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 __LINUX_BPF_H__
9 #define __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_ALU64 0x07 /* alu mode in double word width */
20 #define BPF_DW 0x18 /* double word (64-bit) */
21 #define BPF_XADD 0xc0 /* exclusive add */
24 #define BPF_MOV 0xb0 /* mov reg to reg */
25 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
27 /* change endianness of a register */
28 #define BPF_END 0xd0 /* flags for endianness conversion: */
29 #define BPF_TO_LE 0x00 /* convert to little-endian */
30 #define BPF_TO_BE 0x08 /* convert to big-endian */
31 #define BPF_FROM_LE BPF_TO_LE
32 #define BPF_FROM_BE BPF_TO_BE
35 #define BPF_JNE 0x50 /* jump != */
36 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
37 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
38 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
39 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
40 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
41 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
42 #define BPF_CALL 0x80 /* function call */
43 #define BPF_EXIT 0x90 /* function return */
45 /* Register numbers */
61 /* BPF has 10 general purpose 64-bit registers and stack frame. */
62 #define MAX_BPF_REG __MAX_BPF_REG
65 __u8 code; /* opcode */
66 __u8 dst_reg:4; /* dest register */
67 __u8 src_reg:4; /* source register */
68 __s16 off; /* signed offset */
69 __s32 imm; /* signed immediate constant */
72 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
73 struct bpf_lpm_trie_key {
74 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
75 __u8 data[0]; /* Arbitrary size */
78 struct bpf_cgroup_storage_key {
79 __u64 cgroup_inode_id; /* cgroup inode id */
80 __u32 attach_type; /* program attach type */
83 /* BPF syscall commands, see bpf(2) man-page for details. */
98 BPF_PROG_GET_FD_BY_ID,
100 BPF_OBJ_GET_INFO_BY_FD,
102 BPF_RAW_TRACEPOINT_OPEN,
104 BPF_BTF_GET_FD_BY_ID,
112 BPF_MAP_TYPE_PROG_ARRAY,
113 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
114 BPF_MAP_TYPE_PERCPU_HASH,
115 BPF_MAP_TYPE_PERCPU_ARRAY,
116 BPF_MAP_TYPE_STACK_TRACE,
117 BPF_MAP_TYPE_CGROUP_ARRAY,
118 BPF_MAP_TYPE_LRU_HASH,
119 BPF_MAP_TYPE_LRU_PERCPU_HASH,
120 BPF_MAP_TYPE_LPM_TRIE,
121 BPF_MAP_TYPE_ARRAY_OF_MAPS,
122 BPF_MAP_TYPE_HASH_OF_MAPS,
124 BPF_MAP_TYPE_SOCKMAP,
127 BPF_MAP_TYPE_SOCKHASH,
128 BPF_MAP_TYPE_CGROUP_STORAGE,
129 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
133 BPF_PROG_TYPE_UNSPEC,
134 BPF_PROG_TYPE_SOCKET_FILTER,
135 BPF_PROG_TYPE_KPROBE,
136 BPF_PROG_TYPE_SCHED_CLS,
137 BPF_PROG_TYPE_SCHED_ACT,
138 BPF_PROG_TYPE_TRACEPOINT,
140 BPF_PROG_TYPE_PERF_EVENT,
141 BPF_PROG_TYPE_CGROUP_SKB,
142 BPF_PROG_TYPE_CGROUP_SOCK,
143 BPF_PROG_TYPE_LWT_IN,
144 BPF_PROG_TYPE_LWT_OUT,
145 BPF_PROG_TYPE_LWT_XMIT,
146 BPF_PROG_TYPE_SOCK_OPS,
147 BPF_PROG_TYPE_SK_SKB,
148 BPF_PROG_TYPE_CGROUP_DEVICE,
149 BPF_PROG_TYPE_SK_MSG,
150 BPF_PROG_TYPE_RAW_TRACEPOINT,
151 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
152 BPF_PROG_TYPE_LWT_SEG6LOCAL,
153 BPF_PROG_TYPE_LIRC_MODE2,
154 BPF_PROG_TYPE_SK_REUSEPORT,
157 enum bpf_attach_type {
158 BPF_CGROUP_INET_INGRESS,
159 BPF_CGROUP_INET_EGRESS,
160 BPF_CGROUP_INET_SOCK_CREATE,
162 BPF_SK_SKB_STREAM_PARSER,
163 BPF_SK_SKB_STREAM_VERDICT,
166 BPF_CGROUP_INET4_BIND,
167 BPF_CGROUP_INET6_BIND,
168 BPF_CGROUP_INET4_CONNECT,
169 BPF_CGROUP_INET6_CONNECT,
170 BPF_CGROUP_INET4_POST_BIND,
171 BPF_CGROUP_INET6_POST_BIND,
172 BPF_CGROUP_UDP4_SENDMSG,
173 BPF_CGROUP_UDP6_SENDMSG,
175 BPF_CGROUP_UDP4_RECVMSG = 19,
176 BPF_CGROUP_UDP6_RECVMSG,
177 __MAX_BPF_ATTACH_TYPE
180 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
182 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
184 * NONE(default): No further bpf programs allowed in the subtree.
186 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
187 * the program in this cgroup yields to sub-cgroup program.
189 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
190 * that cgroup program gets run in addition to the program in this cgroup.
192 * Only one program is allowed to be attached to a cgroup with
193 * NONE or BPF_F_ALLOW_OVERRIDE flag.
194 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
195 * release old program and attach the new one. Attach flags has to match.
197 * Multiple programs are allowed to be attached to a cgroup with
198 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
199 * (those that were attached first, run first)
200 * The programs of sub-cgroup are executed first, then programs of
201 * this cgroup and then programs of parent cgroup.
202 * When children program makes decision (like picking TCP CA or sock bind)
203 * parent program has a chance to override it.
205 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
206 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
208 * cgrp1 (MULTI progs A, B) ->
209 * cgrp2 (OVERRIDE prog C) ->
210 * cgrp3 (MULTI prog D) ->
211 * cgrp4 (OVERRIDE prog E) ->
212 * cgrp5 (NONE prog F)
213 * the event in cgrp5 triggers execution of F,D,A,B in that order.
214 * if prog F is detached, the execution is E,D,A,B
215 * if prog F and D are detached, the execution is E,A,B
216 * if prog F, E and D are detached, the execution is C,A,B
218 * All eligible programs are executed regardless of return code from
221 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
222 #define BPF_F_ALLOW_MULTI (1U << 1)
224 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
225 * verifier will perform strict alignment checking as if the kernel
226 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
227 * and NET_IP_ALIGN defined to 2.
229 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
231 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
232 #define BPF_PSEUDO_MAP_FD 1
234 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
235 * offset to another bpf function
237 #define BPF_PSEUDO_CALL 1
239 /* flags for BPF_MAP_UPDATE_ELEM command */
240 #define BPF_ANY 0 /* create new element or update existing */
241 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
242 #define BPF_EXIST 2 /* update existing element */
244 /* flags for BPF_MAP_CREATE command */
245 #define BPF_F_NO_PREALLOC (1U << 0)
246 /* Instead of having one common LRU list in the
247 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
248 * which can scale and perform better.
249 * Note, the LRU nodes (including free nodes) cannot be moved
250 * across different LRU lists.
252 #define BPF_F_NO_COMMON_LRU (1U << 1)
253 /* Specify numa node during map creation */
254 #define BPF_F_NUMA_NODE (1U << 2)
256 /* flags for BPF_PROG_QUERY */
257 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
259 #define BPF_OBJ_NAME_LEN 16U
261 /* Flags for accessing BPF object */
262 #define BPF_F_RDONLY (1U << 3)
263 #define BPF_F_WRONLY (1U << 4)
265 /* Flag for stack_map, store build_id+offset instead of pointer */
266 #define BPF_F_STACK_BUILD_ID (1U << 5)
268 enum bpf_stack_build_id_status {
269 /* user space need an empty entry to identify end of a trace */
270 BPF_STACK_BUILD_ID_EMPTY = 0,
271 /* with valid build_id and offset */
272 BPF_STACK_BUILD_ID_VALID = 1,
273 /* couldn't get build_id, fallback to ip */
274 BPF_STACK_BUILD_ID_IP = 2,
277 #define BPF_BUILD_ID_SIZE 20
278 struct bpf_stack_build_id {
280 unsigned char build_id[BPF_BUILD_ID_SIZE];
288 struct { /* anonymous struct used by BPF_MAP_CREATE command */
289 __u32 map_type; /* one of enum bpf_map_type */
290 __u32 key_size; /* size of key in bytes */
291 __u32 value_size; /* size of value in bytes */
292 __u32 max_entries; /* max number of entries in a map */
293 __u32 map_flags; /* BPF_MAP_CREATE related
294 * flags defined above.
296 __u32 inner_map_fd; /* fd pointing to the inner map */
297 __u32 numa_node; /* numa node (effective only if
298 * BPF_F_NUMA_NODE is set).
300 char map_name[BPF_OBJ_NAME_LEN];
301 __u32 map_ifindex; /* ifindex of netdev to create on */
302 __u32 btf_fd; /* fd pointing to a BTF type data */
303 __u32 btf_key_type_id; /* BTF type_id of the key */
304 __u32 btf_value_type_id; /* BTF type_id of the value */
307 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
312 __aligned_u64 next_key;
317 struct { /* anonymous struct used by BPF_PROG_LOAD command */
318 __u32 prog_type; /* one of enum bpf_prog_type */
321 __aligned_u64 license;
322 __u32 log_level; /* verbosity level of verifier */
323 __u32 log_size; /* size of user buffer */
324 __aligned_u64 log_buf; /* user supplied buffer */
325 __u32 kern_version; /* checked when prog_type=kprobe */
327 char prog_name[BPF_OBJ_NAME_LEN];
328 __u32 prog_ifindex; /* ifindex of netdev to prep for */
329 /* For some prog types expected attach type must be known at
330 * load time to verify attach type specific parts of prog
331 * (context accesses, allowed helpers, etc).
333 __u32 expected_attach_type;
336 struct { /* anonymous struct used by BPF_OBJ_* commands */
337 __aligned_u64 pathname;
342 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
343 __u32 target_fd; /* container object to attach to */
344 __u32 attach_bpf_fd; /* eBPF program to attach */
349 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
354 __aligned_u64 data_in;
355 __aligned_u64 data_out;
360 struct { /* anonymous struct used by BPF_*_GET_*_ID */
371 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
377 struct { /* anonymous struct used by BPF_PROG_QUERY command */
378 __u32 target_fd; /* container object to query */
382 __aligned_u64 prog_ids;
391 struct { /* anonymous struct for BPF_BTF_LOAD */
393 __aligned_u64 btf_log_buf;
400 __u32 pid; /* input: pid */
401 __u32 fd; /* input: fd */
402 __u32 flags; /* input: flags */
403 __u32 buf_len; /* input/output: buf len */
404 __aligned_u64 buf; /* input/output:
405 * tp_name for tracepoint
407 * filename for uprobe
409 __u32 prog_id; /* output: prod_id */
410 __u32 fd_type; /* output: BPF_FD_TYPE_* */
411 __u64 probe_offset; /* output: probe_offset */
412 __u64 probe_addr; /* output: probe_addr */
414 } __attribute__((aligned(8)));
416 /* The description below is an attempt at providing documentation to eBPF
417 * developers about the multiple available eBPF helper functions. It can be
418 * parsed and used to produce a manual page. The workflow is the following,
419 * and requires the rst2man utility:
421 * $ ./scripts/bpf_helpers_doc.py \
422 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
423 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
424 * $ man /tmp/bpf-helpers.7
426 * Note that in order to produce this external documentation, some RST
427 * formatting is used in the descriptions to get "bold" and "italics" in
428 * manual pages. Also note that the few trailing white spaces are
429 * intentional, removing them would break paragraphs for rst2man.
431 * Start of BPF helper function descriptions:
433 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
435 * Perform a lookup in *map* for an entry associated to *key*.
437 * Map value associated to *key*, or **NULL** if no entry was
440 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
442 * Add or update the value of the entry associated to *key* in
443 * *map* with *value*. *flags* is one of:
446 * The entry for *key* must not exist in the map.
448 * The entry for *key* must already exist in the map.
450 * No condition on the existence of the entry for *key*.
452 * Flag value **BPF_NOEXIST** cannot be used for maps of types
453 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
454 * elements always exist), the helper would return an error.
456 * 0 on success, or a negative error in case of failure.
458 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
460 * Delete entry with *key* from *map*.
462 * 0 on success, or a negative error in case of failure.
464 * int bpf_probe_read(void *dst, u32 size, const void *src)
466 * For tracing programs, safely attempt to read *size* bytes from
467 * address *src* and store the data in *dst*.
469 * 0 on success, or a negative error in case of failure.
471 * u64 bpf_ktime_get_ns(void)
473 * Return the time elapsed since system boot, in nanoseconds.
477 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
479 * This helper is a "printk()-like" facility for debugging. It
480 * prints a message defined by format *fmt* (of size *fmt_size*)
481 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
482 * available. It can take up to three additional **u64**
483 * arguments (as an eBPF helpers, the total number of arguments is
486 * Each time the helper is called, it appends a line to the trace.
487 * The format of the trace is customizable, and the exact output
488 * one will get depends on the options set in
489 * *\/sys/kernel/debug/tracing/trace_options* (see also the
490 * *README* file under the same directory). However, it usually
491 * defaults to something like:
495 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
499 * * ``telnet`` is the name of the current task.
500 * * ``470`` is the PID of the current task.
501 * * ``001`` is the CPU number on which the task is
503 * * In ``.N..``, each character refers to a set of
504 * options (whether irqs are enabled, scheduling
505 * options, whether hard/softirqs are running, level of
506 * preempt_disabled respectively). **N** means that
507 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
509 * * ``419421.045894`` is a timestamp.
510 * * ``0x00000001`` is a fake value used by BPF for the
511 * instruction pointer register.
512 * * ``<formatted msg>`` is the message formatted with
515 * The conversion specifiers supported by *fmt* are similar, but
516 * more limited than for printk(). They are **%d**, **%i**,
517 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
518 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
519 * of field, padding with zeroes, etc.) is available, and the
520 * helper will return **-EINVAL** (but print nothing) if it
521 * encounters an unknown specifier.
523 * Also, note that **bpf_trace_printk**\ () is slow, and should
524 * only be used for debugging purposes. For this reason, a notice
525 * bloc (spanning several lines) is printed to kernel logs and
526 * states that the helper should not be used "for production use"
527 * the first time this helper is used (or more precisely, when
528 * **trace_printk**\ () buffers are allocated). For passing values
529 * to user space, perf events should be preferred.
531 * The number of bytes written to the buffer, or a negative error
532 * in case of failure.
534 * u32 bpf_get_prandom_u32(void)
536 * Get a pseudo-random number.
538 * From a security point of view, this helper uses its own
539 * pseudo-random internal state, and cannot be used to infer the
540 * seed of other random functions in the kernel. However, it is
541 * essential to note that the generator used by the helper is not
542 * cryptographically secure.
544 * A random 32-bit unsigned value.
546 * u32 bpf_get_smp_processor_id(void)
548 * Get the SMP (symmetric multiprocessing) processor id. Note that
549 * all programs run with preemption disabled, which means that the
550 * SMP processor id is stable during all the execution of the
553 * The SMP id of the processor running the program.
555 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
557 * Store *len* bytes from address *from* into the packet
558 * associated to *skb*, at *offset*. *flags* are a combination of
559 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
560 * checksum for the packet after storing the bytes) and
561 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
562 * **->swhash** and *skb*\ **->l4hash** to 0).
564 * A call to this helper is susceptible to change the underlaying
565 * packet buffer. Therefore, at load time, all checks on pointers
566 * previously done by the verifier are invalidated and must be
567 * performed again, if the helper is used in combination with
568 * direct packet access.
570 * 0 on success, or a negative error in case of failure.
572 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
574 * Recompute the layer 3 (e.g. IP) checksum for the packet
575 * associated to *skb*. Computation is incremental, so the helper
576 * must know the former value of the header field that was
577 * modified (*from*), the new value of this field (*to*), and the
578 * number of bytes (2 or 4) for this field, stored in *size*.
579 * Alternatively, it is possible to store the difference between
580 * the previous and the new values of the header field in *to*, by
581 * setting *from* and *size* to 0. For both methods, *offset*
582 * indicates the location of the IP checksum within the packet.
584 * This helper works in combination with **bpf_csum_diff**\ (),
585 * which does not update the checksum in-place, but offers more
586 * flexibility and can handle sizes larger than 2 or 4 for the
587 * checksum to update.
589 * A call to this helper is susceptible to change the underlaying
590 * packet buffer. Therefore, at load time, all checks on pointers
591 * previously done by the verifier are invalidated and must be
592 * performed again, if the helper is used in combination with
593 * direct packet access.
595 * 0 on success, or a negative error in case of failure.
597 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
599 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
600 * packet associated to *skb*. Computation is incremental, so the
601 * helper must know the former value of the header field that was
602 * modified (*from*), the new value of this field (*to*), and the
603 * number of bytes (2 or 4) for this field, stored on the lowest
604 * four bits of *flags*. Alternatively, it is possible to store
605 * the difference between the previous and the new values of the
606 * header field in *to*, by setting *from* and the four lowest
607 * bits of *flags* to 0. For both methods, *offset* indicates the
608 * location of the IP checksum within the packet. In addition to
609 * the size of the field, *flags* can be added (bitwise OR) actual
610 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
611 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
612 * for updates resulting in a null checksum the value is set to
613 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
614 * the checksum is to be computed against a pseudo-header.
616 * This helper works in combination with **bpf_csum_diff**\ (),
617 * which does not update the checksum in-place, but offers more
618 * flexibility and can handle sizes larger than 2 or 4 for the
619 * checksum to update.
621 * A call to this helper is susceptible to change the underlaying
622 * packet buffer. Therefore, at load time, all checks on pointers
623 * previously done by the verifier are invalidated and must be
624 * performed again, if the helper is used in combination with
625 * direct packet access.
627 * 0 on success, or a negative error in case of failure.
629 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
631 * This special helper is used to trigger a "tail call", or in
632 * other words, to jump into another eBPF program. The same stack
633 * frame is used (but values on stack and in registers for the
634 * caller are not accessible to the callee). This mechanism allows
635 * for program chaining, either for raising the maximum number of
636 * available eBPF instructions, or to execute given programs in
637 * conditional blocks. For security reasons, there is an upper
638 * limit to the number of successive tail calls that can be
641 * Upon call of this helper, the program attempts to jump into a
642 * program referenced at index *index* in *prog_array_map*, a
643 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
644 * *ctx*, a pointer to the context.
646 * If the call succeeds, the kernel immediately runs the first
647 * instruction of the new program. This is not a function call,
648 * and it never returns to the previous program. If the call
649 * fails, then the helper has no effect, and the caller continues
650 * to run its subsequent instructions. A call can fail if the
651 * destination program for the jump does not exist (i.e. *index*
652 * is superior to the number of entries in *prog_array_map*), or
653 * if the maximum number of tail calls has been reached for this
654 * chain of programs. This limit is defined in the kernel by the
655 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
656 * which is currently set to 32.
658 * 0 on success, or a negative error in case of failure.
660 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
662 * Clone and redirect the packet associated to *skb* to another
663 * net device of index *ifindex*. Both ingress and egress
664 * interfaces can be used for redirection. The **BPF_F_INGRESS**
665 * value in *flags* is used to make the distinction (ingress path
666 * is selected if the flag is present, egress path otherwise).
667 * This is the only flag supported for now.
669 * In comparison with **bpf_redirect**\ () helper,
670 * **bpf_clone_redirect**\ () has the associated cost of
671 * duplicating the packet buffer, but this can be executed out of
672 * the eBPF program. Conversely, **bpf_redirect**\ () is more
673 * efficient, but it is handled through an action code where the
674 * redirection happens only after the eBPF program has returned.
676 * A call to this helper is susceptible to change the underlaying
677 * packet buffer. Therefore, at load time, all checks on pointers
678 * previously done by the verifier are invalidated and must be
679 * performed again, if the helper is used in combination with
680 * direct packet access.
682 * 0 on success, or a negative error in case of failure.
684 * u64 bpf_get_current_pid_tgid(void)
686 * A 64-bit integer containing the current tgid and pid, and
688 * *current_task*\ **->tgid << 32 \|**
689 * *current_task*\ **->pid**.
691 * u64 bpf_get_current_uid_gid(void)
693 * A 64-bit integer containing the current GID and UID, and
694 * created as such: *current_gid* **<< 32 \|** *current_uid*.
696 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
698 * Copy the **comm** attribute of the current task into *buf* of
699 * *size_of_buf*. The **comm** attribute contains the name of
700 * the executable (excluding the path) for the current task. The
701 * *size_of_buf* must be strictly positive. On success, the
702 * helper makes sure that the *buf* is NUL-terminated. On failure,
703 * it is filled with zeroes.
705 * 0 on success, or a negative error in case of failure.
707 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
709 * Retrieve the classid for the current task, i.e. for the net_cls
710 * cgroup to which *skb* belongs.
712 * This helper can be used on TC egress path, but not on ingress.
714 * The net_cls cgroup provides an interface to tag network packets
715 * based on a user-provided identifier for all traffic coming from
716 * the tasks belonging to the related cgroup. See also the related
717 * kernel documentation, available from the Linux sources in file
718 * *Documentation/cgroup-v1/net_cls.txt*.
720 * The Linux kernel has two versions for cgroups: there are
721 * cgroups v1 and cgroups v2. Both are available to users, who can
722 * use a mixture of them, but note that the net_cls cgroup is for
723 * cgroup v1 only. This makes it incompatible with BPF programs
724 * run on cgroups, which is a cgroup-v2-only feature (a socket can
725 * only hold data for one version of cgroups at a time).
727 * This helper is only available is the kernel was compiled with
728 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
729 * "**y**" or to "**m**".
731 * The classid, or 0 for the default unconfigured classid.
733 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
735 * Push a *vlan_tci* (VLAN tag control information) of protocol
736 * *vlan_proto* to the packet associated to *skb*, then update
737 * the checksum. Note that if *vlan_proto* is different from
738 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
739 * be **ETH_P_8021Q**.
741 * A call to this helper is susceptible to change the underlaying
742 * packet buffer. Therefore, at load time, all checks on pointers
743 * previously done by the verifier are invalidated and must be
744 * performed again, if the helper is used in combination with
745 * direct packet access.
747 * 0 on success, or a negative error in case of failure.
749 * int bpf_skb_vlan_pop(struct sk_buff *skb)
751 * Pop a VLAN header from the packet associated to *skb*.
753 * A call to this helper is susceptible to change the underlaying
754 * packet buffer. Therefore, at load time, all checks on pointers
755 * previously done by the verifier are invalidated and must be
756 * performed again, if the helper is used in combination with
757 * direct packet access.
759 * 0 on success, or a negative error in case of failure.
761 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
763 * Get tunnel metadata. This helper takes a pointer *key* to an
764 * empty **struct bpf_tunnel_key** of **size**, that will be
765 * filled with tunnel metadata for the packet associated to *skb*.
766 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
767 * indicates that the tunnel is based on IPv6 protocol instead of
770 * The **struct bpf_tunnel_key** is an object that generalizes the
771 * principal parameters used by various tunneling protocols into a
772 * single struct. This way, it can be used to easily make a
773 * decision based on the contents of the encapsulation header,
774 * "summarized" in this struct. In particular, it holds the IP
775 * address of the remote end (IPv4 or IPv6, depending on the case)
776 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
777 * this struct exposes the *key*\ **->tunnel_id**, which is
778 * generally mapped to a VNI (Virtual Network Identifier), making
779 * it programmable together with the **bpf_skb_set_tunnel_key**\
782 * Let's imagine that the following code is part of a program
783 * attached to the TC ingress interface, on one end of a GRE
784 * tunnel, and is supposed to filter out all messages coming from
785 * remote ends with IPv4 address other than 10.0.0.1:
790 * struct bpf_tunnel_key key = {};
792 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
794 * return TC_ACT_SHOT; // drop packet
796 * if (key.remote_ipv4 != 0x0a000001)
797 * return TC_ACT_SHOT; // drop packet
799 * return TC_ACT_OK; // accept packet
801 * This interface can also be used with all encapsulation devices
802 * that can operate in "collect metadata" mode: instead of having
803 * one network device per specific configuration, the "collect
804 * metadata" mode only requires a single device where the
805 * configuration can be extracted from this helper.
807 * This can be used together with various tunnels such as VXLan,
808 * Geneve, GRE or IP in IP (IPIP).
810 * 0 on success, or a negative error in case of failure.
812 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
814 * Populate tunnel metadata for packet associated to *skb.* The
815 * tunnel metadata is set to the contents of *key*, of *size*. The
816 * *flags* can be set to a combination of the following values:
818 * **BPF_F_TUNINFO_IPV6**
819 * Indicate that the tunnel is based on IPv6 protocol
821 * **BPF_F_ZERO_CSUM_TX**
822 * For IPv4 packets, add a flag to tunnel metadata
823 * indicating that checksum computation should be skipped
824 * and checksum set to zeroes.
825 * **BPF_F_DONT_FRAGMENT**
826 * Add a flag to tunnel metadata indicating that the
827 * packet should not be fragmented.
828 * **BPF_F_SEQ_NUMBER**
829 * Add a flag to tunnel metadata indicating that a
830 * sequence number should be added to tunnel header before
831 * sending the packet. This flag was added for GRE
832 * encapsulation, but might be used with other protocols
833 * as well in the future.
835 * Here is a typical usage on the transmit path:
839 * struct bpf_tunnel_key key;
841 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
842 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
844 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
845 * helper for additional information.
847 * 0 on success, or a negative error in case of failure.
849 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
851 * Read the value of a perf event counter. This helper relies on a
852 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
853 * the perf event counter is selected when *map* is updated with
854 * perf event file descriptors. The *map* is an array whose size
855 * is the number of available CPUs, and each cell contains a value
856 * relative to one CPU. The value to retrieve is indicated by
857 * *flags*, that contains the index of the CPU to look up, masked
858 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
859 * **BPF_F_CURRENT_CPU** to indicate that the value for the
860 * current CPU should be retrieved.
862 * Note that before Linux 4.13, only hardware perf event can be
865 * Also, be aware that the newer helper
866 * **bpf_perf_event_read_value**\ () is recommended over
867 * **bpf_perf_event_read**\ () in general. The latter has some ABI
868 * quirks where error and counter value are used as a return code
869 * (which is wrong to do since ranges may overlap). This issue is
870 * fixed with **bpf_perf_event_read_value**\ (), which at the same
871 * time provides more features over the **bpf_perf_event_read**\
872 * () interface. Please refer to the description of
873 * **bpf_perf_event_read_value**\ () for details.
875 * The value of the perf event counter read from the map, or a
876 * negative error code in case of failure.
878 * int bpf_redirect(u32 ifindex, u64 flags)
880 * Redirect the packet to another net device of index *ifindex*.
881 * This helper is somewhat similar to **bpf_clone_redirect**\
882 * (), except that the packet is not cloned, which provides
883 * increased performance.
885 * Except for XDP, both ingress and egress interfaces can be used
886 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
887 * to make the distinction (ingress path is selected if the flag
888 * is present, egress path otherwise). Currently, XDP only
889 * supports redirection to the egress interface, and accepts no
892 * The same effect can be attained with the more generic
893 * **bpf_redirect_map**\ (), which requires specific maps to be
894 * used but offers better performance.
896 * For XDP, the helper returns **XDP_REDIRECT** on success or
897 * **XDP_ABORTED** on error. For other program types, the values
898 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
901 * u32 bpf_get_route_realm(struct sk_buff *skb)
903 * Retrieve the realm or the route, that is to say the
904 * **tclassid** field of the destination for the *skb*. The
905 * indentifier retrieved is a user-provided tag, similar to the
906 * one used with the net_cls cgroup (see description for
907 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
908 * held by a route (a destination entry), not by a task.
910 * Retrieving this identifier works with the clsact TC egress hook
911 * (see also **tc-bpf(8)**), or alternatively on conventional
912 * classful egress qdiscs, but not on TC ingress path. In case of
913 * clsact TC egress hook, this has the advantage that, internally,
914 * the destination entry has not been dropped yet in the transmit
915 * path. Therefore, the destination entry does not need to be
916 * artificially held via **netif_keep_dst**\ () for a classful
917 * qdisc until the *skb* is freed.
919 * This helper is available only if the kernel was compiled with
920 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
922 * The realm of the route for the packet associated to *skb*, or 0
925 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
927 * Write raw *data* blob into a special BPF perf event held by
928 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
929 * event must have the following attributes: **PERF_SAMPLE_RAW**
930 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
931 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
933 * The *flags* are used to indicate the index in *map* for which
934 * the value must be put, masked with **BPF_F_INDEX_MASK**.
935 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
936 * to indicate that the index of the current CPU core should be
939 * The value to write, of *size*, is passed through eBPF stack and
942 * The context of the program *ctx* needs also be passed to the
945 * On user space, a program willing to read the values needs to
946 * call **perf_event_open**\ () on the perf event (either for
947 * one or for all CPUs) and to store the file descriptor into the
948 * *map*. This must be done before the eBPF program can send data
949 * into it. An example is available in file
950 * *samples/bpf/trace_output_user.c* in the Linux kernel source
951 * tree (the eBPF program counterpart is in
952 * *samples/bpf/trace_output_kern.c*).
954 * **bpf_perf_event_output**\ () achieves better performance
955 * than **bpf_trace_printk**\ () for sharing data with user
956 * space, and is much better suitable for streaming data from eBPF
959 * Note that this helper is not restricted to tracing use cases
960 * and can be used with programs attached to TC or XDP as well,
961 * where it allows for passing data to user space listeners. Data
964 * * Only custom structs,
965 * * Only the packet payload, or
966 * * A combination of both.
968 * 0 on success, or a negative error in case of failure.
970 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
972 * This helper was provided as an easy way to load data from a
973 * packet. It can be used to load *len* bytes from *offset* from
974 * the packet associated to *skb*, into the buffer pointed by
977 * Since Linux 4.7, usage of this helper has mostly been replaced
978 * by "direct packet access", enabling packet data to be
979 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
980 * pointing respectively to the first byte of packet data and to
981 * the byte after the last byte of packet data. However, it
982 * remains useful if one wishes to read large quantities of data
983 * at once from a packet into the eBPF stack.
985 * 0 on success, or a negative error in case of failure.
987 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
989 * Walk a user or a kernel stack and return its id. To achieve
990 * this, the helper needs *ctx*, which is a pointer to the context
991 * on which the tracing program is executed, and a pointer to a
992 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
994 * The last argument, *flags*, holds the number of stack frames to
995 * skip (from 0 to 255), masked with
996 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
997 * a combination of the following flags:
999 * **BPF_F_USER_STACK**
1000 * Collect a user space stack instead of a kernel stack.
1001 * **BPF_F_FAST_STACK_CMP**
1002 * Compare stacks by hash only.
1003 * **BPF_F_REUSE_STACKID**
1004 * If two different stacks hash into the same *stackid*,
1005 * discard the old one.
1007 * The stack id retrieved is a 32 bit long integer handle which
1008 * can be further combined with other data (including other stack
1009 * ids) and used as a key into maps. This can be useful for
1010 * generating a variety of graphs (such as flame graphs or off-cpu
1013 * For walking a stack, this helper is an improvement over
1014 * **bpf_probe_read**\ (), which can be used with unrolled loops
1015 * but is not efficient and consumes a lot of eBPF instructions.
1016 * Instead, **bpf_get_stackid**\ () can collect up to
1017 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1018 * this limit can be controlled with the **sysctl** program, and
1019 * that it should be manually increased in order to profile long
1020 * user stacks (such as stacks for Java programs). To do so, use:
1024 * # sysctl kernel.perf_event_max_stack=<new value>
1026 * The positive or null stack id on success, or a negative error
1027 * in case of failure.
1029 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1031 * Compute a checksum difference, from the raw buffer pointed by
1032 * *from*, of length *from_size* (that must be a multiple of 4),
1033 * towards the raw buffer pointed by *to*, of size *to_size*
1034 * (same remark). An optional *seed* can be added to the value
1035 * (this can be cascaded, the seed may come from a previous call
1038 * This is flexible enough to be used in several ways:
1040 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1041 * checksum, it can be used when pushing new data.
1042 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1043 * checksum, it can be used when removing data from a packet.
1044 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1045 * can be used to compute a diff. Note that *from_size* and
1046 * *to_size* do not need to be equal.
1048 * This helper can be used in combination with
1049 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1050 * which one can feed in the difference computed with
1051 * **bpf_csum_diff**\ ().
1053 * The checksum result, or a negative error code in case of
1056 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1058 * Retrieve tunnel options metadata for the packet associated to
1059 * *skb*, and store the raw tunnel option data to the buffer *opt*
1062 * This helper can be used with encapsulation devices that can
1063 * operate in "collect metadata" mode (please refer to the related
1064 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1065 * more details). A particular example where this can be used is
1066 * in combination with the Geneve encapsulation protocol, where it
1067 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1068 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1069 * the eBPF program. This allows for full customization of these
1072 * The size of the option data retrieved.
1074 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1076 * Set tunnel options metadata for the packet associated to *skb*
1077 * to the option data contained in the raw buffer *opt* of *size*.
1079 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1080 * helper for additional information.
1082 * 0 on success, or a negative error in case of failure.
1084 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1086 * Change the protocol of the *skb* to *proto*. Currently
1087 * supported are transition from IPv4 to IPv6, and from IPv6 to
1088 * IPv4. The helper takes care of the groundwork for the
1089 * transition, including resizing the socket buffer. The eBPF
1090 * program is expected to fill the new headers, if any, via
1091 * **skb_store_bytes**\ () and to recompute the checksums with
1092 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1093 * (). The main case for this helper is to perform NAT64
1094 * operations out of an eBPF program.
1096 * Internally, the GSO type is marked as dodgy so that headers are
1097 * checked and segments are recalculated by the GSO/GRO engine.
1098 * The size for GSO target is adapted as well.
1100 * All values for *flags* are reserved for future usage, and must
1103 * A call to this helper is susceptible to change the underlaying
1104 * packet buffer. Therefore, at load time, all checks on pointers
1105 * previously done by the verifier are invalidated and must be
1106 * performed again, if the helper is used in combination with
1107 * direct packet access.
1109 * 0 on success, or a negative error in case of failure.
1111 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1113 * Change the packet type for the packet associated to *skb*. This
1114 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1115 * the eBPF program does not have a write access to *skb*\
1116 * **->pkt_type** beside this helper. Using a helper here allows
1117 * for graceful handling of errors.
1119 * The major use case is to change incoming *skb*s to
1120 * **PACKET_HOST** in a programmatic way instead of having to
1121 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1124 * Note that *type* only allows certain values. At this time, they
1129 * **PACKET_BROADCAST**
1130 * Send packet to all.
1131 * **PACKET_MULTICAST**
1132 * Send packet to group.
1133 * **PACKET_OTHERHOST**
1134 * Send packet to someone else.
1136 * 0 on success, or a negative error in case of failure.
1138 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1140 * Check whether *skb* is a descendant of the cgroup2 held by
1141 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1143 * The return value depends on the result of the test, and can be:
1145 * * 0, if the *skb* failed the cgroup2 descendant test.
1146 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1147 * * A negative error code, if an error occurred.
1149 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1151 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1152 * not set, in particular if the hash was cleared due to mangling,
1153 * recompute this hash. Later accesses to the hash can be done
1154 * directly with *skb*\ **->hash**.
1156 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1157 * prototype with **bpf_skb_change_proto**\ (), or calling
1158 * **bpf_skb_store_bytes**\ () with the
1159 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1160 * the hash and to trigger a new computation for the next call to
1161 * **bpf_get_hash_recalc**\ ().
1165 * u64 bpf_get_current_task(void)
1167 * A pointer to the current task struct.
1169 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1171 * Attempt in a safe way to write *len* bytes from the buffer
1172 * *src* to *dst* in memory. It only works for threads that are in
1173 * user context, and *dst* must be a valid user space address.
1175 * This helper should not be used to implement any kind of
1176 * security mechanism because of TOC-TOU attacks, but rather to
1177 * debug, divert, and manipulate execution of semi-cooperative
1180 * Keep in mind that this feature is meant for experiments, and it
1181 * has a risk of crashing the system and running programs.
1182 * Therefore, when an eBPF program using this helper is attached,
1183 * a warning including PID and process name is printed to kernel
1186 * 0 on success, or a negative error in case of failure.
1188 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1190 * Check whether the probe is being run is the context of a given
1191 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1192 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1194 * The return value depends on the result of the test, and can be:
1196 * * 0, if current task belongs to the cgroup2.
1197 * * 1, if current task does not belong to the cgroup2.
1198 * * A negative error code, if an error occurred.
1200 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1202 * Resize (trim or grow) the packet associated to *skb* to the
1203 * new *len*. The *flags* are reserved for future usage, and must
1206 * The basic idea is that the helper performs the needed work to
1207 * change the size of the packet, then the eBPF program rewrites
1208 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1209 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1210 * and others. This helper is a slow path utility intended for
1211 * replies with control messages. And because it is targeted for
1212 * slow path, the helper itself can afford to be slow: it
1213 * implicitly linearizes, unclones and drops offloads from the
1216 * A call to this helper is susceptible to change the underlaying
1217 * packet buffer. Therefore, at load time, all checks on pointers
1218 * previously done by the verifier are invalidated and must be
1219 * performed again, if the helper is used in combination with
1220 * direct packet access.
1222 * 0 on success, or a negative error in case of failure.
1224 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1226 * Pull in non-linear data in case the *skb* is non-linear and not
1227 * all of *len* are part of the linear section. Make *len* bytes
1228 * from *skb* readable and writable. If a zero value is passed for
1229 * *len*, then the whole length of the *skb* is pulled.
1231 * This helper is only needed for reading and writing with direct
1234 * For direct packet access, testing that offsets to access
1235 * are within packet boundaries (test on *skb*\ **->data_end**) is
1236 * susceptible to fail if offsets are invalid, or if the requested
1237 * data is in non-linear parts of the *skb*. On failure the
1238 * program can just bail out, or in the case of a non-linear
1239 * buffer, use a helper to make the data available. The
1240 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1241 * the data. Another one consists in using **bpf_skb_pull_data**
1242 * to pull in once the non-linear parts, then retesting and
1243 * eventually access the data.
1245 * At the same time, this also makes sure the *skb* is uncloned,
1246 * which is a necessary condition for direct write. As this needs
1247 * to be an invariant for the write part only, the verifier
1248 * detects writes and adds a prologue that is calling
1249 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1250 * the very beginning in case it is indeed cloned.
1252 * A call to this helper is susceptible to change the underlaying
1253 * packet buffer. Therefore, at load time, all checks on pointers
1254 * previously done by the verifier are invalidated and must be
1255 * performed again, if the helper is used in combination with
1256 * direct packet access.
1258 * 0 on success, or a negative error in case of failure.
1260 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1262 * Add the checksum *csum* into *skb*\ **->csum** in case the
1263 * driver has supplied a checksum for the entire packet into that
1264 * field. Return an error otherwise. This helper is intended to be
1265 * used in combination with **bpf_csum_diff**\ (), in particular
1266 * when the checksum needs to be updated after data has been
1267 * written into the packet through direct packet access.
1269 * The checksum on success, or a negative error code in case of
1272 * void bpf_set_hash_invalid(struct sk_buff *skb)
1274 * Invalidate the current *skb*\ **->hash**. It can be used after
1275 * mangling on headers through direct packet access, in order to
1276 * indicate that the hash is outdated and to trigger a
1277 * recalculation the next time the kernel tries to access this
1278 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1280 * int bpf_get_numa_node_id(void)
1282 * Return the id of the current NUMA node. The primary use case
1283 * for this helper is the selection of sockets for the local NUMA
1284 * node, when the program is attached to sockets using the
1285 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1286 * but the helper is also available to other eBPF program types,
1287 * similarly to **bpf_get_smp_processor_id**\ ().
1289 * The id of current NUMA node.
1291 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1293 * Grows headroom of packet associated to *skb* and adjusts the
1294 * offset of the MAC header accordingly, adding *len* bytes of
1295 * space. It automatically extends and reallocates memory as
1298 * This helper can be used on a layer 3 *skb* to push a MAC header
1299 * for redirection into a layer 2 device.
1301 * All values for *flags* are reserved for future usage, and must
1304 * A call to this helper is susceptible to change the underlaying
1305 * packet buffer. Therefore, at load time, all checks on pointers
1306 * previously done by the verifier are invalidated and must be
1307 * performed again, if the helper is used in combination with
1308 * direct packet access.
1310 * 0 on success, or a negative error in case of failure.
1312 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1314 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1315 * it is possible to use a negative value for *delta*. This helper
1316 * can be used to prepare the packet for pushing or popping
1319 * A call to this helper is susceptible to change the underlaying
1320 * packet buffer. Therefore, at load time, all checks on pointers
1321 * previously done by the verifier are invalidated and must be
1322 * performed again, if the helper is used in combination with
1323 * direct packet access.
1325 * 0 on success, or a negative error in case of failure.
1327 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1329 * Copy a NUL terminated string from an unsafe address
1330 * *unsafe_ptr* to *dst*. The *size* should include the
1331 * terminating NUL byte. In case the string length is smaller than
1332 * *size*, the target is not padded with further NUL bytes. If the
1333 * string length is larger than *size*, just *size*-1 bytes are
1334 * copied and the last byte is set to NUL.
1336 * On success, the length of the copied string is returned. This
1337 * makes this helper useful in tracing programs for reading
1338 * strings, and more importantly to get its length at runtime. See
1339 * the following snippet:
1343 * SEC("kprobe/sys_open")
1344 * void bpf_sys_open(struct pt_regs *ctx)
1346 * char buf[PATHLEN]; // PATHLEN is defined to 256
1347 * int res = bpf_probe_read_str(buf, sizeof(buf),
1350 * // Consume buf, for example push it to
1351 * // userspace via bpf_perf_event_output(); we
1352 * // can use res (the string length) as event
1353 * // size, after checking its boundaries.
1356 * In comparison, using **bpf_probe_read()** helper here instead
1357 * to read the string would require to estimate the length at
1358 * compile time, and would often result in copying more memory
1361 * Another useful use case is when parsing individual process
1362 * arguments or individual environment variables navigating
1363 * *current*\ **->mm->arg_start** and *current*\
1364 * **->mm->env_start**: using this helper and the return value,
1365 * one can quickly iterate at the right offset of the memory area.
1367 * On success, the strictly positive length of the string,
1368 * including the trailing NUL character. On error, a negative
1371 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1373 * If the **struct sk_buff** pointed by *skb* has a known socket,
1374 * retrieve the cookie (generated by the kernel) of this socket.
1375 * If no cookie has been set yet, generate a new cookie. Once
1376 * generated, the socket cookie remains stable for the life of the
1377 * socket. This helper can be useful for monitoring per socket
1378 * networking traffic statistics as it provides a unique socket
1379 * identifier per namespace.
1381 * A 8-byte long non-decreasing number on success, or 0 if the
1382 * socket field is missing inside *skb*.
1384 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1386 * Equivalent to bpf_get_socket_cookie() helper that accepts
1387 * *skb*, but gets socket from **struct bpf_sock_addr** contex.
1389 * A 8-byte long non-decreasing number.
1391 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1393 * Equivalent to bpf_get_socket_cookie() helper that accepts
1394 * *skb*, but gets socket from **struct bpf_sock_ops** contex.
1396 * A 8-byte long non-decreasing number.
1398 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1400 * The owner UID of the socket associated to *skb*. If the socket
1401 * is **NULL**, or if it is not a full socket (i.e. if it is a
1402 * time-wait or a request socket instead), **overflowuid** value
1403 * is returned (note that **overflowuid** might also be the actual
1404 * UID value for the socket).
1406 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1408 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1413 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1415 * Emulate a call to **setsockopt()** on the socket associated to
1416 * *bpf_socket*, which must be a full socket. The *level* at
1417 * which the option resides and the name *optname* of the option
1418 * must be specified, see **setsockopt(2)** for more information.
1419 * The option value of length *optlen* is pointed by *optval*.
1421 * This helper actually implements a subset of **setsockopt()**.
1422 * It supports the following *level*\ s:
1424 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1425 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1426 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1427 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1428 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1429 * **TCP_BPF_SNDCWND_CLAMP**.
1430 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1431 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1433 * 0 on success, or a negative error in case of failure.
1435 * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags)
1437 * Grow or shrink the room for data in the packet associated to
1438 * *skb* by *len_diff*, and according to the selected *mode*.
1440 * There is a single supported mode at this time:
1442 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1443 * (room space is added or removed below the layer 3 header).
1445 * All values for *flags* are reserved for future usage, and must
1448 * A call to this helper is susceptible to change the underlaying
1449 * packet buffer. Therefore, at load time, all checks on pointers
1450 * previously done by the verifier are invalidated and must be
1451 * performed again, if the helper is used in combination with
1452 * direct packet access.
1454 * 0 on success, or a negative error in case of failure.
1456 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1458 * Redirect the packet to the endpoint referenced by *map* at
1459 * index *key*. Depending on its type, this *map* can contain
1460 * references to net devices (for forwarding packets through other
1461 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1462 * but this is only implemented for native XDP (with driver
1463 * support) as of this writing).
1465 * All values for *flags* are reserved for future usage, and must
1468 * When used to redirect packets to net devices, this helper
1469 * provides a high performance increase over **bpf_redirect**\ ().
1470 * This is due to various implementation details of the underlying
1471 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1472 * () tries to send packet as a "bulk" to the device.
1474 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1476 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1478 * Redirect the packet to the socket referenced by *map* (of type
1479 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1480 * egress interfaces can be used for redirection. The
1481 * **BPF_F_INGRESS** value in *flags* is used to make the
1482 * distinction (ingress path is selected if the flag is present,
1483 * egress path otherwise). This is the only flag supported for now.
1485 * **SK_PASS** on success, or **SK_DROP** on error.
1487 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1489 * Add an entry to, or update a *map* referencing sockets. The
1490 * *skops* is used as a new value for the entry associated to
1491 * *key*. *flags* is one of:
1494 * The entry for *key* must not exist in the map.
1496 * The entry for *key* must already exist in the map.
1498 * No condition on the existence of the entry for *key*.
1500 * If the *map* has eBPF programs (parser and verdict), those will
1501 * be inherited by the socket being added. If the socket is
1502 * already attached to eBPF programs, this results in an error.
1504 * 0 on success, or a negative error in case of failure.
1506 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1508 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1509 * *delta* (which can be positive or negative). Note that this
1510 * operation modifies the address stored in *xdp_md*\ **->data**,
1511 * so the latter must be loaded only after the helper has been
1514 * The use of *xdp_md*\ **->data_meta** is optional and programs
1515 * are not required to use it. The rationale is that when the
1516 * packet is processed with XDP (e.g. as DoS filter), it is
1517 * possible to push further meta data along with it before passing
1518 * to the stack, and to give the guarantee that an ingress eBPF
1519 * program attached as a TC classifier on the same device can pick
1520 * this up for further post-processing. Since TC works with socket
1521 * buffers, it remains possible to set from XDP the **mark** or
1522 * **priority** pointers, or other pointers for the socket buffer.
1523 * Having this scratch space generic and programmable allows for
1524 * more flexibility as the user is free to store whatever meta
1527 * A call to this helper is susceptible to change the underlaying
1528 * packet buffer. Therefore, at load time, all checks on pointers
1529 * previously done by the verifier are invalidated and must be
1530 * performed again, if the helper is used in combination with
1531 * direct packet access.
1533 * 0 on success, or a negative error in case of failure.
1535 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1537 * Read the value of a perf event counter, and store it into *buf*
1538 * of size *buf_size*. This helper relies on a *map* of type
1539 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1540 * counter is selected when *map* is updated with perf event file
1541 * descriptors. The *map* is an array whose size is the number of
1542 * available CPUs, and each cell contains a value relative to one
1543 * CPU. The value to retrieve is indicated by *flags*, that
1544 * contains the index of the CPU to look up, masked with
1545 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1546 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1547 * current CPU should be retrieved.
1549 * This helper behaves in a way close to
1550 * **bpf_perf_event_read**\ () helper, save that instead of
1551 * just returning the value observed, it fills the *buf*
1552 * structure. This allows for additional data to be retrieved: in
1553 * particular, the enabled and running times (in *buf*\
1554 * **->enabled** and *buf*\ **->running**, respectively) are
1555 * copied. In general, **bpf_perf_event_read_value**\ () is
1556 * recommended over **bpf_perf_event_read**\ (), which has some
1557 * ABI issues and provides fewer functionalities.
1559 * These values are interesting, because hardware PMU (Performance
1560 * Monitoring Unit) counters are limited resources. When there are
1561 * more PMU based perf events opened than available counters,
1562 * kernel will multiplex these events so each event gets certain
1563 * percentage (but not all) of the PMU time. In case that
1564 * multiplexing happens, the number of samples or counter value
1565 * will not reflect the case compared to when no multiplexing
1566 * occurs. This makes comparison between different runs difficult.
1567 * Typically, the counter value should be normalized before
1568 * comparing to other experiments. The usual normalization is done
1573 * normalized_counter = counter * t_enabled / t_running
1575 * Where t_enabled is the time enabled for event and t_running is
1576 * the time running for event since last normalization. The
1577 * enabled and running times are accumulated since the perf event
1578 * open. To achieve scaling factor between two invocations of an
1579 * eBPF program, users can can use CPU id as the key (which is
1580 * typical for perf array usage model) to remember the previous
1581 * value and do the calculation inside the eBPF program.
1583 * 0 on success, or a negative error in case of failure.
1585 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1587 * For en eBPF program attached to a perf event, retrieve the
1588 * value of the event counter associated to *ctx* and store it in
1589 * the structure pointed by *buf* and of size *buf_size*. Enabled
1590 * and running times are also stored in the structure (see
1591 * description of helper **bpf_perf_event_read_value**\ () for
1594 * 0 on success, or a negative error in case of failure.
1596 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1598 * Emulate a call to **getsockopt()** on the socket associated to
1599 * *bpf_socket*, which must be a full socket. The *level* at
1600 * which the option resides and the name *optname* of the option
1601 * must be specified, see **getsockopt(2)** for more information.
1602 * The retrieved value is stored in the structure pointed by
1603 * *opval* and of length *optlen*.
1605 * This helper actually implements a subset of **getsockopt()**.
1606 * It supports the following *level*\ s:
1608 * * **IPPROTO_TCP**, which supports *optname*
1609 * **TCP_CONGESTION**.
1610 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1611 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1613 * 0 on success, or a negative error in case of failure.
1615 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1617 * Used for error injection, this helper uses kprobes to override
1618 * the return value of the probed function, and to set it to *rc*.
1619 * The first argument is the context *regs* on which the kprobe
1622 * This helper works by setting setting the PC (program counter)
1623 * to an override function which is run in place of the original
1624 * probed function. This means the probed function is not run at
1625 * all. The replacement function just returns with the required
1628 * This helper has security implications, and thus is subject to
1629 * restrictions. It is only available if the kernel was compiled
1630 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1631 * option, and in this case it only works on functions tagged with
1632 * **ALLOW_ERROR_INJECTION** in the kernel code.
1634 * Also, the helper is only available for the architectures having
1635 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1636 * x86 architecture is the only one to support this feature.
1640 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1642 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1643 * for the full TCP socket associated to *bpf_sock_ops* to
1646 * The primary use of this field is to determine if there should
1647 * be calls to eBPF programs of type
1648 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1649 * code. A program of the same type can change its value, per
1650 * connection and as necessary, when the connection is
1651 * established. This field is directly accessible for reading, but
1652 * this helper must be used for updates in order to return an
1653 * error if an eBPF program tries to set a callback that is not
1654 * supported in the current kernel.
1656 * The supported callback values that *argval* can combine are:
1658 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1659 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1660 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1662 * Here are some examples of where one could call such eBPF
1666 * * When a packet is retransmitted.
1667 * * When the connection terminates.
1668 * * When a packet is sent.
1669 * * When a packet is received.
1671 * Code **-EINVAL** if the socket is not a full TCP socket;
1672 * otherwise, a positive number containing the bits that could not
1673 * be set is returned (which comes down to 0 if all bits were set
1676 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1678 * This helper is used in programs implementing policies at the
1679 * socket level. If the message *msg* is allowed to pass (i.e. if
1680 * the verdict eBPF program returns **SK_PASS**), redirect it to
1681 * the socket referenced by *map* (of type
1682 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1683 * egress interfaces can be used for redirection. The
1684 * **BPF_F_INGRESS** value in *flags* is used to make the
1685 * distinction (ingress path is selected if the flag is present,
1686 * egress path otherwise). This is the only flag supported for now.
1688 * **SK_PASS** on success, or **SK_DROP** on error.
1690 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1692 * For socket policies, apply the verdict of the eBPF program to
1693 * the next *bytes* (number of bytes) of message *msg*.
1695 * For example, this helper can be used in the following cases:
1697 * * A single **sendmsg**\ () or **sendfile**\ () system call
1698 * contains multiple logical messages that the eBPF program is
1699 * supposed to read and for which it should apply a verdict.
1700 * * An eBPF program only cares to read the first *bytes* of a
1701 * *msg*. If the message has a large payload, then setting up
1702 * and calling the eBPF program repeatedly for all bytes, even
1703 * though the verdict is already known, would create unnecessary
1706 * When called from within an eBPF program, the helper sets a
1707 * counter internal to the BPF infrastructure, that is used to
1708 * apply the last verdict to the next *bytes*. If *bytes* is
1709 * smaller than the current data being processed from a
1710 * **sendmsg**\ () or **sendfile**\ () system call, the first
1711 * *bytes* will be sent and the eBPF program will be re-run with
1712 * the pointer for start of data pointing to byte number *bytes*
1713 * **+ 1**. If *bytes* is larger than the current data being
1714 * processed, then the eBPF verdict will be applied to multiple
1715 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1718 * Note that if a socket closes with the internal counter holding
1719 * a non-zero value, this is not a problem because data is not
1720 * being buffered for *bytes* and is sent as it is received.
1724 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1726 * For socket policies, prevent the execution of the verdict eBPF
1727 * program for message *msg* until *bytes* (byte number) have been
1730 * This can be used when one needs a specific number of bytes
1731 * before a verdict can be assigned, even if the data spans
1732 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1733 * case would be a user calling **sendmsg**\ () repeatedly with
1734 * 1-byte long message segments. Obviously, this is bad for
1735 * performance, but it is still valid. If the eBPF program needs
1736 * *bytes* bytes to validate a header, this helper can be used to
1737 * prevent the eBPF program to be called again until *bytes* have
1742 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1744 * For socket policies, pull in non-linear data from user space
1745 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1746 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1749 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1750 * *msg* it can only parse data that the (**data**, **data_end**)
1751 * pointers have already consumed. For **sendmsg**\ () hooks this
1752 * is likely the first scatterlist element. But for calls relying
1753 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1754 * be the range (**0**, **0**) because the data is shared with
1755 * user space and by default the objective is to avoid allowing
1756 * user space to modify data while (or after) eBPF verdict is
1757 * being decided. This helper can be used to pull in data and to
1758 * set the start and end pointer to given values. Data will be
1759 * copied if necessary (i.e. if data was not linear and if start
1760 * and end pointers do not point to the same chunk).
1762 * A call to this helper is susceptible to change the underlaying
1763 * packet buffer. Therefore, at load time, all checks on pointers
1764 * previously done by the verifier are invalidated and must be
1765 * performed again, if the helper is used in combination with
1766 * direct packet access.
1768 * All values for *flags* are reserved for future usage, and must
1771 * 0 on success, or a negative error in case of failure.
1773 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1775 * Bind the socket associated to *ctx* to the address pointed by
1776 * *addr*, of length *addr_len*. This allows for making outgoing
1777 * connection from the desired IP address, which can be useful for
1778 * example when all processes inside a cgroup should use one
1779 * single IP address on a host that has multiple IP configured.
1781 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1782 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1783 * **AF_INET6**). Looking for a free port to bind to can be
1784 * expensive, therefore binding to port is not permitted by the
1785 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1786 * must be set to zero.
1788 * 0 on success, or a negative error in case of failure.
1790 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1792 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1793 * only possible to shrink the packet as of this writing,
1794 * therefore *delta* must be a negative integer.
1796 * A call to this helper is susceptible to change the underlaying
1797 * packet buffer. Therefore, at load time, all checks on pointers
1798 * previously done by the verifier are invalidated and must be
1799 * performed again, if the helper is used in combination with
1800 * direct packet access.
1802 * 0 on success, or a negative error in case of failure.
1804 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1806 * Retrieve the XFRM state (IP transform framework, see also
1807 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1809 * The retrieved value is stored in the **struct bpf_xfrm_state**
1810 * pointed by *xfrm_state* and of length *size*.
1812 * All values for *flags* are reserved for future usage, and must
1815 * This helper is available only if the kernel was compiled with
1816 * **CONFIG_XFRM** configuration option.
1818 * 0 on success, or a negative error in case of failure.
1820 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1822 * Return a user or a kernel stack in bpf program provided buffer.
1823 * To achieve this, the helper needs *ctx*, which is a pointer
1824 * to the context on which the tracing program is executed.
1825 * To store the stacktrace, the bpf program provides *buf* with
1826 * a nonnegative *size*.
1828 * The last argument, *flags*, holds the number of stack frames to
1829 * skip (from 0 to 255), masked with
1830 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1831 * the following flags:
1833 * **BPF_F_USER_STACK**
1834 * Collect a user space stack instead of a kernel stack.
1835 * **BPF_F_USER_BUILD_ID**
1836 * Collect buildid+offset instead of ips for user stack,
1837 * only valid if **BPF_F_USER_STACK** is also specified.
1839 * **bpf_get_stack**\ () can collect up to
1840 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1841 * to sufficient large buffer size. Note that
1842 * this limit can be controlled with the **sysctl** program, and
1843 * that it should be manually increased in order to profile long
1844 * user stacks (such as stacks for Java programs). To do so, use:
1848 * # sysctl kernel.perf_event_max_stack=<new value>
1850 * A non-negative value equal to or less than *size* on success,
1851 * or a negative error in case of failure.
1853 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1855 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1856 * it provides an easy way to load *len* bytes from *offset*
1857 * from the packet associated to *skb*, into the buffer pointed
1858 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1859 * a fifth argument *start_header* exists in order to select a
1860 * base offset to start from. *start_header* can be one of:
1862 * **BPF_HDR_START_MAC**
1863 * Base offset to load data from is *skb*'s mac header.
1864 * **BPF_HDR_START_NET**
1865 * Base offset to load data from is *skb*'s network header.
1867 * In general, "direct packet access" is the preferred method to
1868 * access packet data, however, this helper is in particular useful
1869 * in socket filters where *skb*\ **->data** does not always point
1870 * to the start of the mac header and where "direct packet access"
1873 * 0 on success, or a negative error in case of failure.
1875 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1877 * Do FIB lookup in kernel tables using parameters in *params*.
1878 * If lookup is successful and result shows packet is to be
1879 * forwarded, the neighbor tables are searched for the nexthop.
1880 * If successful (ie., FIB lookup shows forwarding and nexthop
1881 * is resolved), the nexthop address is returned in ipv4_dst
1882 * or ipv6_dst based on family, smac is set to mac address of
1883 * egress device, dmac is set to nexthop mac address, rt_metric
1884 * is set to metric from route (IPv4/IPv6 only), and ifindex
1885 * is set to the device index of the nexthop from the FIB lookup.
1887 * *plen* argument is the size of the passed in struct.
1888 * *flags* argument can be a combination of one or more of the
1891 * **BPF_FIB_LOOKUP_DIRECT**
1892 * Do a direct table lookup vs full lookup using FIB
1894 * **BPF_FIB_LOOKUP_OUTPUT**
1895 * Perform lookup from an egress perspective (default is
1898 * *ctx* is either **struct xdp_md** for XDP programs or
1899 * **struct sk_buff** tc cls_act programs.
1901 * * < 0 if any input argument is invalid
1902 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1903 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1904 * packet is not forwarded or needs assist from full stack
1906 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1908 * Add an entry to, or update a sockhash *map* referencing sockets.
1909 * The *skops* is used as a new value for the entry associated to
1910 * *key*. *flags* is one of:
1913 * The entry for *key* must not exist in the map.
1915 * The entry for *key* must already exist in the map.
1917 * No condition on the existence of the entry for *key*.
1919 * If the *map* has eBPF programs (parser and verdict), those will
1920 * be inherited by the socket being added. If the socket is
1921 * already attached to eBPF programs, this results in an error.
1923 * 0 on success, or a negative error in case of failure.
1925 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1927 * This helper is used in programs implementing policies at the
1928 * socket level. If the message *msg* is allowed to pass (i.e. if
1929 * the verdict eBPF program returns **SK_PASS**), redirect it to
1930 * the socket referenced by *map* (of type
1931 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1932 * egress interfaces can be used for redirection. The
1933 * **BPF_F_INGRESS** value in *flags* is used to make the
1934 * distinction (ingress path is selected if the flag is present,
1935 * egress path otherwise). This is the only flag supported for now.
1937 * **SK_PASS** on success, or **SK_DROP** on error.
1939 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1941 * This helper is used in programs implementing policies at the
1942 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1943 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1944 * to the socket referenced by *map* (of type
1945 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1946 * egress interfaces can be used for redirection. The
1947 * **BPF_F_INGRESS** value in *flags* is used to make the
1948 * distinction (ingress path is selected if the flag is present,
1949 * egress otherwise). This is the only flag supported for now.
1951 * **SK_PASS** on success, or **SK_DROP** on error.
1953 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
1955 * Encapsulate the packet associated to *skb* within a Layer 3
1956 * protocol header. This header is provided in the buffer at
1957 * address *hdr*, with *len* its size in bytes. *type* indicates
1958 * the protocol of the header and can be one of:
1960 * **BPF_LWT_ENCAP_SEG6**
1961 * IPv6 encapsulation with Segment Routing Header
1962 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
1963 * the IPv6 header is computed by the kernel.
1964 * **BPF_LWT_ENCAP_SEG6_INLINE**
1965 * Only works if *skb* contains an IPv6 packet. Insert a
1966 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
1969 * A call to this helper is susceptible to change the underlaying
1970 * packet buffer. Therefore, at load time, all checks on pointers
1971 * previously done by the verifier are invalidated and must be
1972 * performed again, if the helper is used in combination with
1973 * direct packet access.
1975 * 0 on success, or a negative error in case of failure.
1977 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
1979 * Store *len* bytes from address *from* into the packet
1980 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
1981 * inside the outermost IPv6 Segment Routing Header can be
1982 * modified through this helper.
1984 * A call to this helper is susceptible to change the underlaying
1985 * packet buffer. Therefore, at load time, all checks on pointers
1986 * previously done by the verifier are invalidated and must be
1987 * performed again, if the helper is used in combination with
1988 * direct packet access.
1990 * 0 on success, or a negative error in case of failure.
1992 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
1994 * Adjust the size allocated to TLVs in the outermost IPv6
1995 * Segment Routing Header contained in the packet associated to
1996 * *skb*, at position *offset* by *delta* bytes. Only offsets
1997 * after the segments are accepted. *delta* can be as well
1998 * positive (growing) as negative (shrinking).
2000 * A call to this helper is susceptible to change the underlaying
2001 * packet buffer. Therefore, at load time, all checks on pointers
2002 * previously done by the verifier are invalidated and must be
2003 * performed again, if the helper is used in combination with
2004 * direct packet access.
2006 * 0 on success, or a negative error in case of failure.
2008 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2010 * Apply an IPv6 Segment Routing action of type *action* to the
2011 * packet associated to *skb*. Each action takes a parameter
2012 * contained at address *param*, and of length *param_len* bytes.
2013 * *action* can be one of:
2015 * **SEG6_LOCAL_ACTION_END_X**
2016 * End.X action: Endpoint with Layer-3 cross-connect.
2017 * Type of *param*: **struct in6_addr**.
2018 * **SEG6_LOCAL_ACTION_END_T**
2019 * End.T action: Endpoint with specific IPv6 table lookup.
2020 * Type of *param*: **int**.
2021 * **SEG6_LOCAL_ACTION_END_B6**
2022 * End.B6 action: Endpoint bound to an SRv6 policy.
2023 * Type of param: **struct ipv6_sr_hdr**.
2024 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2025 * End.B6.Encap action: Endpoint bound to an SRv6
2026 * encapsulation policy.
2027 * Type of param: **struct ipv6_sr_hdr**.
2029 * A call to this helper is susceptible to change the underlaying
2030 * packet buffer. Therefore, at load time, all checks on pointers
2031 * previously done by the verifier are invalidated and must be
2032 * performed again, if the helper is used in combination with
2033 * direct packet access.
2035 * 0 on success, or a negative error in case of failure.
2037 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2039 * This helper is used in programs implementing IR decoding, to
2040 * report a successfully decoded key press with *scancode*,
2041 * *toggle* value in the given *protocol*. The scancode will be
2042 * translated to a keycode using the rc keymap, and reported as
2043 * an input key down event. After a period a key up event is
2044 * generated. This period can be extended by calling either
2045 * **bpf_rc_keydown** () again with the same values, or calling
2046 * **bpf_rc_repeat** ().
2048 * Some protocols include a toggle bit, in case the button was
2049 * released and pressed again between consecutive scancodes.
2051 * The *ctx* should point to the lirc sample as passed into
2054 * The *protocol* is the decoded protocol number (see
2055 * **enum rc_proto** for some predefined values).
2057 * This helper is only available is the kernel was compiled with
2058 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2063 * int bpf_rc_repeat(void *ctx)
2065 * This helper is used in programs implementing IR decoding, to
2066 * report a successfully decoded repeat key message. This delays
2067 * the generation of a key up event for previously generated
2070 * Some IR protocols like NEC have a special IR message for
2071 * repeating last button, for when a button is held down.
2073 * The *ctx* should point to the lirc sample as passed into
2076 * This helper is only available is the kernel was compiled with
2077 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2082 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2084 * Return the cgroup v2 id of the socket associated with the *skb*.
2085 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2086 * helper for cgroup v1 by providing a tag resp. identifier that
2087 * can be matched on or used for map lookups e.g. to implement
2088 * policy. The cgroup v2 id of a given path in the hierarchy is
2089 * exposed in user space through the f_handle API in order to get
2090 * to the same 64-bit id.
2092 * This helper can be used on TC egress path, but not on ingress,
2093 * and is available only if the kernel was compiled with the
2094 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2096 * The id is returned or 0 in case the id could not be retrieved.
2098 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2100 * Return id of cgroup v2 that is ancestor of cgroup associated
2101 * with the *skb* at the *ancestor_level*. The root cgroup is at
2102 * *ancestor_level* zero and each step down the hierarchy
2103 * increments the level. If *ancestor_level* == level of cgroup
2104 * associated with *skb*, then return value will be same as that
2105 * of **bpf_skb_cgroup_id**\ ().
2107 * The helper is useful to implement policies based on cgroups
2108 * that are upper in hierarchy than immediate cgroup associated
2111 * The format of returned id and helper limitations are same as in
2112 * **bpf_skb_cgroup_id**\ ().
2114 * The id is returned or 0 in case the id could not be retrieved.
2116 * u64 bpf_get_current_cgroup_id(void)
2118 * A 64-bit integer containing the current cgroup id based
2119 * on the cgroup within which the current task is running.
2121 * void* get_local_storage(void *map, u64 flags)
2123 * Get the pointer to the local storage area.
2124 * The type and the size of the local storage is defined
2125 * by the *map* argument.
2126 * The *flags* meaning is specific for each map type,
2127 * and has to be 0 for cgroup local storage.
2129 * Depending on the bpf program type, a local storage area
2130 * can be shared between multiple instances of the bpf program,
2131 * running simultaneously.
2133 * A user should care about the synchronization by himself.
2134 * For example, by using the BPF_STX_XADD instruction to alter
2137 * Pointer to the local storage area.
2139 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2141 * Select a SO_REUSEPORT sk from a BPF_MAP_TYPE_REUSEPORT_ARRAY map
2142 * It checks the selected sk is matching the incoming
2143 * request in the skb.
2145 * 0 on success, or a negative error in case of failure.
2147 #define __BPF_FUNC_MAPPER(FN) \
2149 FN(map_lookup_elem), \
2150 FN(map_update_elem), \
2151 FN(map_delete_elem), \
2155 FN(get_prandom_u32), \
2156 FN(get_smp_processor_id), \
2157 FN(skb_store_bytes), \
2158 FN(l3_csum_replace), \
2159 FN(l4_csum_replace), \
2161 FN(clone_redirect), \
2162 FN(get_current_pid_tgid), \
2163 FN(get_current_uid_gid), \
2164 FN(get_current_comm), \
2165 FN(get_cgroup_classid), \
2166 FN(skb_vlan_push), \
2168 FN(skb_get_tunnel_key), \
2169 FN(skb_set_tunnel_key), \
2170 FN(perf_event_read), \
2172 FN(get_route_realm), \
2173 FN(perf_event_output), \
2174 FN(skb_load_bytes), \
2177 FN(skb_get_tunnel_opt), \
2178 FN(skb_set_tunnel_opt), \
2179 FN(skb_change_proto), \
2180 FN(skb_change_type), \
2181 FN(skb_under_cgroup), \
2182 FN(get_hash_recalc), \
2183 FN(get_current_task), \
2184 FN(probe_write_user), \
2185 FN(current_task_under_cgroup), \
2186 FN(skb_change_tail), \
2187 FN(skb_pull_data), \
2189 FN(set_hash_invalid), \
2190 FN(get_numa_node_id), \
2191 FN(skb_change_head), \
2192 FN(xdp_adjust_head), \
2193 FN(probe_read_str), \
2194 FN(get_socket_cookie), \
2195 FN(get_socket_uid), \
2198 FN(skb_adjust_room), \
2200 FN(sk_redirect_map), \
2201 FN(sock_map_update), \
2202 FN(xdp_adjust_meta), \
2203 FN(perf_event_read_value), \
2204 FN(perf_prog_read_value), \
2206 FN(override_return), \
2207 FN(sock_ops_cb_flags_set), \
2208 FN(msg_redirect_map), \
2209 FN(msg_apply_bytes), \
2210 FN(msg_cork_bytes), \
2211 FN(msg_pull_data), \
2213 FN(xdp_adjust_tail), \
2214 FN(skb_get_xfrm_state), \
2216 FN(skb_load_bytes_relative), \
2218 FN(sock_hash_update), \
2219 FN(msg_redirect_hash), \
2220 FN(sk_redirect_hash), \
2221 FN(lwt_push_encap), \
2222 FN(lwt_seg6_store_bytes), \
2223 FN(lwt_seg6_adjust_srh), \
2224 FN(lwt_seg6_action), \
2227 FN(skb_cgroup_id), \
2228 FN(get_current_cgroup_id), \
2229 FN(get_local_storage), \
2230 FN(sk_select_reuseport), \
2231 FN(skb_ancestor_cgroup_id),
2233 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2234 * function eBPF program intends to call
2236 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2238 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2241 #undef __BPF_ENUM_FN
2243 /* All flags used by eBPF helper functions, placed here. */
2245 /* BPF_FUNC_skb_store_bytes flags. */
2246 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2247 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2249 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2250 * First 4 bits are for passing the header field size.
2252 #define BPF_F_HDR_FIELD_MASK 0xfULL
2254 /* BPF_FUNC_l4_csum_replace flags. */
2255 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2256 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2257 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2259 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2260 #define BPF_F_INGRESS (1ULL << 0)
2262 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2263 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2265 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2266 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2267 #define BPF_F_USER_STACK (1ULL << 8)
2268 /* flags used by BPF_FUNC_get_stackid only. */
2269 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2270 #define BPF_F_REUSE_STACKID (1ULL << 10)
2271 /* flags used by BPF_FUNC_get_stack only. */
2272 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2274 /* BPF_FUNC_skb_set_tunnel_key flags. */
2275 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2276 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2277 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2279 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2280 * BPF_FUNC_perf_event_read_value flags.
2282 #define BPF_F_INDEX_MASK 0xffffffffULL
2283 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2284 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2285 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2287 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2288 enum bpf_adj_room_mode {
2292 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2293 enum bpf_hdr_start_off {
2298 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2299 enum bpf_lwt_encap_mode {
2301 BPF_LWT_ENCAP_SEG6_INLINE
2304 /* user accessible mirror of in-kernel sk_buff.
2305 * new fields can only be added to the end of this structure
2311 __u32 queue_mapping;
2317 __u32 ingress_ifindex;
2327 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2329 __u32 remote_ip4; /* Stored in network byte order */
2330 __u32 local_ip4; /* Stored in network byte order */
2331 __u32 remote_ip6[4]; /* Stored in network byte order */
2332 __u32 local_ip6[4]; /* Stored in network byte order */
2333 __u32 remote_port; /* Stored in network byte order */
2334 __u32 local_port; /* stored in host byte order */
2340 struct bpf_tunnel_key {
2344 __u32 remote_ipv6[4];
2348 __u16 tunnel_ext; /* Padding, future use. */
2352 /* user accessible mirror of in-kernel xfrm_state.
2353 * new fields can only be added to the end of this structure
2355 struct bpf_xfrm_state {
2357 __u32 spi; /* Stored in network byte order */
2359 __u16 ext; /* Padding, future use. */
2361 __u32 remote_ipv4; /* Stored in network byte order */
2362 __u32 remote_ipv6[4]; /* Stored in network byte order */
2366 /* Generic BPF return codes which all BPF program types may support.
2367 * The values are binary compatible with their TC_ACT_* counter-part to
2368 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2371 * XDP is handled seprately, see XDP_*.
2379 /* >127 are reserved for prog type specific return codes */
2389 __u32 src_ip4; /* Allows 1,2,4-byte read.
2390 * Stored in network byte order.
2392 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2393 * Stored in network byte order.
2395 __u32 src_port; /* Allows 4-byte read.
2396 * Stored in host byte order
2400 #define XDP_PACKET_HEADROOM 256
2402 /* User return codes for XDP prog type.
2403 * A valid XDP program must return one of these defined values. All other
2404 * return codes are reserved for future use. Unknown return codes will
2405 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2415 /* user accessible metadata for XDP packet hook
2416 * new fields must be added to the end of this structure
2422 /* Below access go through struct xdp_rxq_info */
2423 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2424 __u32 rx_queue_index; /* rxq->queue_index */
2432 /* user accessible metadata for SK_MSG packet hook, new fields must
2433 * be added to the end of this structure
2440 __u32 remote_ip4; /* Stored in network byte order */
2441 __u32 local_ip4; /* Stored in network byte order */
2442 __u32 remote_ip6[4]; /* Stored in network byte order */
2443 __u32 local_ip6[4]; /* Stored in network byte order */
2444 __u32 remote_port; /* Stored in network byte order */
2445 __u32 local_port; /* stored in host byte order */
2448 struct sk_reuseport_md {
2450 * Start of directly accessible data. It begins from
2451 * the tcp/udp header.
2454 void *data_end; /* End of directly accessible data */
2456 * Total length of packet (starting from the tcp/udp header).
2457 * Note that the directly accessible bytes (data_end - data)
2458 * could be less than this "len". Those bytes could be
2459 * indirectly read by a helper "bpf_skb_load_bytes()".
2463 * Eth protocol in the mac header (network byte order). e.g.
2464 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2467 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2468 __u32 bind_inany; /* Is sock bound to an INANY address? */
2469 __u32 hash; /* A hash of the packet 4 tuples */
2472 #define BPF_TAG_SIZE 8
2474 struct bpf_prog_info {
2477 __u8 tag[BPF_TAG_SIZE];
2478 __u32 jited_prog_len;
2479 __u32 xlated_prog_len;
2480 __aligned_u64 jited_prog_insns;
2481 __aligned_u64 xlated_prog_insns;
2482 __u64 load_time; /* ns since boottime */
2483 __u32 created_by_uid;
2485 __aligned_u64 map_ids;
2486 char name[BPF_OBJ_NAME_LEN];
2488 __u32 gpl_compatible:1;
2489 __u32 :31; /* alignment pad */
2492 __u32 nr_jited_ksyms;
2493 __u32 nr_jited_func_lens;
2494 __aligned_u64 jited_ksyms;
2495 __aligned_u64 jited_func_lens;
2496 } __attribute__((aligned(8)));
2498 struct bpf_map_info {
2505 char name[BPF_OBJ_NAME_LEN];
2511 __u32 btf_key_type_id;
2512 __u32 btf_value_type_id;
2513 } __attribute__((aligned(8)));
2515 struct bpf_btf_info {
2519 } __attribute__((aligned(8)));
2521 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2522 * by user and intended to be used by socket (e.g. to bind to, depends on
2523 * attach attach type).
2525 struct bpf_sock_addr {
2526 __u32 user_family; /* Allows 4-byte read, but no write. */
2527 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2528 * Stored in network byte order.
2530 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2531 * Stored in network byte order.
2533 __u32 user_port; /* Allows 4-byte read and write.
2534 * Stored in network byte order
2536 __u32 family; /* Allows 4-byte read, but no write */
2537 __u32 type; /* Allows 4-byte read, but no write */
2538 __u32 protocol; /* Allows 4-byte read, but no write */
2539 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2540 * Stored in network byte order.
2542 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2543 * Stored in network byte order.
2547 /* User bpf_sock_ops struct to access socket values and specify request ops
2548 * and their replies.
2549 * Some of this fields are in network (bigendian) byte order and may need
2550 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2551 * New fields can only be added at the end of this structure
2553 struct bpf_sock_ops {
2556 __u32 args[4]; /* Optionally passed to bpf program */
2557 __u32 reply; /* Returned by bpf program */
2558 __u32 replylong[4]; /* Optionally returned by bpf prog */
2561 __u32 remote_ip4; /* Stored in network byte order */
2562 __u32 local_ip4; /* Stored in network byte order */
2563 __u32 remote_ip6[4]; /* Stored in network byte order */
2564 __u32 local_ip6[4]; /* Stored in network byte order */
2565 __u32 remote_port; /* Stored in network byte order */
2566 __u32 local_port; /* stored in host byte order */
2567 __u32 is_fullsock; /* Some TCP fields are only valid if
2568 * there is a full socket. If not, the
2569 * fields read as zero.
2572 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2573 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2582 __u32 rate_delivered;
2583 __u32 rate_interval_us;
2586 __u32 total_retrans;
2590 __u32 data_segs_out;
2594 __u64 bytes_received;
2598 /* Definitions for bpf_sock_ops_cb_flags */
2599 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2600 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2601 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2602 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2603 * supported cb flags
2606 /* List of known BPF sock_ops operators.
2607 * New entries can only be added at the end
2611 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2612 * -1 if default value should be used
2614 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2615 * window (in packets) or -1 if default
2616 * value should be used
2618 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2619 * active connection is initialized
2621 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2622 * active connection is
2625 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2626 * passive connection is
2629 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2632 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2633 * based on the path and may be
2634 * dependent on the congestion control
2635 * algorithm. In general it indicates
2636 * a congestion threshold. RTTs above
2637 * this indicate congestion
2639 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2640 * Arg1: value of icsk_retransmits
2641 * Arg2: value of icsk_rto
2642 * Arg3: whether RTO has expired
2644 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2645 * Arg1: sequence number of 1st byte
2647 * Arg3: return value of
2648 * tcp_transmit_skb (0 => success)
2650 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2654 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
2655 * socket transition to LISTEN state.
2659 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2660 * changes between the TCP and BPF versions. Ideally this should never happen.
2661 * If it does, we need to add code to convert them before calling
2662 * the BPF sock_ops function.
2665 BPF_TCP_ESTABLISHED = 1,
2675 BPF_TCP_CLOSING, /* Now a valid state */
2676 BPF_TCP_NEW_SYN_RECV,
2678 BPF_TCP_MAX_STATES /* Leave at the end! */
2681 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
2682 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
2684 struct bpf_perf_event_value {
2690 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2691 #define BPF_DEVCG_ACC_READ (1ULL << 1)
2692 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2694 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2695 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2697 struct bpf_cgroup_dev_ctx {
2698 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2704 struct bpf_raw_tracepoint_args {
2708 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
2709 * OUTPUT: Do lookup from egress perspective; default is ingress
2711 #define BPF_FIB_LOOKUP_DIRECT (1U << 0)
2712 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
2715 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
2716 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
2717 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
2718 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
2719 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
2720 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
2721 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
2722 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
2723 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
2726 struct bpf_fib_lookup {
2727 /* input: network family for lookup (AF_INET, AF_INET6)
2728 * output: network family of egress nexthop
2732 /* set if lookup is to consider L4 data - e.g., FIB rules */
2737 /* total length of packet from network header - used for MTU check */
2740 /* input: L3 device index for lookup
2741 * output: device index from FIB lookup
2746 /* inputs to lookup */
2747 __u8 tos; /* AF_INET */
2748 __be32 flowinfo; /* AF_INET6, flow_label + priority */
2750 /* output: metric of fib result (IPv4/IPv6 only) */
2756 __u32 ipv6_src[4]; /* in6_addr; network order */
2759 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2760 * network header. output: bpf_fib_lookup sets to gateway address
2761 * if FIB lookup returns gateway route
2765 __u32 ipv6_dst[4]; /* in6_addr; network order */
2769 __be16 h_vlan_proto;
2771 __u8 smac[6]; /* ETH_ALEN */
2772 __u8 dmac[6]; /* ETH_ALEN */
2775 enum bpf_task_fd_type {
2776 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
2777 BPF_FD_TYPE_TRACEPOINT, /* tp name */
2778 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
2779 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
2780 BPF_FD_TYPE_UPROBE, /* filename + offset */
2781 BPF_FD_TYPE_URETPROBE, /* filename + offset */
2784 #endif /* __LINUX_BPF_H__ */
projects / cross.git / blob