path.
3162 * The *flags* argument must be zero.
3164 * 0 on success, or a negative error in case of failure:
3166 * **-EINVAL** if specified *flags* are not supported.
3168 * **-ENOENT** if the socket is unavailable for assignment.
3170 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3172 * **-EOPNOTSUPP** if the operation is not supported, for example
3173 * a call from outside of TC ingress.
3175 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3178 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3180 * Helper is overloaded depending on BPF program type. This
3181 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
3183 * Select the *sk* as a result of a socket lookup.
3185 * For the operation to succeed passed socket must be compatible
3186 * with the packet description provided by the *ctx* object.
3188 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
3189 * be an exact match. While IP family (**AF_INET** or
3190 * **AF_INET6**) must be compatible, that is IPv6 sockets
3191 * that are not v6-only can be selected for IPv4 packets.
3193 * Only TCP listeners and UDP unconnected sockets can be
3194 * selected. *sk* can also be NULL to reset any previous
3197 * *flags* argument can combination of following values:
3199 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
3200 * socket selection, potentially done by a BPF program
3201 * that ran before us.
3203 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
3204 * load-balancing within reuseport group for the socket
3207 * On success *ctx->sk* will point to the selected socket.
3210 * 0 on success, or a negative errno in case of failure.
3212 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
3213 * not compatible with packet family (*ctx->family*).
3215 * * **-EEXIST** if socket has been already selected,
3216 * potentially by another program, and
3217 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
3219 * * **-EINVAL** if unsupported flags were specified.
3221 * * **-EPROTOTYPE** if socket L4 protocol
3222 * (*sk->protocol*) doesn't match packet protocol
3223 * (*ctx->protocol*).
3225 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
3226 * state (TCP listening or UDP unconnected).
3228 * u64 bpf_ktime_get_boot_ns(void)
3230 * Return the time elapsed since system boot, in nanoseconds.
3231 * Does include the time the system was suspended.
3232 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
3236 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3238 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3239 * out the format string.
3240 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3241 * the format string itself. The *data* and *data_len* are format string
3242 * arguments. The *data* are a **u64** array and corresponding format string
3243 * values are stored in the array. For strings and pointers where pointees
3244 * are accessed, only the pointer values are stored in the *data* array.
3245 * The *data_len* is the size of *data* in bytes.
3247 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3248 * Reading kernel memory may fail due to either invalid address or
3249 * valid address but requiring a major memory fault. If reading kernel memory
3250 * fails, the string for **%s** will be an empty string, and the ip
3251 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
3252 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
3254 * 0 on success, or a negative error in case of failure:
3256 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3257 * by returning 1 from bpf program.
3259 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3261 * **-E2BIG** if *fmt* contains too many format specifiers.
3263 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3265 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
3267 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
3268 * The *m* represents the seq_file. The *data* and *len* represent the
3269 * data to write in bytes.
3271 * 0 on success, or a negative error in case of failure:
3273 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3275 * u64 bpf_sk_cgroup_id(void *sk)
3277 * Return the cgroup v2 id of the socket *sk*.
3279 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
3280 * returned from **bpf_sk_lookup_xxx**\ (),
3281 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3282 * same as in **bpf_skb_cgroup_id**\ ().
3284 * This helper is available only if the kernel was compiled with
3285 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3287 * The id is returned or 0 in case the id could not be retrieved.
3289 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
3291 * Return id of cgroup v2 that is ancestor of cgroup associated
3292 * with the *sk* at the *ancestor_level*. The root cgroup is at
3293 * *ancestor_level* zero and each step down the hierarchy
3294 * increments the level. If *ancestor_level* == level of cgroup
3295 * associated with *sk*, then return value will be same as that
3296 * of **bpf_sk_cgroup_id**\ ().
3298 * The helper is useful to implement policies based on cgroups
3299 * that are upper in hierarchy than immediate cgroup associated
3302 * The format of returned id and helper limitations are same as in
3303 * **bpf_sk_cgroup_id**\ ().
3305 * The id is returned or 0 in case the id could not be retrieved.
3307 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
3309 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
3310 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3311 * of new data availability is sent.
3312 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3313 * of new data availability is sent unconditionally.
3315 * 0 on success, or a negative error in case of failure.
3317 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3319 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3321 * Valid pointer with *size* bytes of memory available; NULL,
3324 * void bpf_ringbuf_submit(void *data, u64 flags)
3326 * Submit reserved ring buffer sample, pointed to by *data*.
3327 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3328 * of new data availability is sent.
3329 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3330 * of new data availability is sent unconditionally.
3332 * Nothing. Always succeeds.
3334 * void bpf_ringbuf_discard(void *data, u64 flags)
3336 * Discard reserved ring buffer sample, pointed to by *data*.
3337 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3338 * of new data availability is sent.
3339 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3340 * of new data availability is sent unconditionally.
3342 * Nothing. Always succeeds.
3344 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3346 * Query various characteristics of provided ring buffer. What
3347 * exactly is queries is determined by *flags*:
3349 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
3350 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
3351 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
3352 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
3354 * Data returned is just a momentary snapshot of actual values
3355 * and could be inaccurate, so this facility should be used to
3356 * power heuristics and for reporting, not to make 100% correct
3359 * Requested value, or 0, if *flags* are not recognized.
3361 * long bpf_csum_level(struct sk_buff *skb, u64 level)
3363 * Change the skbs checksum level by one layer up or down, or
3364 * reset it entirely to none in order to have the stack perform
3365 * checksum validation. The level is applicable to the following
3366 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
3367 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
3368 * through **bpf_skb_adjust_room**\ () helper with passing in
3369 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
3370 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
3371 * the UDP header is removed. Similarly, an encap of the latter
3372 * into the former could be accompanied by a helper call to
3373 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
3374 * skb is still intended to be processed in higher layers of the
3375 * stack instead of just egressing at tc.
3377 * There are three supported level settings at this time:
3379 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
3380 * with CHECKSUM_UNNECESSARY.
3381 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
3382 * with CHECKSUM_UNNECESSARY.
3383 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
3384 * sets CHECKSUM_NONE to force checksum validation by the stack.
3385 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
3388 * 0 on success, or a negative error in case of failure. In the
3389 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
3390 * is returned or the error code -EACCES in case the skb is not
3391 * subject to CHECKSUM_UNNECESSARY.
3393 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
3395 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
3397 * *sk* if casting is valid, or **NULL** otherwise.
3399 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
3401 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
3403 * *sk* if casting is valid, or **NULL** otherwise.
3405 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
3407 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
3409 * *sk* if casting is valid, or **NULL** otherwise.
3411 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
3413 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
3415 * *sk* if casting is valid, or **NULL** otherwise.
3417 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
3419 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
3421 * *sk* if casting is valid, or **NULL** otherwise.
3423 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
3425 * Return a user or a kernel stack in bpf program provided buffer.
3426 * To achieve this, the helper needs *task*, which is a valid
3427 * pointer to **struct task_struct**. To store the stacktrace, the
3428 * bpf program provides *buf* with a nonnegative *size*.
3430 * The last argument, *flags*, holds the number of stack frames to
3431 * skip (from 0 to 255), masked with
3432 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3433 * the following flags:
3435 * **BPF_F_USER_STACK**
3436 * Collect a user space stack instead of a kernel stack.
3437 * **BPF_F_USER_BUILD_ID**
3438 * Collect buildid+offset instead of ips for user stack,
3439 * only valid if **BPF_F_USER_STACK** is also specified.
3441 * **bpf_get_task_stack**\ () can collect up to
3442 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3443 * to sufficient large buffer size. Note that
3444 * this limit can be controlled with the **sysctl** program, and
3445 * that it should be manually increased in order to profile long
3446 * user stacks (such as stacks for Java programs). To do so, use:
3450 * # sysctl kernel.perf_event_max_stack=<new value>
3452 * A non-negative value equal to or less than *size* on success,
3453 * or a negative error in case of failure.
3455 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
3457 * Load header option. Support reading a particular TCP header
3458 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
3460 * If *flags* is 0, it will search the option from the
3461 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
3462 * has details on what skb_data contains under different
3463 * *skops*\ **->op**.
3465 * The first byte of the *searchby_res* specifies the
3466 * kind that it wants to search.
3468 * If the searching kind is an experimental kind
3469 * (i.e. 253 or 254 according to RFC6994). It also
3470 * needs to specify the "magic" which is either
3471 * 2 bytes or 4 bytes. It then also needs to
3472 * specify the size of the magic by using
3473 * the 2nd byte which is "kind-length" of a TCP
3474 * header option and the "kind-length" also
3475 * includes the first 2 bytes "kind" and "kind-length"
3476 * itself as a normal TCP header option also does.
3478 * For example, to search experimental kind 254 with
3479 * 2 byte magic 0xeB9F, the searchby_res should be
3480 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
3482 * To search for the standard window scale option (3),
3483 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
3484 * Note, kind-length must be 0 for regular option.
3486 * Searching for No-Op (0) and End-of-Option-List (1) are
3489 * *len* must be at least 2 bytes which is the minimal size
3490 * of a header option.
3494 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
3495 * saved_syn packet or the just-received syn packet.
3498 * > 0 when found, the header option is copied to *searchby_res*.
3499 * The return value is the total length copied. On failure, a
3500 * negative error code is returned:
3502 * **-EINVAL** if a parameter is invalid.
3504 * **-ENOMSG** if the option is not found.
3506 * **-ENOENT** if no syn packet is available when
3507 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
3509 * **-ENOSPC** if there is not enough space. Only *len* number of
3512 * **-EFAULT** on failure to parse the header options in the
3515 * **-EPERM** if the helper cannot be used under the current
3516 * *skops*\ **->op**.
3518 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
3520 * Store header option. The data will be copied
3521 * from buffer *from* with length *len* to the TCP header.
3523 * The buffer *from* should have the whole option that
3524 * includes the kind, kind-length, and the actual
3525 * option data. The *len* must be at least kind-length
3526 * long. The kind-length does not have to be 4 byte
3527 * aligned. The kernel will take care of the padding
3528 * and setting the 4 bytes aligned value to th->doff.
3530 * This helper will check for duplicated option
3531 * by searching the same option in the outgoing skb.
3533 * This helper can only be called during
3534 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
3537 * 0 on success, or negative error in case of failure:
3539 * **-EINVAL** If param is invalid.
3541 * **-ENOSPC** if there is not enough space in the header.
3542 * Nothing has been written
3544 * **-EEXIST** if the option already exists.
3546 * **-EFAULT** on failrue to parse the existing header options.
3548 * **-EPERM** if the helper cannot be used under the current
3549 * *skops*\ **->op**.
3551 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
3553 * Reserve *len* bytes for the bpf header option. The
3554 * space will be used by **bpf_store_hdr_opt**\ () later in
3555 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
3557 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
3558 * the total number of bytes will be reserved.
3560 * This helper can only be called during
3561 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
3564 * 0 on success, or negative error in case of failure:
3566 * **-EINVAL** if a parameter is invalid.
3568 * **-ENOSPC** if there is not enough space in the header.
3570 * **-EPERM** if the helper cannot be used under the current
3571 * *skops*\ **->op**.
3573 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
3575 * Get a bpf_local_storage from an *inode*.
3577 * Logically, it could be thought of as getting the value from
3578 * a *map* with *inode* as the **key**. From this
3579 * perspective, the usage is not much different from
3580 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
3581 * helper enforces the key must be an inode and the map must also
3582 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
3584 * Underneath, the value is stored locally at *inode* instead of
3585 * the *map*. The *map* is used as the bpf-local-storage
3586 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3587 * searched against all bpf_local_storage residing at *inode*.
3589 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
3590 * used such that a new bpf_local_storage will be
3591 * created if one does not exist. *value* can be used
3592 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
3593 * the initial value of a bpf_local_storage. If *value* is
3594 * **NULL**, the new bpf_local_storage will be zero initialized.
3596 * A bpf_local_storage pointer is returned on success.
3598 * **NULL** if not found or there was an error in adding
3599 * a new bpf_local_storage.
3601 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
3603 * Delete a bpf_local_storage from an *inode*.
3607 * **-ENOENT** if the bpf_local_storage cannot be found.
3609 * long bpf_d_path(struct path *path, char *buf, u32 sz)
3611 * Return full path for given **struct path** object, which
3612 * needs to be the kernel BTF *path* object. The path is
3613 * returned in the provided buffer *buf* of size *sz* and
3614 * is zero terminated.
3617 * On success, the strictly positive length of the string,
3618 * including the trailing NUL character. On error, a negative
3621 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
3623 * Read *size* bytes from user space address *user_ptr* and store
3624 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
3626 * 0 on success, or a negative error in case of failure.
3628 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
3630 * Use BTF to store a string representation of *ptr*->ptr in *str*,
3631 * using *ptr*->type_id. This value should specify the type
3632 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
3633 * can be used to look up vmlinux BTF type ids. Traversing the
3634 * data structure using BTF, the type information and values are
3635 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
3636 * the pointer data is carried out to avoid kernel crashes during
3637 * operation. Smaller types can use string space on the stack;
3638 * larger programs can use map data to store the string
3641 * The string can be subsequently shared with userspace via
3642 * bpf_perf_event_output() or ring buffer interfaces.
3643 * bpf_trace_printk() is to be avoided as it places too small
3644 * a limit on string size to be useful.
3646 * *flags* is a combination of
3649 * no formatting around type information
3651 * no struct/union member names/types
3653 * show raw (unobfuscated) pointer values;
3654 * equivalent to printk specifier %px.
3656 * show zero-valued struct/union members; they
3657 * are not displayed by default
3660 * The number of bytes that were written (or would have been
3661 * written if output had to be truncated due to string size),
3662 * or a negative error in cases of failure.
3664 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
3666 * Use BTF to write to seq_write a string representation of
3667 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
3668 * *flags* are identical to those used for bpf_snprintf_btf.
3670 * 0 on success or a negative error in case of failure.
3672 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
3674 * See **bpf_get_cgroup_classid**\ () for the main description.
3675 * This helper differs from **bpf_get_cgroup_classid**\ () in that
3676 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
3677 * associated socket instead of the current process.
3679 * The id is returned or 0 in case the id could not be retrieved.
3681 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
3683 * Redirect the packet to another net device of index *ifindex*
3684 * and fill in L2 addresses from neighboring subsystem. This helper
3685 * is somewhat similar to **bpf_redirect**\ (), except that it
3686 * populates L2 addresses as well, meaning, internally, the helper
3687 * relies on the neighbor lookup for the L2 address of the nexthop.
3689 * The helper will perform a FIB lookup based on the skb's
3690 * networking header to get the address of the next hop, unless
3691 * this is supplied by the caller in the *params* argument. The
3692 * *plen* argument indicates the len of *params* and should be set
3693 * to 0 if *params* is NULL.
3695 * The *flags* argument is reserved and must be 0. The helper is
3696 * currently only supported for tc BPF program types, and enabled
3697 * for IPv4 and IPv6 protocols.
3699 * The helper returns **TC_ACT_REDIRECT** on success or
3700 * **TC_ACT_SHOT** on error.
3702 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
3704 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
3705 * pointer to the percpu kernel variable on *cpu*. A ksym is an
3706 * extern variable decorated with '__ksym'. For ksym, there is a
3707 * global var (either static or global) defined of the same name
3708 * in the kernel. The ksym is percpu if the global var is percpu.
3709 * The returned pointer points to the global percpu var on *cpu*.
3711 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
3712 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
3713 * happens if *cpu* is larger than nr_cpu_ids. The caller of
3714 * bpf_per_cpu_ptr() must check the returned value.
3716 * A pointer pointing to the kernel percpu variable on *cpu*, or
3717 * NULL, if *cpu* is invalid.
3719 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
3721 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
3722 * pointer to the percpu kernel variable on this cpu. See the
3723 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
3725 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
3726 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
3727 * never return NULL.
3729 * A pointer pointing to the kernel percpu variable on this cpu.
3731 * long bpf_redirect_peer(u32 ifindex, u64 flags)
3733 * Redirect the packet to another net device of index *ifindex*.
3734 * This helper is somewhat similar to **bpf_redirect**\ (), except
3735 * that the redirection happens to the *ifindex*' peer device and
3736 * the netns switch takes place from ingress to ingress without
3737 * going through the CPU's backlog queue.
3739 * The *flags* argument is reserved and must be 0. The helper is
3740 * currently only supported for tc BPF program types at the ingress
3741 * hook and for veth device types. The peer device must reside in a
3742 * different network namespace.
3744 * The helper returns **TC_ACT_REDIRECT** on success or
3745 * **TC_ACT_SHOT** on error.
3747 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
3749 * Get a bpf_local_storage from the *task*.
3751 * Logically, it could be thought of as getting the value from
3752 * a *map* with *task* as the **key**. From this
3753 * perspective, the usage is not much different from
3754 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
3755 * helper enforces the key must be an task_struct and the map must also
3756 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
3758 * Underneath, the value is stored locally at *task* instead of
3759 * the *map*. The *map* is used as the bpf-local-storage
3760 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3761 * searched against all bpf_local_storage residing at *task*.
3763 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
3764 * used such that a new bpf_local_storage will be
3765 * created if one does not exist. *value* can be used
3766 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
3767 * the initial value of a bpf_local_storage. If *value* is
3768 * **NULL**, the new bpf_local_storage will be zero initialized.
3770 * A bpf_local_storage pointer is returned on success.
3772 * **NULL** if not found or there was an error in adding
3773 * a new bpf_local_storage.
3775 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
3777 * Delete a bpf_local_storage from a *task*.
3781 * **-ENOENT** if the bpf_local_storage cannot be found.
3783 * struct task_struct *bpf_get_current_task_btf(void)
3785 * Return a BTF pointer to the "current" task.
3786 * This pointer can also be used in helpers that accept an
3787 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
3789 * Pointer to the current task.
3791 #define __BPF_FUNC_MAPPER(FN) \
3793 FN(map_lookup_elem), \
3794 FN(map_update_elem), \
3795 FN(map_delete_elem), \
3799 FN(get_prandom_u32), \
3800 FN(get_smp_processor_id), \
3801 FN(skb_store_bytes), \
3802 FN(l3_csum_replace), \
3803 FN(l4_csum_replace), \
3805 FN(clone_redirect), \
3806 FN(get_current_pid_tgid), \
3807 FN(get_current_uid_gid), \
3808 FN(get_current_comm), \
3809 FN(get_cgroup_classid), \
3810 FN(skb_vlan_push), \
3812 FN(skb_get_tunnel_key), \
3813 FN(skb_set_tunnel_key), \
3814 FN(perf_event_read), \
3816 FN(get_route_realm), \
3817 FN(perf_event_output), \
3818 FN(skb_load_bytes), \
3821 FN(skb_get_tunnel_opt), \
3822 FN(skb_set_tunnel_opt), \
3823 FN(skb_change_proto), \
3824 FN(skb_change_type), \
3825 FN(skb_under_cgroup), \
3826 FN(get_hash_recalc), \
3827 FN(get_current_task), \
3828 FN(probe_write_user), \
3829 FN(current_task_under_cgroup), \
3830 FN(skb_change_tail), \
3831 FN(skb_pull_data), \
3833 FN(set_hash_invalid), \
3834 FN(get_numa_node_id), \
3835 FN(skb_change_head), \
3836 FN(xdp_adjust_head), \
3837 FN(probe_read_str), \
3838 FN(get_socket_cookie), \
3839 FN(get_socket_uid), \
3842 FN(skb_adjust_room), \
3844 FN(sk_redirect_map), \
3845 FN(sock_map_update), \
3846 FN(xdp_adjust_meta), \
3847 FN(perf_event_read_value), \
3848 FN(perf_prog_read_value), \
3850 FN(override_return), \
3851 FN(sock_ops_cb_flags_set), \
3852 FN(msg_redirect_map), \
3853 FN(msg_apply_bytes), \
3854 FN(msg_cork_bytes), \
3855 FN(msg_pull_data), \
3857 FN(xdp_adjust_tail), \
3858 FN(skb_get_xfrm_state), \
3860 FN(skb_load_bytes_relative), \
3862 FN(sock_hash_update), \
3863 FN(msg_redirect_hash), \
3864 FN(sk_redirect_hash), \
3865 FN(lwt_push_encap), \
3866 FN(lwt_seg6_store_bytes), \
3867 FN(lwt_seg6_adjust_srh), \
3868 FN(lwt_seg6_action), \
3871 FN(skb_cgroup_id), \
3872 FN(get_current_cgroup_id), \
3873 FN(get_local_storage), \
3874 FN(sk_select_reuseport), \
3875 FN(skb_ancestor_cgroup_id), \
3876 FN(sk_lookup_tcp), \
3877 FN(sk_lookup_udp), \
3879 FN(map_push_elem), \
3881 FN(map_peek_elem), \
3882 FN(msg_push_data), \
3884 FN(rc_pointer_rel), \
3889 FN(skb_ecn_set_ce), \
3890 FN(get_listener_sock), \
3891 FN(skc_lookup_tcp), \
3892 FN(tcp_check_syncookie), \
3893 FN(sysctl_get_name), \
3894 FN(sysctl_get_current_value), \
3895 FN(sysctl_get_new_value), \
3896 FN(sysctl_set_new_value), \
3899 FN(sk_storage_get), \
3900 FN(sk_storage_delete), \
3902 FN(tcp_gen_syncookie), \
3904 FN(probe_read_user), \
3905 FN(probe_read_kernel), \
3906 FN(probe_read_user_str), \
3907 FN(probe_read_kernel_str), \
3909 FN(send_signal_thread), \
3911 FN(read_branch_records), \
3912 FN(get_ns_current_pid_tgid), \
3914 FN(get_netns_cookie), \
3915 FN(get_current_ancestor_cgroup_id), \
3917 FN(ktime_get_boot_ns), \
3921 FN(sk_ancestor_cgroup_id), \
3922 FN(ringbuf_output), \
3923 FN(ringbuf_reserve), \
3924 FN(ringbuf_submit), \
3925 FN(ringbuf_discard), \
3926 FN(ringbuf_query), \
3928 FN(skc_to_tcp6_sock), \
3929 FN(skc_to_tcp_sock), \
3930 FN(skc_to_tcp_timewait_sock), \
3931 FN(skc_to_tcp_request_sock), \
3932 FN(skc_to_udp6_sock), \
3933 FN(get_task_stack), \
3935 FN(store_hdr_opt), \
3936 FN(reserve_hdr_opt), \
3937 FN(inode_storage_get), \
3938 FN(inode_storage_delete), \
3940 FN(copy_from_user), \
3942 FN(seq_printf_btf), \
3943 FN(skb_cgroup_classid), \
3944 FN(redirect_neigh), \
3945 FN(bpf_per_cpu_ptr), \
3946 FN(bpf_this_cpu_ptr), \
3947 FN(redirect_peer), \
3948 FN(task_storage_get), \
3949 FN(task_storage_delete), \
3950 FN(get_current_task_btf), \
3953 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
3954 * function eBPF program intends to call
3956 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
3958 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
3961 #undef __BPF_ENUM_FN
3963 /* All flags used by eBPF helper functions, placed here. */
3965 /* BPF_FUNC_skb_store_bytes flags. */
3967 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3968 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3971 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3972 * First 4 bits are for passing the header field size.
3975 BPF_F_HDR_FIELD_MASK = 0xfULL,
3978 /* BPF_FUNC_l4_csum_replace flags. */
3980 BPF_F_PSEUDO_HDR = (1ULL << 4),
3981 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3982 BPF_F_MARK_ENFORCE = (1ULL << 6),
3985 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
3987 BPF_F_INGRESS = (1ULL << 0),
3990 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
3992 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3995 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
3997 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3998 BPF_F_USER_STACK = (1ULL << 8),
3999 /* flags used by BPF_FUNC_get_stackid only. */
4000 BPF_F_FAST_STACK_CMP = (1ULL << 9),
4001 BPF_F_REUSE_STACKID = (1ULL << 10),
4002 /* flags used by BPF_FUNC_get_stack only. */
4003 BPF_F_USER_BUILD_ID = (1ULL << 11),
4006 /* BPF_FUNC_skb_set_tunnel_key flags. */
4008 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
4009 BPF_F_DONT_FRAGMENT = (1ULL << 2),
4010 BPF_F_SEQ_NUMBER = (1ULL << 3),
4013 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
4014 * BPF_FUNC_perf_event_read_value flags.
4017 BPF_F_INDEX_MASK = 0xffffffffULL,
4018 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
4019 /* BPF_FUNC_perf_event_output for sk_buff input context. */
4020 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
4023 /* Current network namespace */
4025 BPF_F_CURRENT_NETNS = (-1L),
4028 /* BPF_FUNC_csum_level level values. */
4030 BPF_CSUM_LEVEL_QUERY,
4033 BPF_CSUM_LEVEL_RESET,
4036 /* BPF_FUNC_skb_adjust_room flags. */
4038 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
4039 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
4040 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
4041 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
4042 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
4043 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
4047 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
4048 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
4051 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
4052 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
4053 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
4055 /* BPF_FUNC_sysctl_get_name flags. */
4057 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
4060 /* BPF_FUNC_<kernel_obj>_storage_get flags */
4062 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
4063 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
4064 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
4066 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
4069 /* BPF_FUNC_read_branch_records flags. */
4071 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
4074 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
4075 * BPF_FUNC_bpf_ringbuf_output flags.
4078 BPF_RB_NO_WAKEUP = (1ULL << 0),
4079 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
4082 /* BPF_FUNC_bpf_ringbuf_query flags */
4084 BPF_RB_AVAIL_DATA = 0,
4085 BPF_RB_RING_SIZE = 1,
4086 BPF_RB_CONS_POS = 2,
4087 BPF_RB_PROD_POS = 3,
4090 /* BPF ring buffer constants */
4092 BPF_RINGBUF_BUSY_BIT = (1U << 31),
4093 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
4094 BPF_RINGBUF_HDR_SZ = 8,
4097 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
4099 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
4100 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
4103 /* Mode for BPF_FUNC_skb_adjust_room helper. */
4104 enum bpf_adj_room_mode {
4109 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
4110 enum bpf_hdr_start_off {
4115 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
4116 enum bpf_lwt_encap_mode {
4118 BPF_LWT_ENCAP_SEG6_INLINE,
4122 #define __bpf_md_ptr(type, name) \
4126 } __attribute__((aligned(8)))
4128 /* user accessible mirror of in-kernel sk_buff.
4129 * new fields can only be added to the end of this structure
4135 __u32 queue_mapping;
4141 __u32 ingress_ifindex;
4151 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
4153 __u32 remote_ip4; /* Stored in network byte order */
4154 __u32 local_ip4; /* Stored in network byte order */
4155 __u32 remote_ip6[4]; /* Stored in network byte order */
4156 __u32 local_ip6[4]; /* Stored in network byte order */
4157 __u32 remote_port; /* Stored in network byte order */
4158 __u32 local_port; /* stored in host byte order */
4162 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
4166 __bpf_md_ptr(struct bpf_sock *, sk);
4170 struct bpf_tunnel_key {
4174 __u32 remote_ipv6[4];
4178 __u16 tunnel_ext; /* Padding, future use. */
4182 /* user accessible mirror of in-kernel xfrm_state.
4183 * new fields can only be added to the end of this structure
4185 struct bpf_xfrm_state {
4187 __u32 spi; /* Stored in network byte order */
4189 __u16 ext; /* Padding, future use. */
4191 __u32 remote_ipv4; /* Stored in network byte order */
4192 __u32 remote_ipv6[4]; /* Stored in network byte order */
4196 /* Generic BPF return codes which all BPF program types may support.
4197 * The values are binary compatible with their TC_ACT_* counter-part to
4198 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
4201 * XDP is handled seprately, see XDP_*.
4209 /* >127 are reserved for prog type specific return codes.
4211 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
4212 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
4213 * changed and should be routed based on its new L3 header.
4214 * (This is an L3 redirect, as opposed to L2 redirect
4215 * represented by BPF_REDIRECT above).
4217 BPF_LWT_REROUTE = 128,
4227 /* IP address also allows 1 and 2 bytes access */
4230 __u32 src_port; /* host byte order */
4231 __u32 dst_port; /* network byte order */
4235 __s32 rx_queue_mapping;
4238 struct bpf_tcp_sock {
4239 __u32 snd_cwnd; /* Sending congestion window */
4240 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
4242 __u32 snd_ssthresh; /* Slow start size threshold */
4243 __u32 rcv_nxt; /* What we want to receive next */
4244 __u32 snd_nxt; /* Next sequence we send */
4245 __u32 snd_una; /* First byte we want an ack for */
4246 __u32 mss_cache; /* Cached effective mss, not including SACKS */
4247 __u32 ecn_flags; /* ECN status bits. */
4248 __u32 rate_delivered; /* saved rate sample: packets delivered */
4249 __u32 rate_interval_us; /* saved rate sample: time elapsed */
4250 __u32 packets_out; /* Packets which are "in flight" */
4251 __u32 retrans_out; /* Retransmitted packets out */
4252 __u32 total_retrans; /* Total retransmits for entire connection */
4253 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
4254 * total number of segments in.
4256 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
4257 * total number of data segments in.
4259 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
4260 * The total number of segments sent.
4262 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
4263 * total number of data segments sent.
4265 __u32 lost_out; /* Lost packets */
4266 __u32 sacked_out; /* SACK'd packets */
4267 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
4268 * sum(delta(rcv_nxt)), or how many bytes
4271 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
4272 * sum(delta(snd_una)), or how many bytes
4275 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
4276 * total number of DSACK blocks received
4278 __u32 delivered; /* Total data packets delivered incl. rexmits */
4279 __u32 delivered_ce; /* Like the above but only ECE marked packets */
4280 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
4283 struct bpf_sock_tuple {
4300 struct bpf_xdp_sock {
4304 #define XDP_PACKET_HEADROOM 256
4306 /* User return codes for XDP prog type.
4307 * A valid XDP program must return one of these defined values. All other
4308 * return codes are reserved for future use. Unknown return codes will
4309 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
4319 /* user accessible metadata for XDP packet hook
4320 * new fields must be added to the end of this structure
4326 /* Below access go through struct xdp_rxq_info */
4327 __u32 ingress_ifindex; /* rxq->dev->ifindex */
4328 __u32 rx_queue_index; /* rxq->queue_index */
4330 __u32 egress_ifindex; /* txq->dev->ifindex */
4333 /* DEVMAP map-value layout
4335 * The struct data-layout of map-value is a configuration interface.
4336 * New members can only be added to the end of this structure.
4338 struct bpf_devmap_val {
4339 __u32 ifindex; /* device index */
4341 int fd; /* prog fd on map write */
4342 __u32 id; /* prog id on map read */
4346 /* CPUMAP map-value layout
4348 * The struct data-layout of map-value is a configuration interface.
4349 * New members can only be added to the end of this structure.
4351 struct bpf_cpumap_val {
4352 __u32 qsize; /* queue size to remote target CPU */
4354 int fd; /* prog fd on map write */
4355 __u32 id; /* prog id on map read */
4364 /* user accessible metadata for SK_MSG packet hook, new fields must
4365 * be added to the end of this structure
4368 __bpf_md_ptr(void *, data);
4369 __bpf_md_ptr(void *, data_end);
4372 __u32 remote_ip4; /* Stored in network byte order */
4373 __u32 local_ip4; /* Stored in network byte order */
4374 __u32 remote_ip6[4]; /* Stored in network byte order */
4375 __u32 local_ip6[4]; /* Stored in network byte order */
4376 __u32 remote_port; /* Stored in network byte order */
4377 __u32 local_port; /* stored in host byte order */
4378 __u32 size; /* Total size of sk_msg */
4380 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
4383 struct sk_reuseport_md {
4385 * Start of directly accessible data. It begins from
4386 * the tcp/udp header.
4388 __bpf_md_ptr(void *, data);
4389 /* End of directly accessible data */
4390 __bpf_md_ptr(void *, data_end);
4392 * Total length of packet (starting from the tcp/udp header).
4393 * Note that the directly accessible bytes (data_end - data)
4394 * could be less than this "len". Those bytes could be
4395 * indirectly read by a helper "bpf_skb_load_bytes()".
4399 * Eth protocol in the mac header (network byte order). e.g.
4400 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
4403 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
4404 __u32 bind_inany; /* Is sock bound to an INANY address? */
4405 __u32 hash; /* A hash of the packet 4 tuples */
4408 #define BPF_TAG_SIZE 8
4410 struct bpf_prog_info {
4413 __u8 tag[BPF_TAG_SIZE];
4414 __u32 jited_prog_len;
4415 __u32 xlated_prog_len;
4416 __aligned_u64 jited_prog_insns;
4417 __aligned_u64 xlated_prog_insns;
4418 __u64 load_time; /* ns since boottime */
4419 __u32 created_by_uid;
4421 __aligned_u64 map_ids;
4422 char name[BPF_OBJ_NAME_LEN];
4424 __u32 gpl_compatible:1;
4425 __u32 :31; /* alignment pad */
4428 __u32 nr_jited_ksyms;
4429 __u32 nr_jited_func_lens;
4430 __aligned_u64 jited_ksyms;
4431 __aligned_u64 jited_func_lens;
4433 __u32 func_info_rec_size;
4434 __aligned_u64 func_info;
4437 __aligned_u64 line_info;
4438 __aligned_u64 jited_line_info;
4439 __u32 nr_jited_line_info;
4440 __u32 line_info_rec_size;
4441 __u32 jited_line_info_rec_size;
4443 __aligned_u64 prog_tags;
4446 } __attribute__((aligned(8)));
4448 struct bpf_map_info {
4455 char name[BPF_OBJ_NAME_LEN];
4457 __u32 btf_vmlinux_value_type_id;
4461 __u32 btf_key_type_id;
4462 __u32 btf_value_type_id;
4463 } __attribute__((aligned(8)));
4465 struct bpf_btf_info {
4472 } __attribute__((aligned(8)));
4474 struct bpf_link_info {
4480 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
4481 __u32 tp_name_len; /* in/out: tp_name buffer len */
4491 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
4492 __u32 target_name_len; /* in/out: target_name buffer len */
4507 } __attribute__((aligned(8)));
4509 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
4510 * by user and intended to be used by socket (e.g. to bind to, depends on
4513 struct bpf_sock_addr {
4514 __u32 user_family; /* Allows 4-byte read, but no write. */
4515 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4516 * Stored in network byte order.
4518 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4519 * Stored in network byte order.
4521 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
4522 * Stored in network byte order
4524 __u32 family; /* Allows 4-byte read, but no write */
4525 __u32 type; /* Allows 4-byte read, but no write */
4526 __u32 protocol; /* Allows 4-byte read, but no write */
4527 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4528 * Stored in network byte order.
4530 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4531 * Stored in network byte order.
4533 __bpf_md_ptr(struct bpf_sock *, sk);
4536 /* User bpf_sock_ops struct to access socket values and specify request ops
4537 * and their replies.
4538 * Some of this fields are in network (bigendian) byte order and may need
4539 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
4540 * New fields can only be added at the end of this structure
4542 struct bpf_sock_ops {
4545 __u32 args[4]; /* Optionally passed to bpf program */
4546 __u32 reply; /* Returned by bpf program */
4547 __u32 replylong[4]; /* Optionally returned by bpf prog */
4550 __u32 remote_ip4; /* Stored in network byte order */
4551 __u32 local_ip4; /* Stored in network byte order */
4552 __u32 remote_ip6[4]; /* Stored in network byte order */
4553 __u32 local_ip6[4]; /* Stored in network byte order */
4554 __u32 remote_port; /* Stored in network byte order */
4555 __u32 local_port; /* stored in host byte order */
4556 __u32 is_fullsock; /* Some TCP fields are only valid if
4557 * there is a full socket. If not, the
4558 * fields read as zero.
4561 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
4562 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
4571 __u32 rate_delivered;
4572 __u32 rate_interval_us;
4575 __u32 total_retrans;
4579 __u32 data_segs_out;
4583 __u64 bytes_received;
4585 __bpf_md_ptr(struct bpf_sock *, sk);
4586 /* [skb_data, skb_data_end) covers the whole TCP header.
4588 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
4589 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
4590 * header has not been written.
4591 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
4592 * been written so far.
4593 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
4595 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
4598 * bpf_load_hdr_opt() can also be used to read a particular option.
4600 __bpf_md_ptr(void *, skb_data);
4601 __bpf_md_ptr(void *, skb_data_end);
4602 __u32 skb_len; /* The total length of a packet.
4603 * It includes the header, options,
4606 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
4607 * an easy way to check for tcp_flags
4608 * without parsing skb_data.
4610 * In particular, the skb_tcp_flags
4611 * will still be available in
4612 * BPF_SOCK_OPS_HDR_OPT_LEN even though
4613 * the outgoing header has not
4618 /* Definitions for bpf_sock_ops_cb_flags */
4620 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
4621 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
4622 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
4623 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
4624 /* Call bpf for all received TCP headers. The bpf prog will be
4625 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4627 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4628 * for the header option related helpers that will be useful
4629 * to the bpf programs.
4631 * It could be used at the client/active side (i.e. connect() side)
4632 * when the server told it that the server was in syncookie
4633 * mode and required the active side to resend the bpf-written
4634 * options. The active side can keep writing the bpf-options until
4635 * it received a valid packet from the server side to confirm
4636 * the earlier packet (and options) has been received. The later
4637 * example patch is using it like this at the active side when the
4638 * server is in syncookie mode.
4640 * The bpf prog will usually turn this off in the common cases.
4642 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
4643 /* Call bpf when kernel has received a header option that
4644 * the kernel cannot handle. The bpf prog will be called under
4645 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
4647 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4648 * for the header option related helpers that will be useful
4649 * to the bpf programs.
4651 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
4652 /* Call bpf when the kernel is writing header options for the
4653 * outgoing packet. The bpf prog will first be called
4654 * to reserve space in a skb under
4655 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
4656 * the bpf prog will be called to write the header option(s)
4657 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4659 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
4660 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
4661 * related helpers that will be useful to the bpf programs.
4663 * The kernel gets its chance to reserve space and write
4664 * options first before the BPF program does.
4666 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
4667 /* Mask of all currently supported cb flags */
4668 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
4671 /* List of known BPF sock_ops operators.
4672 * New entries can only be added at the end
4676 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
4677 * -1 if default value should be used
4679 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
4680 * window (in packets) or -1 if default
4681 * value should be used
4683 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
4684 * active connection is initialized
4686 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
4687 * active connection is
4690 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
4691 * passive connection is
4694 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
4697 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
4698 * based on the path and may be
4699 * dependent on the congestion control
4700 * algorithm. In general it indicates
4701 * a congestion threshold. RTTs above
4702 * this indicate congestion
4704 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
4705 * Arg1: value of icsk_retransmits
4706 * Arg2: value of icsk_rto
4707 * Arg3: whether RTO has expired
4709 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
4710 * Arg1: sequence number of 1st byte
4712 * Arg3: return value of
4713 * tcp_transmit_skb (0 => success)
4715 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
4719 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
4720 * socket transition to LISTEN state.
4722 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
4724 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
4725 * It will be called to handle
4726 * the packets received at
4727 * an already established
4730 * sock_ops->skb_data:
4731 * Referring to the received skb.
4732 * It covers the TCP header only.
4734 * bpf_load_hdr_opt() can also
4735 * be used to search for a
4736 * particular option.
4738 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
4739 * header option later in
4740 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4741 * Arg1: bool want_cookie. (in
4742 * writing SYNACK only)
4744 * sock_ops->skb_data:
4745 * Not available because no header has
4748 * sock_ops->skb_tcp_flags:
4749 * The tcp_flags of the
4750 * outgoing skb. (e.g. SYN, ACK, FIN).
4752 * bpf_reserve_hdr_opt() should
4753 * be used to reserve space.
4755 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
4756 * Arg1: bool want_cookie. (in
4757 * writing SYNACK only)
4759 * sock_ops->skb_data:
4760 * Referring to the outgoing skb.
4761 * It covers the TCP header
4762 * that has already been written
4763 * by the kernel and the
4764 * earlier bpf-progs.
4766 * sock_ops->skb_tcp_flags:
4767 * The tcp_flags of the outgoing
4768 * skb. (e.g. SYN, ACK, FIN).
4770 * bpf_store_hdr_opt() should
4771 * be used to write the
4774 * bpf_load_hdr_opt() can also
4775 * be used to search for a
4776 * particular option that
4777 * has already been written
4778 * by the kernel or the
4779 * earlier bpf-progs.
4783 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
4784 * changes between the TCP and BPF versions. Ideally this should never happen.
4785 * If it does, we need to add code to convert them before calling
4786 * the BPF sock_ops function.
4789 BPF_TCP_ESTABLISHED = 1,
4799 BPF_TCP_CLOSING, /* Now a valid state */
4800 BPF_TCP_NEW_SYN_RECV,
4802 BPF_TCP_MAX_STATES /* Leave at the end! */
4806 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4807 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
4808 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
4809 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
4810 /* Copy the SYN pkt to optval
4812 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
4813 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
4814 * to only getting from the saved_syn. It can either get the
4817 * 1. the just-received SYN packet (only available when writing the
4818 * SYNACK). It will be useful when it is not necessary to
4819 * save the SYN packet for latter use. It is also the only way
4820 * to get the SYN during syncookie mode because the syn
4821 * packet cannot be saved during syncookie.
4825 * 2. the earlier saved syn which was done by
4826 * bpf_setsockopt(TCP_SAVE_SYN).
4828 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
4829 * SYN packet is obtained.
4831 * If the bpf-prog does not need the IP[46] header, the
4832 * bpf-prog can avoid parsing the IP header by using
4833 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
4834 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
4836 * >0: Total number of bytes copied
4837 * -ENOSPC: Not enough space in optval. Only optlen number of
4839 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
4840 * is not saved by setsockopt(TCP_SAVE_SYN).
4842 TCP_BPF_SYN = 1005, /* Copy the TCP header */
4843 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
4844 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
4848 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
4851 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
4852 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4855 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
4856 * total option spaces
4857 * required for an established
4858 * sk in order to calculate the
4859 * MSS. No skb is actually
4862 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
4863 * when sending a SYN.
4867 struct bpf_perf_event_value {
4874 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
4875 BPF_DEVCG_ACC_READ = (1ULL << 1),
4876 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
4880 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
4881 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
4884 struct bpf_cgroup_dev_ctx {
4885 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
4891 struct bpf_raw_tracepoint_args {
4895 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
4896 * OUTPUT: Do lookup from egress perspective; default is ingress
4899 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
4900 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
4904 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
4905 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
4906 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
4907 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
4908 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
4909 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
4910 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
4911 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
4912 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
4915 struct bpf_fib_lookup {
4916 /* input: network family for lookup (AF_INET, AF_INET6)
4917 * output: network family of egress nexthop
4921 /* set if lookup is to consider L4 data - e.g., FIB rules */
4926 /* total length of packet from network header - used for MTU check */
4929 /* input: L3 device index for lookup
4930 * output: device index from FIB lookup
4935 /* inputs to lookup */
4936 __u8 tos; /* AF_INET */
4937 __be32 flowinfo; /* AF_INET6, flow_label + priority */
4939 /* output: metric of fib result (IPv4/IPv6 only) */
4945 __u32 ipv6_src[4]; /* in6_addr; network order */
4948 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
4949 * network header. output: bpf_fib_lookup sets to gateway address
4950 * if FIB lookup returns gateway route
4954 __u32 ipv6_dst[4]; /* in6_addr; network order */
4958 __be16 h_vlan_proto;
4960 __u8 smac[6]; /* ETH_ALEN */
4961 __u8 dmac[6]; /* ETH_ALEN */
4964 struct bpf_redir_neigh {
4965 /* network family for lookup (AF_INET, AF_INET6) */
4967 /* network address of nexthop; skips fib lookup to find gateway */
4970 __u32 ipv6_nh[4]; /* in6_addr; network order */
4974 enum bpf_task_fd_type {
4975 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
4976 BPF_FD_TYPE_TRACEPOINT, /* tp name */
4977 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
4978 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
4979 BPF_FD_TYPE_UPROBE, /* filename + offset */
4980 BPF_FD_TYPE_URETPROBE, /* filename + offset */
4984 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
4985 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
4986 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
4989 struct bpf_flow_keys {
4992 __u16 addr_proto; /* ETH_P_* of valid addrs */
5006 __u32 ipv6_src[4]; /* in6_addr; network order */
5007 __u32 ipv6_dst[4]; /* in6_addr; network order */
5014 struct bpf_func_info {
5019 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
5020 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
5022 struct bpf_line_info {
5024 __u32 file_name_off;
5029 struct bpf_spin_lock {
5034 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
5035 * Allows 1,2,4-byte read, but no write.
5037 __u32 file_pos; /* Sysctl file position to read from, write to.
5038 * Allows 1,2,4-byte read an 4-byte write.
5042 struct bpf_sockopt {
5043 __bpf_md_ptr(struct bpf_sock *, sk);
5044 __bpf_md_ptr(void *, optval);
5045 __bpf_md_ptr(void *, optval_end);
5053 struct bpf_pidns_info {
5058 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
5059 struct bpf_sk_lookup {
5060 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
5062 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
5063 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
5064 __u32 remote_ip4; /* Network byte order */
5065 __u32 remote_ip6[4]; /* Network byte order */
5066 __u32 remote_port; /* Network byte order */
5067 __u32 local_ip4; /* Network byte order */
5068 __u32 local_ip6[4]; /* Network byte order */
5069 __u32 local_port; /* Host byte order */
5073 * struct btf_ptr is used for typed pointer representation; the
5074 * type id is used to render the pointer data as the appropriate type
5075 * via the bpf_snprintf_btf() helper described above. A flags field -
5076 * potentially to specify additional details about the BTF pointer
5077 * (rather than its mode of display) - is included for future use.
5078 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
5083 __u32 flags; /* BTF ptr flags; unused at present. */
5087 * Flags to control bpf_snprintf_btf() behaviour.
5088 * - BTF_F_COMPACT: no formatting around type information
5089 * - BTF_F_NONAME: no struct/union member names/types
5090 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
5091 * equivalent to %px.
5092 * - BTF_F_ZERO: show zero-valued struct/union members; they
5093 * are not displayed by default
5096 BTF_F_COMPACT = (1ULL << 0),
5097 BTF_F_NONAME = (1ULL << 1),
5098 BTF_F_PTR_RAW = (1ULL << 2),
5099 BTF_F_ZERO = (1ULL << 3),
5102 #endif /* _UAPI__LINUX_BPF_H__ */