]> git.ipfire.org Git - thirdparty/kernel/stable.git/blob - include/uapi/linux/bpf.h
Merge tag 'drm/tegra/for-5.1-rc5' of git://anongit.freedesktop.org/tegra/linux into...
[thirdparty/kernel/stable.git] / include / uapi / linux / bpf.h
1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3 *
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.
7 */
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
10
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
13
14 /* Extended instruction set based on top of classic BPF */
15
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
19
20 /* ld/ldx fields */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_XADD 0xc0 /* exclusive add */
23
24 /* alu/jmp fields */
25 #define BPF_MOV 0xb0 /* mov reg to reg */
26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
27
28 /* change endianness of a register */
29 #define BPF_END 0xd0 /* flags for endianness conversion: */
30 #define BPF_TO_LE 0x00 /* convert to little-endian */
31 #define BPF_TO_BE 0x08 /* convert to big-endian */
32 #define BPF_FROM_LE BPF_TO_LE
33 #define BPF_FROM_BE BPF_TO_BE
34
35 /* jmp encodings */
36 #define BPF_JNE 0x50 /* jump != */
37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
43 #define BPF_CALL 0x80 /* function call */
44 #define BPF_EXIT 0x90 /* function return */
45
46 /* Register numbers */
47 enum {
48 BPF_REG_0 = 0,
49 BPF_REG_1,
50 BPF_REG_2,
51 BPF_REG_3,
52 BPF_REG_4,
53 BPF_REG_5,
54 BPF_REG_6,
55 BPF_REG_7,
56 BPF_REG_8,
57 BPF_REG_9,
58 BPF_REG_10,
59 __MAX_BPF_REG,
60 };
61
62 /* BPF has 10 general purpose 64-bit registers and stack frame. */
63 #define MAX_BPF_REG __MAX_BPF_REG
64
65 struct bpf_insn {
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
71 };
72
73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74 struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data[0]; /* Arbitrary size */
77 };
78
79 struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
82 };
83
84 /* BPF syscall commands, see bpf(2) man-page for details. */
85 enum bpf_cmd {
86 BPF_MAP_CREATE,
87 BPF_MAP_LOOKUP_ELEM,
88 BPF_MAP_UPDATE_ELEM,
89 BPF_MAP_DELETE_ELEM,
90 BPF_MAP_GET_NEXT_KEY,
91 BPF_PROG_LOAD,
92 BPF_OBJ_PIN,
93 BPF_OBJ_GET,
94 BPF_PROG_ATTACH,
95 BPF_PROG_DETACH,
96 BPF_PROG_TEST_RUN,
97 BPF_PROG_GET_NEXT_ID,
98 BPF_MAP_GET_NEXT_ID,
99 BPF_PROG_GET_FD_BY_ID,
100 BPF_MAP_GET_FD_BY_ID,
101 BPF_OBJ_GET_INFO_BY_FD,
102 BPF_PROG_QUERY,
103 BPF_RAW_TRACEPOINT_OPEN,
104 BPF_BTF_LOAD,
105 BPF_BTF_GET_FD_BY_ID,
106 BPF_TASK_FD_QUERY,
107 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
108 };
109
110 enum bpf_map_type {
111 BPF_MAP_TYPE_UNSPEC,
112 BPF_MAP_TYPE_HASH,
113 BPF_MAP_TYPE_ARRAY,
114 BPF_MAP_TYPE_PROG_ARRAY,
115 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
116 BPF_MAP_TYPE_PERCPU_HASH,
117 BPF_MAP_TYPE_PERCPU_ARRAY,
118 BPF_MAP_TYPE_STACK_TRACE,
119 BPF_MAP_TYPE_CGROUP_ARRAY,
120 BPF_MAP_TYPE_LRU_HASH,
121 BPF_MAP_TYPE_LRU_PERCPU_HASH,
122 BPF_MAP_TYPE_LPM_TRIE,
123 BPF_MAP_TYPE_ARRAY_OF_MAPS,
124 BPF_MAP_TYPE_HASH_OF_MAPS,
125 BPF_MAP_TYPE_DEVMAP,
126 BPF_MAP_TYPE_SOCKMAP,
127 BPF_MAP_TYPE_CPUMAP,
128 BPF_MAP_TYPE_XSKMAP,
129 BPF_MAP_TYPE_SOCKHASH,
130 BPF_MAP_TYPE_CGROUP_STORAGE,
131 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
132 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
133 BPF_MAP_TYPE_QUEUE,
134 BPF_MAP_TYPE_STACK,
135 };
136
137 /* Note that tracing related programs such as
138 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
139 * are not subject to a stable API since kernel internal data
140 * structures can change from release to release and may
141 * therefore break existing tracing BPF programs. Tracing BPF
142 * programs correspond to /a/ specific kernel which is to be
143 * analyzed, and not /a/ specific kernel /and/ all future ones.
144 */
145 enum bpf_prog_type {
146 BPF_PROG_TYPE_UNSPEC,
147 BPF_PROG_TYPE_SOCKET_FILTER,
148 BPF_PROG_TYPE_KPROBE,
149 BPF_PROG_TYPE_SCHED_CLS,
150 BPF_PROG_TYPE_SCHED_ACT,
151 BPF_PROG_TYPE_TRACEPOINT,
152 BPF_PROG_TYPE_XDP,
153 BPF_PROG_TYPE_PERF_EVENT,
154 BPF_PROG_TYPE_CGROUP_SKB,
155 BPF_PROG_TYPE_CGROUP_SOCK,
156 BPF_PROG_TYPE_LWT_IN,
157 BPF_PROG_TYPE_LWT_OUT,
158 BPF_PROG_TYPE_LWT_XMIT,
159 BPF_PROG_TYPE_SOCK_OPS,
160 BPF_PROG_TYPE_SK_SKB,
161 BPF_PROG_TYPE_CGROUP_DEVICE,
162 BPF_PROG_TYPE_SK_MSG,
163 BPF_PROG_TYPE_RAW_TRACEPOINT,
164 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
165 BPF_PROG_TYPE_LWT_SEG6LOCAL,
166 BPF_PROG_TYPE_LIRC_MODE2,
167 BPF_PROG_TYPE_SK_REUSEPORT,
168 BPF_PROG_TYPE_FLOW_DISSECTOR,
169 };
170
171 enum bpf_attach_type {
172 BPF_CGROUP_INET_INGRESS,
173 BPF_CGROUP_INET_EGRESS,
174 BPF_CGROUP_INET_SOCK_CREATE,
175 BPF_CGROUP_SOCK_OPS,
176 BPF_SK_SKB_STREAM_PARSER,
177 BPF_SK_SKB_STREAM_VERDICT,
178 BPF_CGROUP_DEVICE,
179 BPF_SK_MSG_VERDICT,
180 BPF_CGROUP_INET4_BIND,
181 BPF_CGROUP_INET6_BIND,
182 BPF_CGROUP_INET4_CONNECT,
183 BPF_CGROUP_INET6_CONNECT,
184 BPF_CGROUP_INET4_POST_BIND,
185 BPF_CGROUP_INET6_POST_BIND,
186 BPF_CGROUP_UDP4_SENDMSG,
187 BPF_CGROUP_UDP6_SENDMSG,
188 BPF_LIRC_MODE2,
189 BPF_FLOW_DISSECTOR,
190 __MAX_BPF_ATTACH_TYPE
191 };
192
193 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
194
195 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
196 *
197 * NONE(default): No further bpf programs allowed in the subtree.
198 *
199 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
200 * the program in this cgroup yields to sub-cgroup program.
201 *
202 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
203 * that cgroup program gets run in addition to the program in this cgroup.
204 *
205 * Only one program is allowed to be attached to a cgroup with
206 * NONE or BPF_F_ALLOW_OVERRIDE flag.
207 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
208 * release old program and attach the new one. Attach flags has to match.
209 *
210 * Multiple programs are allowed to be attached to a cgroup with
211 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
212 * (those that were attached first, run first)
213 * The programs of sub-cgroup are executed first, then programs of
214 * this cgroup and then programs of parent cgroup.
215 * When children program makes decision (like picking TCP CA or sock bind)
216 * parent program has a chance to override it.
217 *
218 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
219 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
220 * Ex1:
221 * cgrp1 (MULTI progs A, B) ->
222 * cgrp2 (OVERRIDE prog C) ->
223 * cgrp3 (MULTI prog D) ->
224 * cgrp4 (OVERRIDE prog E) ->
225 * cgrp5 (NONE prog F)
226 * the event in cgrp5 triggers execution of F,D,A,B in that order.
227 * if prog F is detached, the execution is E,D,A,B
228 * if prog F and D are detached, the execution is E,A,B
229 * if prog F, E and D are detached, the execution is C,A,B
230 *
231 * All eligible programs are executed regardless of return code from
232 * earlier programs.
233 */
234 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
235 #define BPF_F_ALLOW_MULTI (1U << 1)
236
237 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
238 * verifier will perform strict alignment checking as if the kernel
239 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
240 * and NET_IP_ALIGN defined to 2.
241 */
242 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
243
244 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
245 * verifier will allow any alignment whatsoever. On platforms
246 * with strict alignment requirements for loads ands stores (such
247 * as sparc and mips) the verifier validates that all loads and
248 * stores provably follow this requirement. This flag turns that
249 * checking and enforcement off.
250 *
251 * It is mostly used for testing when we want to validate the
252 * context and memory access aspects of the verifier, but because
253 * of an unaligned access the alignment check would trigger before
254 * the one we are interested in.
255 */
256 #define BPF_F_ANY_ALIGNMENT (1U << 1)
257
258 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
259 #define BPF_PSEUDO_MAP_FD 1
260
261 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
262 * offset to another bpf function
263 */
264 #define BPF_PSEUDO_CALL 1
265
266 /* flags for BPF_MAP_UPDATE_ELEM command */
267 #define BPF_ANY 0 /* create new element or update existing */
268 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
269 #define BPF_EXIST 2 /* update existing element */
270 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
271
272 /* flags for BPF_MAP_CREATE command */
273 #define BPF_F_NO_PREALLOC (1U << 0)
274 /* Instead of having one common LRU list in the
275 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
276 * which can scale and perform better.
277 * Note, the LRU nodes (including free nodes) cannot be moved
278 * across different LRU lists.
279 */
280 #define BPF_F_NO_COMMON_LRU (1U << 1)
281 /* Specify numa node during map creation */
282 #define BPF_F_NUMA_NODE (1U << 2)
283
284 #define BPF_OBJ_NAME_LEN 16U
285
286 /* Flags for accessing BPF object */
287 #define BPF_F_RDONLY (1U << 3)
288 #define BPF_F_WRONLY (1U << 4)
289
290 /* Flag for stack_map, store build_id+offset instead of pointer */
291 #define BPF_F_STACK_BUILD_ID (1U << 5)
292
293 /* Zero-initialize hash function seed. This should only be used for testing. */
294 #define BPF_F_ZERO_SEED (1U << 6)
295
296 /* flags for BPF_PROG_QUERY */
297 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
298
299 enum bpf_stack_build_id_status {
300 /* user space need an empty entry to identify end of a trace */
301 BPF_STACK_BUILD_ID_EMPTY = 0,
302 /* with valid build_id and offset */
303 BPF_STACK_BUILD_ID_VALID = 1,
304 /* couldn't get build_id, fallback to ip */
305 BPF_STACK_BUILD_ID_IP = 2,
306 };
307
308 #define BPF_BUILD_ID_SIZE 20
309 struct bpf_stack_build_id {
310 __s32 status;
311 unsigned char build_id[BPF_BUILD_ID_SIZE];
312 union {
313 __u64 offset;
314 __u64 ip;
315 };
316 };
317
318 union bpf_attr {
319 struct { /* anonymous struct used by BPF_MAP_CREATE command */
320 __u32 map_type; /* one of enum bpf_map_type */
321 __u32 key_size; /* size of key in bytes */
322 __u32 value_size; /* size of value in bytes */
323 __u32 max_entries; /* max number of entries in a map */
324 __u32 map_flags; /* BPF_MAP_CREATE related
325 * flags defined above.
326 */
327 __u32 inner_map_fd; /* fd pointing to the inner map */
328 __u32 numa_node; /* numa node (effective only if
329 * BPF_F_NUMA_NODE is set).
330 */
331 char map_name[BPF_OBJ_NAME_LEN];
332 __u32 map_ifindex; /* ifindex of netdev to create on */
333 __u32 btf_fd; /* fd pointing to a BTF type data */
334 __u32 btf_key_type_id; /* BTF type_id of the key */
335 __u32 btf_value_type_id; /* BTF type_id of the value */
336 };
337
338 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
339 __u32 map_fd;
340 __aligned_u64 key;
341 union {
342 __aligned_u64 value;
343 __aligned_u64 next_key;
344 };
345 __u64 flags;
346 };
347
348 struct { /* anonymous struct used by BPF_PROG_LOAD command */
349 __u32 prog_type; /* one of enum bpf_prog_type */
350 __u32 insn_cnt;
351 __aligned_u64 insns;
352 __aligned_u64 license;
353 __u32 log_level; /* verbosity level of verifier */
354 __u32 log_size; /* size of user buffer */
355 __aligned_u64 log_buf; /* user supplied buffer */
356 __u32 kern_version; /* not used */
357 __u32 prog_flags;
358 char prog_name[BPF_OBJ_NAME_LEN];
359 __u32 prog_ifindex; /* ifindex of netdev to prep for */
360 /* For some prog types expected attach type must be known at
361 * load time to verify attach type specific parts of prog
362 * (context accesses, allowed helpers, etc).
363 */
364 __u32 expected_attach_type;
365 __u32 prog_btf_fd; /* fd pointing to BTF type data */
366 __u32 func_info_rec_size; /* userspace bpf_func_info size */
367 __aligned_u64 func_info; /* func info */
368 __u32 func_info_cnt; /* number of bpf_func_info records */
369 __u32 line_info_rec_size; /* userspace bpf_line_info size */
370 __aligned_u64 line_info; /* line info */
371 __u32 line_info_cnt; /* number of bpf_line_info records */
372 };
373
374 struct { /* anonymous struct used by BPF_OBJ_* commands */
375 __aligned_u64 pathname;
376 __u32 bpf_fd;
377 __u32 file_flags;
378 };
379
380 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
381 __u32 target_fd; /* container object to attach to */
382 __u32 attach_bpf_fd; /* eBPF program to attach */
383 __u32 attach_type;
384 __u32 attach_flags;
385 };
386
387 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
388 __u32 prog_fd;
389 __u32 retval;
390 __u32 data_size_in; /* input: len of data_in */
391 __u32 data_size_out; /* input/output: len of data_out
392 * returns ENOSPC if data_out
393 * is too small.
394 */
395 __aligned_u64 data_in;
396 __aligned_u64 data_out;
397 __u32 repeat;
398 __u32 duration;
399 } test;
400
401 struct { /* anonymous struct used by BPF_*_GET_*_ID */
402 union {
403 __u32 start_id;
404 __u32 prog_id;
405 __u32 map_id;
406 __u32 btf_id;
407 };
408 __u32 next_id;
409 __u32 open_flags;
410 };
411
412 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
413 __u32 bpf_fd;
414 __u32 info_len;
415 __aligned_u64 info;
416 } info;
417
418 struct { /* anonymous struct used by BPF_PROG_QUERY command */
419 __u32 target_fd; /* container object to query */
420 __u32 attach_type;
421 __u32 query_flags;
422 __u32 attach_flags;
423 __aligned_u64 prog_ids;
424 __u32 prog_cnt;
425 } query;
426
427 struct {
428 __u64 name;
429 __u32 prog_fd;
430 } raw_tracepoint;
431
432 struct { /* anonymous struct for BPF_BTF_LOAD */
433 __aligned_u64 btf;
434 __aligned_u64 btf_log_buf;
435 __u32 btf_size;
436 __u32 btf_log_size;
437 __u32 btf_log_level;
438 };
439
440 struct {
441 __u32 pid; /* input: pid */
442 __u32 fd; /* input: fd */
443 __u32 flags; /* input: flags */
444 __u32 buf_len; /* input/output: buf len */
445 __aligned_u64 buf; /* input/output:
446 * tp_name for tracepoint
447 * symbol for kprobe
448 * filename for uprobe
449 */
450 __u32 prog_id; /* output: prod_id */
451 __u32 fd_type; /* output: BPF_FD_TYPE_* */
452 __u64 probe_offset; /* output: probe_offset */
453 __u64 probe_addr; /* output: probe_addr */
454 } task_fd_query;
455 } __attribute__((aligned(8)));
456
457 /* The description below is an attempt at providing documentation to eBPF
458 * developers about the multiple available eBPF helper functions. It can be
459 * parsed and used to produce a manual page. The workflow is the following,
460 * and requires the rst2man utility:
461 *
462 * $ ./scripts/bpf_helpers_doc.py \
463 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
464 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
465 * $ man /tmp/bpf-helpers.7
466 *
467 * Note that in order to produce this external documentation, some RST
468 * formatting is used in the descriptions to get "bold" and "italics" in
469 * manual pages. Also note that the few trailing white spaces are
470 * intentional, removing them would break paragraphs for rst2man.
471 *
472 * Start of BPF helper function descriptions:
473 *
474 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
475 * Description
476 * Perform a lookup in *map* for an entry associated to *key*.
477 * Return
478 * Map value associated to *key*, or **NULL** if no entry was
479 * found.
480 *
481 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
482 * Description
483 * Add or update the value of the entry associated to *key* in
484 * *map* with *value*. *flags* is one of:
485 *
486 * **BPF_NOEXIST**
487 * The entry for *key* must not exist in the map.
488 * **BPF_EXIST**
489 * The entry for *key* must already exist in the map.
490 * **BPF_ANY**
491 * No condition on the existence of the entry for *key*.
492 *
493 * Flag value **BPF_NOEXIST** cannot be used for maps of types
494 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
495 * elements always exist), the helper would return an error.
496 * Return
497 * 0 on success, or a negative error in case of failure.
498 *
499 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
500 * Description
501 * Delete entry with *key* from *map*.
502 * Return
503 * 0 on success, or a negative error in case of failure.
504 *
505 * int bpf_probe_read(void *dst, u32 size, const void *src)
506 * Description
507 * For tracing programs, safely attempt to read *size* bytes from
508 * address *src* and store the data in *dst*.
509 * Return
510 * 0 on success, or a negative error in case of failure.
511 *
512 * u64 bpf_ktime_get_ns(void)
513 * Description
514 * Return the time elapsed since system boot, in nanoseconds.
515 * Return
516 * Current *ktime*.
517 *
518 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
519 * Description
520 * This helper is a "printk()-like" facility for debugging. It
521 * prints a message defined by format *fmt* (of size *fmt_size*)
522 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
523 * available. It can take up to three additional **u64**
524 * arguments (as an eBPF helpers, the total number of arguments is
525 * limited to five).
526 *
527 * Each time the helper is called, it appends a line to the trace.
528 * The format of the trace is customizable, and the exact output
529 * one will get depends on the options set in
530 * *\/sys/kernel/debug/tracing/trace_options* (see also the
531 * *README* file under the same directory). However, it usually
532 * defaults to something like:
533 *
534 * ::
535 *
536 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
537 *
538 * In the above:
539 *
540 * * ``telnet`` is the name of the current task.
541 * * ``470`` is the PID of the current task.
542 * * ``001`` is the CPU number on which the task is
543 * running.
544 * * In ``.N..``, each character refers to a set of
545 * options (whether irqs are enabled, scheduling
546 * options, whether hard/softirqs are running, level of
547 * preempt_disabled respectively). **N** means that
548 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
549 * are set.
550 * * ``419421.045894`` is a timestamp.
551 * * ``0x00000001`` is a fake value used by BPF for the
552 * instruction pointer register.
553 * * ``<formatted msg>`` is the message formatted with
554 * *fmt*.
555 *
556 * The conversion specifiers supported by *fmt* are similar, but
557 * more limited than for printk(). They are **%d**, **%i**,
558 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
559 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
560 * of field, padding with zeroes, etc.) is available, and the
561 * helper will return **-EINVAL** (but print nothing) if it
562 * encounters an unknown specifier.
563 *
564 * Also, note that **bpf_trace_printk**\ () is slow, and should
565 * only be used for debugging purposes. For this reason, a notice
566 * bloc (spanning several lines) is printed to kernel logs and
567 * states that the helper should not be used "for production use"
568 * the first time this helper is used (or more precisely, when
569 * **trace_printk**\ () buffers are allocated). For passing values
570 * to user space, perf events should be preferred.
571 * Return
572 * The number of bytes written to the buffer, or a negative error
573 * in case of failure.
574 *
575 * u32 bpf_get_prandom_u32(void)
576 * Description
577 * Get a pseudo-random number.
578 *
579 * From a security point of view, this helper uses its own
580 * pseudo-random internal state, and cannot be used to infer the
581 * seed of other random functions in the kernel. However, it is
582 * essential to note that the generator used by the helper is not
583 * cryptographically secure.
584 * Return
585 * A random 32-bit unsigned value.
586 *
587 * u32 bpf_get_smp_processor_id(void)
588 * Description
589 * Get the SMP (symmetric multiprocessing) processor id. Note that
590 * all programs run with preemption disabled, which means that the
591 * SMP processor id is stable during all the execution of the
592 * program.
593 * Return
594 * The SMP id of the processor running the program.
595 *
596 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
597 * Description
598 * Store *len* bytes from address *from* into the packet
599 * associated to *skb*, at *offset*. *flags* are a combination of
600 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
601 * checksum for the packet after storing the bytes) and
602 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
603 * **->swhash** and *skb*\ **->l4hash** to 0).
604 *
605 * A call to this helper is susceptible to change the underlaying
606 * packet buffer. Therefore, at load time, all checks on pointers
607 * previously done by the verifier are invalidated and must be
608 * performed again, if the helper is used in combination with
609 * direct packet access.
610 * Return
611 * 0 on success, or a negative error in case of failure.
612 *
613 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
614 * Description
615 * Recompute the layer 3 (e.g. IP) checksum for the packet
616 * associated to *skb*. Computation is incremental, so the helper
617 * must know the former value of the header field that was
618 * modified (*from*), the new value of this field (*to*), and the
619 * number of bytes (2 or 4) for this field, stored in *size*.
620 * Alternatively, it is possible to store the difference between
621 * the previous and the new values of the header field in *to*, by
622 * setting *from* and *size* to 0. For both methods, *offset*
623 * indicates the location of the IP checksum within the packet.
624 *
625 * This helper works in combination with **bpf_csum_diff**\ (),
626 * which does not update the checksum in-place, but offers more
627 * flexibility and can handle sizes larger than 2 or 4 for the
628 * checksum to update.
629 *
630 * A call to this helper is susceptible to change the underlaying
631 * packet buffer. Therefore, at load time, all checks on pointers
632 * previously done by the verifier are invalidated and must be
633 * performed again, if the helper is used in combination with
634 * direct packet access.
635 * Return
636 * 0 on success, or a negative error in case of failure.
637 *
638 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
639 * Description
640 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
641 * packet associated to *skb*. Computation is incremental, so the
642 * helper must know the former value of the header field that was
643 * modified (*from*), the new value of this field (*to*), and the
644 * number of bytes (2 or 4) for this field, stored on the lowest
645 * four bits of *flags*. Alternatively, it is possible to store
646 * the difference between the previous and the new values of the
647 * header field in *to*, by setting *from* and the four lowest
648 * bits of *flags* to 0. For both methods, *offset* indicates the
649 * location of the IP checksum within the packet. In addition to
650 * the size of the field, *flags* can be added (bitwise OR) actual
651 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
652 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
653 * for updates resulting in a null checksum the value is set to
654 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
655 * the checksum is to be computed against a pseudo-header.
656 *
657 * This helper works in combination with **bpf_csum_diff**\ (),
658 * which does not update the checksum in-place, but offers more
659 * flexibility and can handle sizes larger than 2 or 4 for the
660 * checksum to update.
661 *
662 * A call to this helper is susceptible to change the underlaying
663 * packet buffer. Therefore, at load time, all checks on pointers
664 * previously done by the verifier are invalidated and must be
665 * performed again, if the helper is used in combination with
666 * direct packet access.
667 * Return
668 * 0 on success, or a negative error in case of failure.
669 *
670 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
671 * Description
672 * This special helper is used to trigger a "tail call", or in
673 * other words, to jump into another eBPF program. The same stack
674 * frame is used (but values on stack and in registers for the
675 * caller are not accessible to the callee). This mechanism allows
676 * for program chaining, either for raising the maximum number of
677 * available eBPF instructions, or to execute given programs in
678 * conditional blocks. For security reasons, there is an upper
679 * limit to the number of successive tail calls that can be
680 * performed.
681 *
682 * Upon call of this helper, the program attempts to jump into a
683 * program referenced at index *index* in *prog_array_map*, a
684 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
685 * *ctx*, a pointer to the context.
686 *
687 * If the call succeeds, the kernel immediately runs the first
688 * instruction of the new program. This is not a function call,
689 * and it never returns to the previous program. If the call
690 * fails, then the helper has no effect, and the caller continues
691 * to run its subsequent instructions. A call can fail if the
692 * destination program for the jump does not exist (i.e. *index*
693 * is superior to the number of entries in *prog_array_map*), or
694 * if the maximum number of tail calls has been reached for this
695 * chain of programs. This limit is defined in the kernel by the
696 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
697 * which is currently set to 32.
698 * Return
699 * 0 on success, or a negative error in case of failure.
700 *
701 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
702 * Description
703 * Clone and redirect the packet associated to *skb* to another
704 * net device of index *ifindex*. Both ingress and egress
705 * interfaces can be used for redirection. The **BPF_F_INGRESS**
706 * value in *flags* is used to make the distinction (ingress path
707 * is selected if the flag is present, egress path otherwise).
708 * This is the only flag supported for now.
709 *
710 * In comparison with **bpf_redirect**\ () helper,
711 * **bpf_clone_redirect**\ () has the associated cost of
712 * duplicating the packet buffer, but this can be executed out of
713 * the eBPF program. Conversely, **bpf_redirect**\ () is more
714 * efficient, but it is handled through an action code where the
715 * redirection happens only after the eBPF program has returned.
716 *
717 * A call to this helper is susceptible to change the underlaying
718 * packet buffer. Therefore, at load time, all checks on pointers
719 * previously done by the verifier are invalidated and must be
720 * performed again, if the helper is used in combination with
721 * direct packet access.
722 * Return
723 * 0 on success, or a negative error in case of failure.
724 *
725 * u64 bpf_get_current_pid_tgid(void)
726 * Return
727 * A 64-bit integer containing the current tgid and pid, and
728 * created as such:
729 * *current_task*\ **->tgid << 32 \|**
730 * *current_task*\ **->pid**.
731 *
732 * u64 bpf_get_current_uid_gid(void)
733 * Return
734 * A 64-bit integer containing the current GID and UID, and
735 * created as such: *current_gid* **<< 32 \|** *current_uid*.
736 *
737 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
738 * Description
739 * Copy the **comm** attribute of the current task into *buf* of
740 * *size_of_buf*. The **comm** attribute contains the name of
741 * the executable (excluding the path) for the current task. The
742 * *size_of_buf* must be strictly positive. On success, the
743 * helper makes sure that the *buf* is NUL-terminated. On failure,
744 * it is filled with zeroes.
745 * Return
746 * 0 on success, or a negative error in case of failure.
747 *
748 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
749 * Description
750 * Retrieve the classid for the current task, i.e. for the net_cls
751 * cgroup to which *skb* belongs.
752 *
753 * This helper can be used on TC egress path, but not on ingress.
754 *
755 * The net_cls cgroup provides an interface to tag network packets
756 * based on a user-provided identifier for all traffic coming from
757 * the tasks belonging to the related cgroup. See also the related
758 * kernel documentation, available from the Linux sources in file
759 * *Documentation/cgroup-v1/net_cls.txt*.
760 *
761 * The Linux kernel has two versions for cgroups: there are
762 * cgroups v1 and cgroups v2. Both are available to users, who can
763 * use a mixture of them, but note that the net_cls cgroup is for
764 * cgroup v1 only. This makes it incompatible with BPF programs
765 * run on cgroups, which is a cgroup-v2-only feature (a socket can
766 * only hold data for one version of cgroups at a time).
767 *
768 * This helper is only available is the kernel was compiled with
769 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
770 * "**y**" or to "**m**".
771 * Return
772 * The classid, or 0 for the default unconfigured classid.
773 *
774 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
775 * Description
776 * Push a *vlan_tci* (VLAN tag control information) of protocol
777 * *vlan_proto* to the packet associated to *skb*, then update
778 * the checksum. Note that if *vlan_proto* is different from
779 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
780 * be **ETH_P_8021Q**.
781 *
782 * A call to this helper is susceptible to change the underlaying
783 * packet buffer. Therefore, at load time, all checks on pointers
784 * previously done by the verifier are invalidated and must be
785 * performed again, if the helper is used in combination with
786 * direct packet access.
787 * Return
788 * 0 on success, or a negative error in case of failure.
789 *
790 * int bpf_skb_vlan_pop(struct sk_buff *skb)
791 * Description
792 * Pop a VLAN header from the packet associated to *skb*.
793 *
794 * A call to this helper is susceptible to change the underlaying
795 * packet buffer. Therefore, at load time, all checks on pointers
796 * previously done by the verifier are invalidated and must be
797 * performed again, if the helper is used in combination with
798 * direct packet access.
799 * Return
800 * 0 on success, or a negative error in case of failure.
801 *
802 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
803 * Description
804 * Get tunnel metadata. This helper takes a pointer *key* to an
805 * empty **struct bpf_tunnel_key** of **size**, that will be
806 * filled with tunnel metadata for the packet associated to *skb*.
807 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
808 * indicates that the tunnel is based on IPv6 protocol instead of
809 * IPv4.
810 *
811 * The **struct bpf_tunnel_key** is an object that generalizes the
812 * principal parameters used by various tunneling protocols into a
813 * single struct. This way, it can be used to easily make a
814 * decision based on the contents of the encapsulation header,
815 * "summarized" in this struct. In particular, it holds the IP
816 * address of the remote end (IPv4 or IPv6, depending on the case)
817 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
818 * this struct exposes the *key*\ **->tunnel_id**, which is
819 * generally mapped to a VNI (Virtual Network Identifier), making
820 * it programmable together with the **bpf_skb_set_tunnel_key**\
821 * () helper.
822 *
823 * Let's imagine that the following code is part of a program
824 * attached to the TC ingress interface, on one end of a GRE
825 * tunnel, and is supposed to filter out all messages coming from
826 * remote ends with IPv4 address other than 10.0.0.1:
827 *
828 * ::
829 *
830 * int ret;
831 * struct bpf_tunnel_key key = {};
832 *
833 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
834 * if (ret < 0)
835 * return TC_ACT_SHOT; // drop packet
836 *
837 * if (key.remote_ipv4 != 0x0a000001)
838 * return TC_ACT_SHOT; // drop packet
839 *
840 * return TC_ACT_OK; // accept packet
841 *
842 * This interface can also be used with all encapsulation devices
843 * that can operate in "collect metadata" mode: instead of having
844 * one network device per specific configuration, the "collect
845 * metadata" mode only requires a single device where the
846 * configuration can be extracted from this helper.
847 *
848 * This can be used together with various tunnels such as VXLan,
849 * Geneve, GRE or IP in IP (IPIP).
850 * Return
851 * 0 on success, or a negative error in case of failure.
852 *
853 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
854 * Description
855 * Populate tunnel metadata for packet associated to *skb.* The
856 * tunnel metadata is set to the contents of *key*, of *size*. The
857 * *flags* can be set to a combination of the following values:
858 *
859 * **BPF_F_TUNINFO_IPV6**
860 * Indicate that the tunnel is based on IPv6 protocol
861 * instead of IPv4.
862 * **BPF_F_ZERO_CSUM_TX**
863 * For IPv4 packets, add a flag to tunnel metadata
864 * indicating that checksum computation should be skipped
865 * and checksum set to zeroes.
866 * **BPF_F_DONT_FRAGMENT**
867 * Add a flag to tunnel metadata indicating that the
868 * packet should not be fragmented.
869 * **BPF_F_SEQ_NUMBER**
870 * Add a flag to tunnel metadata indicating that a
871 * sequence number should be added to tunnel header before
872 * sending the packet. This flag was added for GRE
873 * encapsulation, but might be used with other protocols
874 * as well in the future.
875 *
876 * Here is a typical usage on the transmit path:
877 *
878 * ::
879 *
880 * struct bpf_tunnel_key key;
881 * populate key ...
882 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
883 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
884 *
885 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
886 * helper for additional information.
887 * Return
888 * 0 on success, or a negative error in case of failure.
889 *
890 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
891 * Description
892 * Read the value of a perf event counter. This helper relies on a
893 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
894 * the perf event counter is selected when *map* is updated with
895 * perf event file descriptors. The *map* is an array whose size
896 * is the number of available CPUs, and each cell contains a value
897 * relative to one CPU. The value to retrieve is indicated by
898 * *flags*, that contains the index of the CPU to look up, masked
899 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
900 * **BPF_F_CURRENT_CPU** to indicate that the value for the
901 * current CPU should be retrieved.
902 *
903 * Note that before Linux 4.13, only hardware perf event can be
904 * retrieved.
905 *
906 * Also, be aware that the newer helper
907 * **bpf_perf_event_read_value**\ () is recommended over
908 * **bpf_perf_event_read**\ () in general. The latter has some ABI
909 * quirks where error and counter value are used as a return code
910 * (which is wrong to do since ranges may overlap). This issue is
911 * fixed with **bpf_perf_event_read_value**\ (), which at the same
912 * time provides more features over the **bpf_perf_event_read**\
913 * () interface. Please refer to the description of
914 * **bpf_perf_event_read_value**\ () for details.
915 * Return
916 * The value of the perf event counter read from the map, or a
917 * negative error code in case of failure.
918 *
919 * int bpf_redirect(u32 ifindex, u64 flags)
920 * Description
921 * Redirect the packet to another net device of index *ifindex*.
922 * This helper is somewhat similar to **bpf_clone_redirect**\
923 * (), except that the packet is not cloned, which provides
924 * increased performance.
925 *
926 * Except for XDP, both ingress and egress interfaces can be used
927 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
928 * to make the distinction (ingress path is selected if the flag
929 * is present, egress path otherwise). Currently, XDP only
930 * supports redirection to the egress interface, and accepts no
931 * flag at all.
932 *
933 * The same effect can be attained with the more generic
934 * **bpf_redirect_map**\ (), which requires specific maps to be
935 * used but offers better performance.
936 * Return
937 * For XDP, the helper returns **XDP_REDIRECT** on success or
938 * **XDP_ABORTED** on error. For other program types, the values
939 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
940 * error.
941 *
942 * u32 bpf_get_route_realm(struct sk_buff *skb)
943 * Description
944 * Retrieve the realm or the route, that is to say the
945 * **tclassid** field of the destination for the *skb*. The
946 * indentifier retrieved is a user-provided tag, similar to the
947 * one used with the net_cls cgroup (see description for
948 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
949 * held by a route (a destination entry), not by a task.
950 *
951 * Retrieving this identifier works with the clsact TC egress hook
952 * (see also **tc-bpf(8)**), or alternatively on conventional
953 * classful egress qdiscs, but not on TC ingress path. In case of
954 * clsact TC egress hook, this has the advantage that, internally,
955 * the destination entry has not been dropped yet in the transmit
956 * path. Therefore, the destination entry does not need to be
957 * artificially held via **netif_keep_dst**\ () for a classful
958 * qdisc until the *skb* is freed.
959 *
960 * This helper is available only if the kernel was compiled with
961 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
962 * Return
963 * The realm of the route for the packet associated to *skb*, or 0
964 * if none was found.
965 *
966 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
967 * Description
968 * Write raw *data* blob into a special BPF perf event held by
969 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
970 * event must have the following attributes: **PERF_SAMPLE_RAW**
971 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
972 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
973 *
974 * The *flags* are used to indicate the index in *map* for which
975 * the value must be put, masked with **BPF_F_INDEX_MASK**.
976 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
977 * to indicate that the index of the current CPU core should be
978 * used.
979 *
980 * The value to write, of *size*, is passed through eBPF stack and
981 * pointed by *data*.
982 *
983 * The context of the program *ctx* needs also be passed to the
984 * helper.
985 *
986 * On user space, a program willing to read the values needs to
987 * call **perf_event_open**\ () on the perf event (either for
988 * one or for all CPUs) and to store the file descriptor into the
989 * *map*. This must be done before the eBPF program can send data
990 * into it. An example is available in file
991 * *samples/bpf/trace_output_user.c* in the Linux kernel source
992 * tree (the eBPF program counterpart is in
993 * *samples/bpf/trace_output_kern.c*).
994 *
995 * **bpf_perf_event_output**\ () achieves better performance
996 * than **bpf_trace_printk**\ () for sharing data with user
997 * space, and is much better suitable for streaming data from eBPF
998 * programs.
999 *
1000 * Note that this helper is not restricted to tracing use cases
1001 * and can be used with programs attached to TC or XDP as well,
1002 * where it allows for passing data to user space listeners. Data
1003 * can be:
1004 *
1005 * * Only custom structs,
1006 * * Only the packet payload, or
1007 * * A combination of both.
1008 * Return
1009 * 0 on success, or a negative error in case of failure.
1010 *
1011 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1012 * Description
1013 * This helper was provided as an easy way to load data from a
1014 * packet. It can be used to load *len* bytes from *offset* from
1015 * the packet associated to *skb*, into the buffer pointed by
1016 * *to*.
1017 *
1018 * Since Linux 4.7, usage of this helper has mostly been replaced
1019 * by "direct packet access", enabling packet data to be
1020 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1021 * pointing respectively to the first byte of packet data and to
1022 * the byte after the last byte of packet data. However, it
1023 * remains useful if one wishes to read large quantities of data
1024 * at once from a packet into the eBPF stack.
1025 * Return
1026 * 0 on success, or a negative error in case of failure.
1027 *
1028 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1029 * Description
1030 * Walk a user or a kernel stack and return its id. To achieve
1031 * this, the helper needs *ctx*, which is a pointer to the context
1032 * on which the tracing program is executed, and a pointer to a
1033 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1034 *
1035 * The last argument, *flags*, holds the number of stack frames to
1036 * skip (from 0 to 255), masked with
1037 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1038 * a combination of the following flags:
1039 *
1040 * **BPF_F_USER_STACK**
1041 * Collect a user space stack instead of a kernel stack.
1042 * **BPF_F_FAST_STACK_CMP**
1043 * Compare stacks by hash only.
1044 * **BPF_F_REUSE_STACKID**
1045 * If two different stacks hash into the same *stackid*,
1046 * discard the old one.
1047 *
1048 * The stack id retrieved is a 32 bit long integer handle which
1049 * can be further combined with other data (including other stack
1050 * ids) and used as a key into maps. This can be useful for
1051 * generating a variety of graphs (such as flame graphs or off-cpu
1052 * graphs).
1053 *
1054 * For walking a stack, this helper is an improvement over
1055 * **bpf_probe_read**\ (), which can be used with unrolled loops
1056 * but is not efficient and consumes a lot of eBPF instructions.
1057 * Instead, **bpf_get_stackid**\ () can collect up to
1058 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1059 * this limit can be controlled with the **sysctl** program, and
1060 * that it should be manually increased in order to profile long
1061 * user stacks (such as stacks for Java programs). To do so, use:
1062 *
1063 * ::
1064 *
1065 * # sysctl kernel.perf_event_max_stack=<new value>
1066 * Return
1067 * The positive or null stack id on success, or a negative error
1068 * in case of failure.
1069 *
1070 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1071 * Description
1072 * Compute a checksum difference, from the raw buffer pointed by
1073 * *from*, of length *from_size* (that must be a multiple of 4),
1074 * towards the raw buffer pointed by *to*, of size *to_size*
1075 * (same remark). An optional *seed* can be added to the value
1076 * (this can be cascaded, the seed may come from a previous call
1077 * to the helper).
1078 *
1079 * This is flexible enough to be used in several ways:
1080 *
1081 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1082 * checksum, it can be used when pushing new data.
1083 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1084 * checksum, it can be used when removing data from a packet.
1085 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1086 * can be used to compute a diff. Note that *from_size* and
1087 * *to_size* do not need to be equal.
1088 *
1089 * This helper can be used in combination with
1090 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1091 * which one can feed in the difference computed with
1092 * **bpf_csum_diff**\ ().
1093 * Return
1094 * The checksum result, or a negative error code in case of
1095 * failure.
1096 *
1097 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1098 * Description
1099 * Retrieve tunnel options metadata for the packet associated to
1100 * *skb*, and store the raw tunnel option data to the buffer *opt*
1101 * of *size*.
1102 *
1103 * This helper can be used with encapsulation devices that can
1104 * operate in "collect metadata" mode (please refer to the related
1105 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1106 * more details). A particular example where this can be used is
1107 * in combination with the Geneve encapsulation protocol, where it
1108 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1109 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1110 * the eBPF program. This allows for full customization of these
1111 * headers.
1112 * Return
1113 * The size of the option data retrieved.
1114 *
1115 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1116 * Description
1117 * Set tunnel options metadata for the packet associated to *skb*
1118 * to the option data contained in the raw buffer *opt* of *size*.
1119 *
1120 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1121 * helper for additional information.
1122 * Return
1123 * 0 on success, or a negative error in case of failure.
1124 *
1125 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1126 * Description
1127 * Change the protocol of the *skb* to *proto*. Currently
1128 * supported are transition from IPv4 to IPv6, and from IPv6 to
1129 * IPv4. The helper takes care of the groundwork for the
1130 * transition, including resizing the socket buffer. The eBPF
1131 * program is expected to fill the new headers, if any, via
1132 * **skb_store_bytes**\ () and to recompute the checksums with
1133 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1134 * (). The main case for this helper is to perform NAT64
1135 * operations out of an eBPF program.
1136 *
1137 * Internally, the GSO type is marked as dodgy so that headers are
1138 * checked and segments are recalculated by the GSO/GRO engine.
1139 * The size for GSO target is adapted as well.
1140 *
1141 * All values for *flags* are reserved for future usage, and must
1142 * be left at zero.
1143 *
1144 * A call to this helper is susceptible to change the underlaying
1145 * packet buffer. Therefore, at load time, all checks on pointers
1146 * previously done by the verifier are invalidated and must be
1147 * performed again, if the helper is used in combination with
1148 * direct packet access.
1149 * Return
1150 * 0 on success, or a negative error in case of failure.
1151 *
1152 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1153 * Description
1154 * Change the packet type for the packet associated to *skb*. This
1155 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1156 * the eBPF program does not have a write access to *skb*\
1157 * **->pkt_type** beside this helper. Using a helper here allows
1158 * for graceful handling of errors.
1159 *
1160 * The major use case is to change incoming *skb*s to
1161 * **PACKET_HOST** in a programmatic way instead of having to
1162 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1163 * example.
1164 *
1165 * Note that *type* only allows certain values. At this time, they
1166 * are:
1167 *
1168 * **PACKET_HOST**
1169 * Packet is for us.
1170 * **PACKET_BROADCAST**
1171 * Send packet to all.
1172 * **PACKET_MULTICAST**
1173 * Send packet to group.
1174 * **PACKET_OTHERHOST**
1175 * Send packet to someone else.
1176 * Return
1177 * 0 on success, or a negative error in case of failure.
1178 *
1179 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1180 * Description
1181 * Check whether *skb* is a descendant of the cgroup2 held by
1182 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1183 * Return
1184 * The return value depends on the result of the test, and can be:
1185 *
1186 * * 0, if the *skb* failed the cgroup2 descendant test.
1187 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1188 * * A negative error code, if an error occurred.
1189 *
1190 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1191 * Description
1192 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1193 * not set, in particular if the hash was cleared due to mangling,
1194 * recompute this hash. Later accesses to the hash can be done
1195 * directly with *skb*\ **->hash**.
1196 *
1197 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1198 * prototype with **bpf_skb_change_proto**\ (), or calling
1199 * **bpf_skb_store_bytes**\ () with the
1200 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1201 * the hash and to trigger a new computation for the next call to
1202 * **bpf_get_hash_recalc**\ ().
1203 * Return
1204 * The 32-bit hash.
1205 *
1206 * u64 bpf_get_current_task(void)
1207 * Return
1208 * A pointer to the current task struct.
1209 *
1210 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1211 * Description
1212 * Attempt in a safe way to write *len* bytes from the buffer
1213 * *src* to *dst* in memory. It only works for threads that are in
1214 * user context, and *dst* must be a valid user space address.
1215 *
1216 * This helper should not be used to implement any kind of
1217 * security mechanism because of TOC-TOU attacks, but rather to
1218 * debug, divert, and manipulate execution of semi-cooperative
1219 * processes.
1220 *
1221 * Keep in mind that this feature is meant for experiments, and it
1222 * has a risk of crashing the system and running programs.
1223 * Therefore, when an eBPF program using this helper is attached,
1224 * a warning including PID and process name is printed to kernel
1225 * logs.
1226 * Return
1227 * 0 on success, or a negative error in case of failure.
1228 *
1229 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1230 * Description
1231 * Check whether the probe is being run is the context of a given
1232 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1233 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1234 * Return
1235 * The return value depends on the result of the test, and can be:
1236 *
1237 * * 0, if the *skb* task belongs to the cgroup2.
1238 * * 1, if the *skb* task does not belong to the cgroup2.
1239 * * A negative error code, if an error occurred.
1240 *
1241 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1242 * Description
1243 * Resize (trim or grow) the packet associated to *skb* to the
1244 * new *len*. The *flags* are reserved for future usage, and must
1245 * be left at zero.
1246 *
1247 * The basic idea is that the helper performs the needed work to
1248 * change the size of the packet, then the eBPF program rewrites
1249 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1250 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1251 * and others. This helper is a slow path utility intended for
1252 * replies with control messages. And because it is targeted for
1253 * slow path, the helper itself can afford to be slow: it
1254 * implicitly linearizes, unclones and drops offloads from the
1255 * *skb*.
1256 *
1257 * A call to this helper is susceptible to change the underlaying
1258 * packet buffer. Therefore, at load time, all checks on pointers
1259 * previously done by the verifier are invalidated and must be
1260 * performed again, if the helper is used in combination with
1261 * direct packet access.
1262 * Return
1263 * 0 on success, or a negative error in case of failure.
1264 *
1265 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1266 * Description
1267 * Pull in non-linear data in case the *skb* is non-linear and not
1268 * all of *len* are part of the linear section. Make *len* bytes
1269 * from *skb* readable and writable. If a zero value is passed for
1270 * *len*, then the whole length of the *skb* is pulled.
1271 *
1272 * This helper is only needed for reading and writing with direct
1273 * packet access.
1274 *
1275 * For direct packet access, testing that offsets to access
1276 * are within packet boundaries (test on *skb*\ **->data_end**) is
1277 * susceptible to fail if offsets are invalid, or if the requested
1278 * data is in non-linear parts of the *skb*. On failure the
1279 * program can just bail out, or in the case of a non-linear
1280 * buffer, use a helper to make the data available. The
1281 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1282 * the data. Another one consists in using **bpf_skb_pull_data**
1283 * to pull in once the non-linear parts, then retesting and
1284 * eventually access the data.
1285 *
1286 * At the same time, this also makes sure the *skb* is uncloned,
1287 * which is a necessary condition for direct write. As this needs
1288 * to be an invariant for the write part only, the verifier
1289 * detects writes and adds a prologue that is calling
1290 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1291 * the very beginning in case it is indeed cloned.
1292 *
1293 * A call to this helper is susceptible to change the underlaying
1294 * packet buffer. Therefore, at load time, all checks on pointers
1295 * previously done by the verifier are invalidated and must be
1296 * performed again, if the helper is used in combination with
1297 * direct packet access.
1298 * Return
1299 * 0 on success, or a negative error in case of failure.
1300 *
1301 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1302 * Description
1303 * Add the checksum *csum* into *skb*\ **->csum** in case the
1304 * driver has supplied a checksum for the entire packet into that
1305 * field. Return an error otherwise. This helper is intended to be
1306 * used in combination with **bpf_csum_diff**\ (), in particular
1307 * when the checksum needs to be updated after data has been
1308 * written into the packet through direct packet access.
1309 * Return
1310 * The checksum on success, or a negative error code in case of
1311 * failure.
1312 *
1313 * void bpf_set_hash_invalid(struct sk_buff *skb)
1314 * Description
1315 * Invalidate the current *skb*\ **->hash**. It can be used after
1316 * mangling on headers through direct packet access, in order to
1317 * indicate that the hash is outdated and to trigger a
1318 * recalculation the next time the kernel tries to access this
1319 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1320 *
1321 * int bpf_get_numa_node_id(void)
1322 * Description
1323 * Return the id of the current NUMA node. The primary use case
1324 * for this helper is the selection of sockets for the local NUMA
1325 * node, when the program is attached to sockets using the
1326 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1327 * but the helper is also available to other eBPF program types,
1328 * similarly to **bpf_get_smp_processor_id**\ ().
1329 * Return
1330 * The id of current NUMA node.
1331 *
1332 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1333 * Description
1334 * Grows headroom of packet associated to *skb* and adjusts the
1335 * offset of the MAC header accordingly, adding *len* bytes of
1336 * space. It automatically extends and reallocates memory as
1337 * required.
1338 *
1339 * This helper can be used on a layer 3 *skb* to push a MAC header
1340 * for redirection into a layer 2 device.
1341 *
1342 * All values for *flags* are reserved for future usage, and must
1343 * be left at zero.
1344 *
1345 * A call to this helper is susceptible to change the underlaying
1346 * packet buffer. Therefore, at load time, all checks on pointers
1347 * previously done by the verifier are invalidated and must be
1348 * performed again, if the helper is used in combination with
1349 * direct packet access.
1350 * Return
1351 * 0 on success, or a negative error in case of failure.
1352 *
1353 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1354 * Description
1355 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1356 * it is possible to use a negative value for *delta*. This helper
1357 * can be used to prepare the packet for pushing or popping
1358 * headers.
1359 *
1360 * A call to this helper is susceptible to change the underlaying
1361 * packet buffer. Therefore, at load time, all checks on pointers
1362 * previously done by the verifier are invalidated and must be
1363 * performed again, if the helper is used in combination with
1364 * direct packet access.
1365 * Return
1366 * 0 on success, or a negative error in case of failure.
1367 *
1368 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1369 * Description
1370 * Copy a NUL terminated string from an unsafe address
1371 * *unsafe_ptr* to *dst*. The *size* should include the
1372 * terminating NUL byte. In case the string length is smaller than
1373 * *size*, the target is not padded with further NUL bytes. If the
1374 * string length is larger than *size*, just *size*-1 bytes are
1375 * copied and the last byte is set to NUL.
1376 *
1377 * On success, the length of the copied string is returned. This
1378 * makes this helper useful in tracing programs for reading
1379 * strings, and more importantly to get its length at runtime. See
1380 * the following snippet:
1381 *
1382 * ::
1383 *
1384 * SEC("kprobe/sys_open")
1385 * void bpf_sys_open(struct pt_regs *ctx)
1386 * {
1387 * char buf[PATHLEN]; // PATHLEN is defined to 256
1388 * int res = bpf_probe_read_str(buf, sizeof(buf),
1389 * ctx->di);
1390 *
1391 * // Consume buf, for example push it to
1392 * // userspace via bpf_perf_event_output(); we
1393 * // can use res (the string length) as event
1394 * // size, after checking its boundaries.
1395 * }
1396 *
1397 * In comparison, using **bpf_probe_read()** helper here instead
1398 * to read the string would require to estimate the length at
1399 * compile time, and would often result in copying more memory
1400 * than necessary.
1401 *
1402 * Another useful use case is when parsing individual process
1403 * arguments or individual environment variables navigating
1404 * *current*\ **->mm->arg_start** and *current*\
1405 * **->mm->env_start**: using this helper and the return value,
1406 * one can quickly iterate at the right offset of the memory area.
1407 * Return
1408 * On success, the strictly positive length of the string,
1409 * including the trailing NUL character. On error, a negative
1410 * value.
1411 *
1412 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1413 * Description
1414 * If the **struct sk_buff** pointed by *skb* has a known socket,
1415 * retrieve the cookie (generated by the kernel) of this socket.
1416 * If no cookie has been set yet, generate a new cookie. Once
1417 * generated, the socket cookie remains stable for the life of the
1418 * socket. This helper can be useful for monitoring per socket
1419 * networking traffic statistics as it provides a unique socket
1420 * identifier per namespace.
1421 * Return
1422 * A 8-byte long non-decreasing number on success, or 0 if the
1423 * socket field is missing inside *skb*.
1424 *
1425 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1426 * Description
1427 * Equivalent to bpf_get_socket_cookie() helper that accepts
1428 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1429 * Return
1430 * A 8-byte long non-decreasing number.
1431 *
1432 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1433 * Description
1434 * Equivalent to bpf_get_socket_cookie() helper that accepts
1435 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1436 * Return
1437 * A 8-byte long non-decreasing number.
1438 *
1439 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1440 * Return
1441 * The owner UID of the socket associated to *skb*. If the socket
1442 * is **NULL**, or if it is not a full socket (i.e. if it is a
1443 * time-wait or a request socket instead), **overflowuid** value
1444 * is returned (note that **overflowuid** might also be the actual
1445 * UID value for the socket).
1446 *
1447 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1448 * Description
1449 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1450 * to value *hash*.
1451 * Return
1452 * 0
1453 *
1454 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1455 * Description
1456 * Emulate a call to **setsockopt()** on the socket associated to
1457 * *bpf_socket*, which must be a full socket. The *level* at
1458 * which the option resides and the name *optname* of the option
1459 * must be specified, see **setsockopt(2)** for more information.
1460 * The option value of length *optlen* is pointed by *optval*.
1461 *
1462 * This helper actually implements a subset of **setsockopt()**.
1463 * It supports the following *level*\ s:
1464 *
1465 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1466 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1467 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1468 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1469 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1470 * **TCP_BPF_SNDCWND_CLAMP**.
1471 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1472 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1473 * Return
1474 * 0 on success, or a negative error in case of failure.
1475 *
1476 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1477 * Description
1478 * Grow or shrink the room for data in the packet associated to
1479 * *skb* by *len_diff*, and according to the selected *mode*.
1480 *
1481 * There is a single supported mode at this time:
1482 *
1483 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1484 * (room space is added or removed below the layer 3 header).
1485 *
1486 * All values for *flags* are reserved for future usage, and must
1487 * be left at zero.
1488 *
1489 * A call to this helper is susceptible to change the underlaying
1490 * packet buffer. Therefore, at load time, all checks on pointers
1491 * previously done by the verifier are invalidated and must be
1492 * performed again, if the helper is used in combination with
1493 * direct packet access.
1494 * Return
1495 * 0 on success, or a negative error in case of failure.
1496 *
1497 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1498 * Description
1499 * Redirect the packet to the endpoint referenced by *map* at
1500 * index *key*. Depending on its type, this *map* can contain
1501 * references to net devices (for forwarding packets through other
1502 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1503 * but this is only implemented for native XDP (with driver
1504 * support) as of this writing).
1505 *
1506 * All values for *flags* are reserved for future usage, and must
1507 * be left at zero.
1508 *
1509 * When used to redirect packets to net devices, this helper
1510 * provides a high performance increase over **bpf_redirect**\ ().
1511 * This is due to various implementation details of the underlying
1512 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1513 * () tries to send packet as a "bulk" to the device.
1514 * Return
1515 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1516 *
1517 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1518 * Description
1519 * Redirect the packet to the socket referenced by *map* (of type
1520 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1521 * egress interfaces can be used for redirection. The
1522 * **BPF_F_INGRESS** value in *flags* is used to make the
1523 * distinction (ingress path is selected if the flag is present,
1524 * egress path otherwise). This is the only flag supported for now.
1525 * Return
1526 * **SK_PASS** on success, or **SK_DROP** on error.
1527 *
1528 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1529 * Description
1530 * Add an entry to, or update a *map* referencing sockets. The
1531 * *skops* is used as a new value for the entry associated to
1532 * *key*. *flags* is one of:
1533 *
1534 * **BPF_NOEXIST**
1535 * The entry for *key* must not exist in the map.
1536 * **BPF_EXIST**
1537 * The entry for *key* must already exist in the map.
1538 * **BPF_ANY**
1539 * No condition on the existence of the entry for *key*.
1540 *
1541 * If the *map* has eBPF programs (parser and verdict), those will
1542 * be inherited by the socket being added. If the socket is
1543 * already attached to eBPF programs, this results in an error.
1544 * Return
1545 * 0 on success, or a negative error in case of failure.
1546 *
1547 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1548 * Description
1549 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1550 * *delta* (which can be positive or negative). Note that this
1551 * operation modifies the address stored in *xdp_md*\ **->data**,
1552 * so the latter must be loaded only after the helper has been
1553 * called.
1554 *
1555 * The use of *xdp_md*\ **->data_meta** is optional and programs
1556 * are not required to use it. The rationale is that when the
1557 * packet is processed with XDP (e.g. as DoS filter), it is
1558 * possible to push further meta data along with it before passing
1559 * to the stack, and to give the guarantee that an ingress eBPF
1560 * program attached as a TC classifier on the same device can pick
1561 * this up for further post-processing. Since TC works with socket
1562 * buffers, it remains possible to set from XDP the **mark** or
1563 * **priority** pointers, or other pointers for the socket buffer.
1564 * Having this scratch space generic and programmable allows for
1565 * more flexibility as the user is free to store whatever meta
1566 * data they need.
1567 *
1568 * A call to this helper is susceptible to change the underlaying
1569 * packet buffer. Therefore, at load time, all checks on pointers
1570 * previously done by the verifier are invalidated and must be
1571 * performed again, if the helper is used in combination with
1572 * direct packet access.
1573 * Return
1574 * 0 on success, or a negative error in case of failure.
1575 *
1576 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1577 * Description
1578 * Read the value of a perf event counter, and store it into *buf*
1579 * of size *buf_size*. This helper relies on a *map* of type
1580 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1581 * counter is selected when *map* is updated with perf event file
1582 * descriptors. The *map* is an array whose size is the number of
1583 * available CPUs, and each cell contains a value relative to one
1584 * CPU. The value to retrieve is indicated by *flags*, that
1585 * contains the index of the CPU to look up, masked with
1586 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1587 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1588 * current CPU should be retrieved.
1589 *
1590 * This helper behaves in a way close to
1591 * **bpf_perf_event_read**\ () helper, save that instead of
1592 * just returning the value observed, it fills the *buf*
1593 * structure. This allows for additional data to be retrieved: in
1594 * particular, the enabled and running times (in *buf*\
1595 * **->enabled** and *buf*\ **->running**, respectively) are
1596 * copied. In general, **bpf_perf_event_read_value**\ () is
1597 * recommended over **bpf_perf_event_read**\ (), which has some
1598 * ABI issues and provides fewer functionalities.
1599 *
1600 * These values are interesting, because hardware PMU (Performance
1601 * Monitoring Unit) counters are limited resources. When there are
1602 * more PMU based perf events opened than available counters,
1603 * kernel will multiplex these events so each event gets certain
1604 * percentage (but not all) of the PMU time. In case that
1605 * multiplexing happens, the number of samples or counter value
1606 * will not reflect the case compared to when no multiplexing
1607 * occurs. This makes comparison between different runs difficult.
1608 * Typically, the counter value should be normalized before
1609 * comparing to other experiments. The usual normalization is done
1610 * as follows.
1611 *
1612 * ::
1613 *
1614 * normalized_counter = counter * t_enabled / t_running
1615 *
1616 * Where t_enabled is the time enabled for event and t_running is
1617 * the time running for event since last normalization. The
1618 * enabled and running times are accumulated since the perf event
1619 * open. To achieve scaling factor between two invocations of an
1620 * eBPF program, users can can use CPU id as the key (which is
1621 * typical for perf array usage model) to remember the previous
1622 * value and do the calculation inside the eBPF program.
1623 * Return
1624 * 0 on success, or a negative error in case of failure.
1625 *
1626 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1627 * Description
1628 * For en eBPF program attached to a perf event, retrieve the
1629 * value of the event counter associated to *ctx* and store it in
1630 * the structure pointed by *buf* and of size *buf_size*. Enabled
1631 * and running times are also stored in the structure (see
1632 * description of helper **bpf_perf_event_read_value**\ () for
1633 * more details).
1634 * Return
1635 * 0 on success, or a negative error in case of failure.
1636 *
1637 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1638 * Description
1639 * Emulate a call to **getsockopt()** on the socket associated to
1640 * *bpf_socket*, which must be a full socket. The *level* at
1641 * which the option resides and the name *optname* of the option
1642 * must be specified, see **getsockopt(2)** for more information.
1643 * The retrieved value is stored in the structure pointed by
1644 * *opval* and of length *optlen*.
1645 *
1646 * This helper actually implements a subset of **getsockopt()**.
1647 * It supports the following *level*\ s:
1648 *
1649 * * **IPPROTO_TCP**, which supports *optname*
1650 * **TCP_CONGESTION**.
1651 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1652 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1653 * Return
1654 * 0 on success, or a negative error in case of failure.
1655 *
1656 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1657 * Description
1658 * Used for error injection, this helper uses kprobes to override
1659 * the return value of the probed function, and to set it to *rc*.
1660 * The first argument is the context *regs* on which the kprobe
1661 * works.
1662 *
1663 * This helper works by setting setting the PC (program counter)
1664 * to an override function which is run in place of the original
1665 * probed function. This means the probed function is not run at
1666 * all. The replacement function just returns with the required
1667 * value.
1668 *
1669 * This helper has security implications, and thus is subject to
1670 * restrictions. It is only available if the kernel was compiled
1671 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1672 * option, and in this case it only works on functions tagged with
1673 * **ALLOW_ERROR_INJECTION** in the kernel code.
1674 *
1675 * Also, the helper is only available for the architectures having
1676 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1677 * x86 architecture is the only one to support this feature.
1678 * Return
1679 * 0
1680 *
1681 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1682 * Description
1683 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1684 * for the full TCP socket associated to *bpf_sock_ops* to
1685 * *argval*.
1686 *
1687 * The primary use of this field is to determine if there should
1688 * be calls to eBPF programs of type
1689 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1690 * code. A program of the same type can change its value, per
1691 * connection and as necessary, when the connection is
1692 * established. This field is directly accessible for reading, but
1693 * this helper must be used for updates in order to return an
1694 * error if an eBPF program tries to set a callback that is not
1695 * supported in the current kernel.
1696 *
1697 * The supported callback values that *argval* can combine are:
1698 *
1699 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1700 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1701 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1702 *
1703 * Here are some examples of where one could call such eBPF
1704 * program:
1705 *
1706 * * When RTO fires.
1707 * * When a packet is retransmitted.
1708 * * When the connection terminates.
1709 * * When a packet is sent.
1710 * * When a packet is received.
1711 * Return
1712 * Code **-EINVAL** if the socket is not a full TCP socket;
1713 * otherwise, a positive number containing the bits that could not
1714 * be set is returned (which comes down to 0 if all bits were set
1715 * as required).
1716 *
1717 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1718 * Description
1719 * This helper is used in programs implementing policies at the
1720 * socket level. If the message *msg* is allowed to pass (i.e. if
1721 * the verdict eBPF program returns **SK_PASS**), redirect it to
1722 * the socket referenced by *map* (of type
1723 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1724 * egress interfaces can be used for redirection. The
1725 * **BPF_F_INGRESS** value in *flags* is used to make the
1726 * distinction (ingress path is selected if the flag is present,
1727 * egress path otherwise). This is the only flag supported for now.
1728 * Return
1729 * **SK_PASS** on success, or **SK_DROP** on error.
1730 *
1731 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1732 * Description
1733 * For socket policies, apply the verdict of the eBPF program to
1734 * the next *bytes* (number of bytes) of message *msg*.
1735 *
1736 * For example, this helper can be used in the following cases:
1737 *
1738 * * A single **sendmsg**\ () or **sendfile**\ () system call
1739 * contains multiple logical messages that the eBPF program is
1740 * supposed to read and for which it should apply a verdict.
1741 * * An eBPF program only cares to read the first *bytes* of a
1742 * *msg*. If the message has a large payload, then setting up
1743 * and calling the eBPF program repeatedly for all bytes, even
1744 * though the verdict is already known, would create unnecessary
1745 * overhead.
1746 *
1747 * When called from within an eBPF program, the helper sets a
1748 * counter internal to the BPF infrastructure, that is used to
1749 * apply the last verdict to the next *bytes*. If *bytes* is
1750 * smaller than the current data being processed from a
1751 * **sendmsg**\ () or **sendfile**\ () system call, the first
1752 * *bytes* will be sent and the eBPF program will be re-run with
1753 * the pointer for start of data pointing to byte number *bytes*
1754 * **+ 1**. If *bytes* is larger than the current data being
1755 * processed, then the eBPF verdict will be applied to multiple
1756 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1757 * consumed.
1758 *
1759 * Note that if a socket closes with the internal counter holding
1760 * a non-zero value, this is not a problem because data is not
1761 * being buffered for *bytes* and is sent as it is received.
1762 * Return
1763 * 0
1764 *
1765 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1766 * Description
1767 * For socket policies, prevent the execution of the verdict eBPF
1768 * program for message *msg* until *bytes* (byte number) have been
1769 * accumulated.
1770 *
1771 * This can be used when one needs a specific number of bytes
1772 * before a verdict can be assigned, even if the data spans
1773 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1774 * case would be a user calling **sendmsg**\ () repeatedly with
1775 * 1-byte long message segments. Obviously, this is bad for
1776 * performance, but it is still valid. If the eBPF program needs
1777 * *bytes* bytes to validate a header, this helper can be used to
1778 * prevent the eBPF program to be called again until *bytes* have
1779 * been accumulated.
1780 * Return
1781 * 0
1782 *
1783 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1784 * Description
1785 * For socket policies, pull in non-linear data from user space
1786 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1787 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1788 * respectively.
1789 *
1790 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1791 * *msg* it can only parse data that the (**data**, **data_end**)
1792 * pointers have already consumed. For **sendmsg**\ () hooks this
1793 * is likely the first scatterlist element. But for calls relying
1794 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1795 * be the range (**0**, **0**) because the data is shared with
1796 * user space and by default the objective is to avoid allowing
1797 * user space to modify data while (or after) eBPF verdict is
1798 * being decided. This helper can be used to pull in data and to
1799 * set the start and end pointer to given values. Data will be
1800 * copied if necessary (i.e. if data was not linear and if start
1801 * and end pointers do not point to the same chunk).
1802 *
1803 * A call to this helper is susceptible to change the underlaying
1804 * packet buffer. Therefore, at load time, all checks on pointers
1805 * previously done by the verifier are invalidated and must be
1806 * performed again, if the helper is used in combination with
1807 * direct packet access.
1808 *
1809 * All values for *flags* are reserved for future usage, and must
1810 * be left at zero.
1811 * Return
1812 * 0 on success, or a negative error in case of failure.
1813 *
1814 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1815 * Description
1816 * Bind the socket associated to *ctx* to the address pointed by
1817 * *addr*, of length *addr_len*. This allows for making outgoing
1818 * connection from the desired IP address, which can be useful for
1819 * example when all processes inside a cgroup should use one
1820 * single IP address on a host that has multiple IP configured.
1821 *
1822 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1823 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1824 * **AF_INET6**). Looking for a free port to bind to can be
1825 * expensive, therefore binding to port is not permitted by the
1826 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1827 * must be set to zero.
1828 * Return
1829 * 0 on success, or a negative error in case of failure.
1830 *
1831 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1832 * Description
1833 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1834 * only possible to shrink the packet as of this writing,
1835 * therefore *delta* must be a negative integer.
1836 *
1837 * A call to this helper is susceptible to change the underlaying
1838 * packet buffer. Therefore, at load time, all checks on pointers
1839 * previously done by the verifier are invalidated and must be
1840 * performed again, if the helper is used in combination with
1841 * direct packet access.
1842 * Return
1843 * 0 on success, or a negative error in case of failure.
1844 *
1845 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1846 * Description
1847 * Retrieve the XFRM state (IP transform framework, see also
1848 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1849 *
1850 * The retrieved value is stored in the **struct bpf_xfrm_state**
1851 * pointed by *xfrm_state* and of length *size*.
1852 *
1853 * All values for *flags* are reserved for future usage, and must
1854 * be left at zero.
1855 *
1856 * This helper is available only if the kernel was compiled with
1857 * **CONFIG_XFRM** configuration option.
1858 * Return
1859 * 0 on success, or a negative error in case of failure.
1860 *
1861 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1862 * Description
1863 * Return a user or a kernel stack in bpf program provided buffer.
1864 * To achieve this, the helper needs *ctx*, which is a pointer
1865 * to the context on which the tracing program is executed.
1866 * To store the stacktrace, the bpf program provides *buf* with
1867 * a nonnegative *size*.
1868 *
1869 * The last argument, *flags*, holds the number of stack frames to
1870 * skip (from 0 to 255), masked with
1871 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1872 * the following flags:
1873 *
1874 * **BPF_F_USER_STACK**
1875 * Collect a user space stack instead of a kernel stack.
1876 * **BPF_F_USER_BUILD_ID**
1877 * Collect buildid+offset instead of ips for user stack,
1878 * only valid if **BPF_F_USER_STACK** is also specified.
1879 *
1880 * **bpf_get_stack**\ () can collect up to
1881 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1882 * to sufficient large buffer size. Note that
1883 * this limit can be controlled with the **sysctl** program, and
1884 * that it should be manually increased in order to profile long
1885 * user stacks (such as stacks for Java programs). To do so, use:
1886 *
1887 * ::
1888 *
1889 * # sysctl kernel.perf_event_max_stack=<new value>
1890 * Return
1891 * A non-negative value equal to or less than *size* on success,
1892 * or a negative error in case of failure.
1893 *
1894 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1895 * Description
1896 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1897 * it provides an easy way to load *len* bytes from *offset*
1898 * from the packet associated to *skb*, into the buffer pointed
1899 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1900 * a fifth argument *start_header* exists in order to select a
1901 * base offset to start from. *start_header* can be one of:
1902 *
1903 * **BPF_HDR_START_MAC**
1904 * Base offset to load data from is *skb*'s mac header.
1905 * **BPF_HDR_START_NET**
1906 * Base offset to load data from is *skb*'s network header.
1907 *
1908 * In general, "direct packet access" is the preferred method to
1909 * access packet data, however, this helper is in particular useful
1910 * in socket filters where *skb*\ **->data** does not always point
1911 * to the start of the mac header and where "direct packet access"
1912 * is not available.
1913 * Return
1914 * 0 on success, or a negative error in case of failure.
1915 *
1916 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1917 * Description
1918 * Do FIB lookup in kernel tables using parameters in *params*.
1919 * If lookup is successful and result shows packet is to be
1920 * forwarded, the neighbor tables are searched for the nexthop.
1921 * If successful (ie., FIB lookup shows forwarding and nexthop
1922 * is resolved), the nexthop address is returned in ipv4_dst
1923 * or ipv6_dst based on family, smac is set to mac address of
1924 * egress device, dmac is set to nexthop mac address, rt_metric
1925 * is set to metric from route (IPv4/IPv6 only), and ifindex
1926 * is set to the device index of the nexthop from the FIB lookup.
1927 *
1928 * *plen* argument is the size of the passed in struct.
1929 * *flags* argument can be a combination of one or more of the
1930 * following values:
1931 *
1932 * **BPF_FIB_LOOKUP_DIRECT**
1933 * Do a direct table lookup vs full lookup using FIB
1934 * rules.
1935 * **BPF_FIB_LOOKUP_OUTPUT**
1936 * Perform lookup from an egress perspective (default is
1937 * ingress).
1938 *
1939 * *ctx* is either **struct xdp_md** for XDP programs or
1940 * **struct sk_buff** tc cls_act programs.
1941 * Return
1942 * * < 0 if any input argument is invalid
1943 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1944 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1945 * packet is not forwarded or needs assist from full stack
1946 *
1947 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1948 * Description
1949 * Add an entry to, or update a sockhash *map* referencing sockets.
1950 * The *skops* is used as a new value for the entry associated to
1951 * *key*. *flags* is one of:
1952 *
1953 * **BPF_NOEXIST**
1954 * The entry for *key* must not exist in the map.
1955 * **BPF_EXIST**
1956 * The entry for *key* must already exist in the map.
1957 * **BPF_ANY**
1958 * No condition on the existence of the entry for *key*.
1959 *
1960 * If the *map* has eBPF programs (parser and verdict), those will
1961 * be inherited by the socket being added. If the socket is
1962 * already attached to eBPF programs, this results in an error.
1963 * Return
1964 * 0 on success, or a negative error in case of failure.
1965 *
1966 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1967 * Description
1968 * This helper is used in programs implementing policies at the
1969 * socket level. If the message *msg* is allowed to pass (i.e. if
1970 * the verdict eBPF program returns **SK_PASS**), redirect it to
1971 * the socket referenced by *map* (of type
1972 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1973 * egress interfaces can be used for redirection. The
1974 * **BPF_F_INGRESS** value in *flags* is used to make the
1975 * distinction (ingress path is selected if the flag is present,
1976 * egress path otherwise). This is the only flag supported for now.
1977 * Return
1978 * **SK_PASS** on success, or **SK_DROP** on error.
1979 *
1980 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1981 * Description
1982 * This helper is used in programs implementing policies at the
1983 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1984 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1985 * to the socket referenced by *map* (of type
1986 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1987 * egress interfaces can be used for redirection. The
1988 * **BPF_F_INGRESS** value in *flags* is used to make the
1989 * distinction (ingress path is selected if the flag is present,
1990 * egress otherwise). This is the only flag supported for now.
1991 * Return
1992 * **SK_PASS** on success, or **SK_DROP** on error.
1993 *
1994 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
1995 * Description
1996 * Encapsulate the packet associated to *skb* within a Layer 3
1997 * protocol header. This header is provided in the buffer at
1998 * address *hdr*, with *len* its size in bytes. *type* indicates
1999 * the protocol of the header and can be one of:
2000 *
2001 * **BPF_LWT_ENCAP_SEG6**
2002 * IPv6 encapsulation with Segment Routing Header
2003 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2004 * the IPv6 header is computed by the kernel.
2005 * **BPF_LWT_ENCAP_SEG6_INLINE**
2006 * Only works if *skb* contains an IPv6 packet. Insert a
2007 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2008 * the IPv6 header.
2009 * **BPF_LWT_ENCAP_IP**
2010 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2011 * must be IPv4 or IPv6, followed by zero or more
2012 * additional headers, up to LWT_BPF_MAX_HEADROOM total
2013 * bytes in all prepended headers. Please note that
2014 * if skb_is_gso(skb) is true, no more than two headers
2015 * can be prepended, and the inner header, if present,
2016 * should be either GRE or UDP/GUE.
2017 *
2018 * BPF_LWT_ENCAP_SEG6*** types can be called by bpf programs of
2019 * type BPF_PROG_TYPE_LWT_IN; BPF_LWT_ENCAP_IP type can be called
2020 * by bpf programs of types BPF_PROG_TYPE_LWT_IN and
2021 * BPF_PROG_TYPE_LWT_XMIT.
2022 *
2023 * A call to this helper is susceptible to change the underlaying
2024 * packet buffer. Therefore, at load time, all checks on pointers
2025 * previously done by the verifier are invalidated and must be
2026 * performed again, if the helper is used in combination with
2027 * direct packet access.
2028 * Return
2029 * 0 on success, or a negative error in case of failure.
2030 *
2031 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2032 * Description
2033 * Store *len* bytes from address *from* into the packet
2034 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2035 * inside the outermost IPv6 Segment Routing Header can be
2036 * modified through this helper.
2037 *
2038 * A call to this helper is susceptible to change the underlaying
2039 * packet buffer. Therefore, at load time, all checks on pointers
2040 * previously done by the verifier are invalidated and must be
2041 * performed again, if the helper is used in combination with
2042 * direct packet access.
2043 * Return
2044 * 0 on success, or a negative error in case of failure.
2045 *
2046 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2047 * Description
2048 * Adjust the size allocated to TLVs in the outermost IPv6
2049 * Segment Routing Header contained in the packet associated to
2050 * *skb*, at position *offset* by *delta* bytes. Only offsets
2051 * after the segments are accepted. *delta* can be as well
2052 * positive (growing) as negative (shrinking).
2053 *
2054 * A call to this helper is susceptible to change the underlaying
2055 * packet buffer. Therefore, at load time, all checks on pointers
2056 * previously done by the verifier are invalidated and must be
2057 * performed again, if the helper is used in combination with
2058 * direct packet access.
2059 * Return
2060 * 0 on success, or a negative error in case of failure.
2061 *
2062 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2063 * Description
2064 * Apply an IPv6 Segment Routing action of type *action* to the
2065 * packet associated to *skb*. Each action takes a parameter
2066 * contained at address *param*, and of length *param_len* bytes.
2067 * *action* can be one of:
2068 *
2069 * **SEG6_LOCAL_ACTION_END_X**
2070 * End.X action: Endpoint with Layer-3 cross-connect.
2071 * Type of *param*: **struct in6_addr**.
2072 * **SEG6_LOCAL_ACTION_END_T**
2073 * End.T action: Endpoint with specific IPv6 table lookup.
2074 * Type of *param*: **int**.
2075 * **SEG6_LOCAL_ACTION_END_B6**
2076 * End.B6 action: Endpoint bound to an SRv6 policy.
2077 * Type of param: **struct ipv6_sr_hdr**.
2078 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2079 * End.B6.Encap action: Endpoint bound to an SRv6
2080 * encapsulation policy.
2081 * Type of param: **struct ipv6_sr_hdr**.
2082 *
2083 * A call to this helper is susceptible to change the underlaying
2084 * packet buffer. Therefore, at load time, all checks on pointers
2085 * previously done by the verifier are invalidated and must be
2086 * performed again, if the helper is used in combination with
2087 * direct packet access.
2088 * Return
2089 * 0 on success, or a negative error in case of failure.
2090 *
2091 * int bpf_rc_repeat(void *ctx)
2092 * Description
2093 * This helper is used in programs implementing IR decoding, to
2094 * report a successfully decoded repeat key message. This delays
2095 * the generation of a key up event for previously generated
2096 * key down event.
2097 *
2098 * Some IR protocols like NEC have a special IR message for
2099 * repeating last button, for when a button is held down.
2100 *
2101 * The *ctx* should point to the lirc sample as passed into
2102 * the program.
2103 *
2104 * This helper is only available is the kernel was compiled with
2105 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2106 * "**y**".
2107 * Return
2108 * 0
2109 *
2110 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2111 * Description
2112 * This helper is used in programs implementing IR decoding, to
2113 * report a successfully decoded key press with *scancode*,
2114 * *toggle* value in the given *protocol*. The scancode will be
2115 * translated to a keycode using the rc keymap, and reported as
2116 * an input key down event. After a period a key up event is
2117 * generated. This period can be extended by calling either
2118 * **bpf_rc_keydown**\ () again with the same values, or calling
2119 * **bpf_rc_repeat**\ ().
2120 *
2121 * Some protocols include a toggle bit, in case the button was
2122 * released and pressed again between consecutive scancodes.
2123 *
2124 * The *ctx* should point to the lirc sample as passed into
2125 * the program.
2126 *
2127 * The *protocol* is the decoded protocol number (see
2128 * **enum rc_proto** for some predefined values).
2129 *
2130 * This helper is only available is the kernel was compiled with
2131 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2132 * "**y**".
2133 * Return
2134 * 0
2135 *
2136 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2137 * Description
2138 * Return the cgroup v2 id of the socket associated with the *skb*.
2139 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2140 * helper for cgroup v1 by providing a tag resp. identifier that
2141 * can be matched on or used for map lookups e.g. to implement
2142 * policy. The cgroup v2 id of a given path in the hierarchy is
2143 * exposed in user space through the f_handle API in order to get
2144 * to the same 64-bit id.
2145 *
2146 * This helper can be used on TC egress path, but not on ingress,
2147 * and is available only if the kernel was compiled with the
2148 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2149 * Return
2150 * The id is returned or 0 in case the id could not be retrieved.
2151 *
2152 * u64 bpf_get_current_cgroup_id(void)
2153 * Return
2154 * A 64-bit integer containing the current cgroup id based
2155 * on the cgroup within which the current task is running.
2156 *
2157 * void *bpf_get_local_storage(void *map, u64 flags)
2158 * Description
2159 * Get the pointer to the local storage area.
2160 * The type and the size of the local storage is defined
2161 * by the *map* argument.
2162 * The *flags* meaning is specific for each map type,
2163 * and has to be 0 for cgroup local storage.
2164 *
2165 * Depending on the BPF program type, a local storage area
2166 * can be shared between multiple instances of the BPF program,
2167 * running simultaneously.
2168 *
2169 * A user should care about the synchronization by himself.
2170 * For example, by using the **BPF_STX_XADD** instruction to alter
2171 * the shared data.
2172 * Return
2173 * A pointer to the local storage area.
2174 *
2175 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2176 * Description
2177 * Select a **SO_REUSEPORT** socket from a
2178 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2179 * It checks the selected socket is matching the incoming
2180 * request in the socket buffer.
2181 * Return
2182 * 0 on success, or a negative error in case of failure.
2183 *
2184 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2185 * Description
2186 * Return id of cgroup v2 that is ancestor of cgroup associated
2187 * with the *skb* at the *ancestor_level*. The root cgroup is at
2188 * *ancestor_level* zero and each step down the hierarchy
2189 * increments the level. If *ancestor_level* == level of cgroup
2190 * associated with *skb*, then return value will be same as that
2191 * of **bpf_skb_cgroup_id**\ ().
2192 *
2193 * The helper is useful to implement policies based on cgroups
2194 * that are upper in hierarchy than immediate cgroup associated
2195 * with *skb*.
2196 *
2197 * The format of returned id and helper limitations are same as in
2198 * **bpf_skb_cgroup_id**\ ().
2199 * Return
2200 * The id is returned or 0 in case the id could not be retrieved.
2201 *
2202 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2203 * Description
2204 * Look for TCP socket matching *tuple*, optionally in a child
2205 * network namespace *netns*. The return value must be checked,
2206 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2207 *
2208 * The *ctx* should point to the context of the program, such as
2209 * the skb or socket (depending on the hook in use). This is used
2210 * to determine the base network namespace for the lookup.
2211 *
2212 * *tuple_size* must be one of:
2213 *
2214 * **sizeof**\ (*tuple*\ **->ipv4**)
2215 * Look for an IPv4 socket.
2216 * **sizeof**\ (*tuple*\ **->ipv6**)
2217 * Look for an IPv6 socket.
2218 *
2219 * If the *netns* is a negative signed 32-bit integer, then the
2220 * socket lookup table in the netns associated with the *ctx* will
2221 * will be used. For the TC hooks, this is the netns of the device
2222 * in the skb. For socket hooks, this is the netns of the socket.
2223 * If *netns* is any other signed 32-bit value greater than or
2224 * equal to zero then it specifies the ID of the netns relative to
2225 * the netns associated with the *ctx*. *netns* values beyond the
2226 * range of 32-bit integers are reserved for future use.
2227 *
2228 * All values for *flags* are reserved for future usage, and must
2229 * be left at zero.
2230 *
2231 * This helper is available only if the kernel was compiled with
2232 * **CONFIG_NET** configuration option.
2233 * Return
2234 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2235 * For sockets with reuseport option, the **struct bpf_sock**
2236 * result is from **reuse->socks**\ [] using the hash of the tuple.
2237 *
2238 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2239 * Description
2240 * Look for UDP socket matching *tuple*, optionally in a child
2241 * network namespace *netns*. The return value must be checked,
2242 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2243 *
2244 * The *ctx* should point to the context of the program, such as
2245 * the skb or socket (depending on the hook in use). This is used
2246 * to determine the base network namespace for the lookup.
2247 *
2248 * *tuple_size* must be one of:
2249 *
2250 * **sizeof**\ (*tuple*\ **->ipv4**)
2251 * Look for an IPv4 socket.
2252 * **sizeof**\ (*tuple*\ **->ipv6**)
2253 * Look for an IPv6 socket.
2254 *
2255 * If the *netns* is a negative signed 32-bit integer, then the
2256 * socket lookup table in the netns associated with the *ctx* will
2257 * will be used. For the TC hooks, this is the netns of the device
2258 * in the skb. For socket hooks, this is the netns of the socket.
2259 * If *netns* is any other signed 32-bit value greater than or
2260 * equal to zero then it specifies the ID of the netns relative to
2261 * the netns associated with the *ctx*. *netns* values beyond the
2262 * range of 32-bit integers are reserved for future use.
2263 *
2264 * All values for *flags* are reserved for future usage, and must
2265 * be left at zero.
2266 *
2267 * This helper is available only if the kernel was compiled with
2268 * **CONFIG_NET** configuration option.
2269 * Return
2270 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2271 * For sockets with reuseport option, the **struct bpf_sock**
2272 * result is from **reuse->socks**\ [] using the hash of the tuple.
2273 *
2274 * int bpf_sk_release(struct bpf_sock *sock)
2275 * Description
2276 * Release the reference held by *sock*. *sock* must be a
2277 * non-**NULL** pointer that was returned from
2278 * **bpf_sk_lookup_xxx**\ ().
2279 * Return
2280 * 0 on success, or a negative error in case of failure.
2281 *
2282 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2283 * Description
2284 * Push an element *value* in *map*. *flags* is one of:
2285 *
2286 * **BPF_EXIST**
2287 * If the queue/stack is full, the oldest element is
2288 * removed to make room for this.
2289 * Return
2290 * 0 on success, or a negative error in case of failure.
2291 *
2292 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2293 * Description
2294 * Pop an element from *map*.
2295 * Return
2296 * 0 on success, or a negative error in case of failure.
2297 *
2298 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2299 * Description
2300 * Get an element from *map* without removing it.
2301 * Return
2302 * 0 on success, or a negative error in case of failure.
2303 *
2304 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2305 * Description
2306 * For socket policies, insert *len* bytes into *msg* at offset
2307 * *start*.
2308 *
2309 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2310 * *msg* it may want to insert metadata or options into the *msg*.
2311 * This can later be read and used by any of the lower layer BPF
2312 * hooks.
2313 *
2314 * This helper may fail if under memory pressure (a malloc
2315 * fails) in these cases BPF programs will get an appropriate
2316 * error and BPF programs will need to handle them.
2317 * Return
2318 * 0 on success, or a negative error in case of failure.
2319 *
2320 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2321 * Description
2322 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2323 * This may result in **ENOMEM** errors under certain situations if
2324 * an allocation and copy are required due to a full ring buffer.
2325 * However, the helper will try to avoid doing the allocation
2326 * if possible. Other errors can occur if input parameters are
2327 * invalid either due to *start* byte not being valid part of *msg*
2328 * payload and/or *pop* value being to large.
2329 * Return
2330 * 0 on success, or a negative error in case of failure.
2331 *
2332 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2333 * Description
2334 * This helper is used in programs implementing IR decoding, to
2335 * report a successfully decoded pointer movement.
2336 *
2337 * The *ctx* should point to the lirc sample as passed into
2338 * the program.
2339 *
2340 * This helper is only available is the kernel was compiled with
2341 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2342 * "**y**".
2343 * Return
2344 * 0
2345 *
2346 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2347 * Description
2348 * Acquire a spinlock represented by the pointer *lock*, which is
2349 * stored as part of a value of a map. Taking the lock allows to
2350 * safely update the rest of the fields in that value. The
2351 * spinlock can (and must) later be released with a call to
2352 * **bpf_spin_unlock**\ (\ *lock*\ ).
2353 *
2354 * Spinlocks in BPF programs come with a number of restrictions
2355 * and constraints:
2356 *
2357 * * **bpf_spin_lock** objects are only allowed inside maps of
2358 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2359 * list could be extended in the future).
2360 * * BTF description of the map is mandatory.
2361 * * The BPF program can take ONE lock at a time, since taking two
2362 * or more could cause dead locks.
2363 * * Only one **struct bpf_spin_lock** is allowed per map element.
2364 * * When the lock is taken, calls (either BPF to BPF or helpers)
2365 * are not allowed.
2366 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2367 * allowed inside a spinlock-ed region.
2368 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2369 * the lock, on all execution paths, before it returns.
2370 * * The BPF program can access **struct bpf_spin_lock** only via
2371 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2372 * helpers. Loading or storing data into the **struct
2373 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2374 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2375 * of the map value must be a struct and have **struct
2376 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2377 * Nested lock inside another struct is not allowed.
2378 * * The **struct bpf_spin_lock** *lock* field in a map value must
2379 * be aligned on a multiple of 4 bytes in that value.
2380 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2381 * the **bpf_spin_lock** field to user space.
2382 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2383 * a BPF program, do not update the **bpf_spin_lock** field.
2384 * * **bpf_spin_lock** cannot be on the stack or inside a
2385 * networking packet (it can only be inside of a map values).
2386 * * **bpf_spin_lock** is available to root only.
2387 * * Tracing programs and socket filter programs cannot use
2388 * **bpf_spin_lock**\ () due to insufficient preemption checks
2389 * (but this may change in the future).
2390 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2391 * Return
2392 * 0
2393 *
2394 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2395 * Description
2396 * Release the *lock* previously locked by a call to
2397 * **bpf_spin_lock**\ (\ *lock*\ ).
2398 * Return
2399 * 0
2400 *
2401 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2402 * Description
2403 * This helper gets a **struct bpf_sock** pointer such
2404 * that all the fields in this **bpf_sock** can be accessed.
2405 * Return
2406 * A **struct bpf_sock** pointer on success, or **NULL** in
2407 * case of failure.
2408 *
2409 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2410 * Description
2411 * This helper gets a **struct bpf_tcp_sock** pointer from a
2412 * **struct bpf_sock** pointer.
2413 * Return
2414 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2415 * case of failure.
2416 *
2417 * int bpf_skb_ecn_set_ce(struct sk_buf *skb)
2418 * Description
2419 * Set ECN (Explicit Congestion Notification) field of IP header
2420 * to **CE** (Congestion Encountered) if current value is **ECT**
2421 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2422 * and IPv4.
2423 * Return
2424 * 1 if the **CE** flag is set (either by the current helper call
2425 * or because it was already present), 0 if it is not set.
2426 *
2427 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2428 * Description
2429 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2430 * **bpf_sk_release**\ () is unnecessary and not allowed.
2431 * Return
2432 * A **struct bpf_sock** pointer on success, or **NULL** in
2433 * case of failure.
2434 */
2435 #define __BPF_FUNC_MAPPER(FN) \
2436 FN(unspec), \
2437 FN(map_lookup_elem), \
2438 FN(map_update_elem), \
2439 FN(map_delete_elem), \
2440 FN(probe_read), \
2441 FN(ktime_get_ns), \
2442 FN(trace_printk), \
2443 FN(get_prandom_u32), \
2444 FN(get_smp_processor_id), \
2445 FN(skb_store_bytes), \
2446 FN(l3_csum_replace), \
2447 FN(l4_csum_replace), \
2448 FN(tail_call), \
2449 FN(clone_redirect), \
2450 FN(get_current_pid_tgid), \
2451 FN(get_current_uid_gid), \
2452 FN(get_current_comm), \
2453 FN(get_cgroup_classid), \
2454 FN(skb_vlan_push), \
2455 FN(skb_vlan_pop), \
2456 FN(skb_get_tunnel_key), \
2457 FN(skb_set_tunnel_key), \
2458 FN(perf_event_read), \
2459 FN(redirect), \
2460 FN(get_route_realm), \
2461 FN(perf_event_output), \
2462 FN(skb_load_bytes), \
2463 FN(get_stackid), \
2464 FN(csum_diff), \
2465 FN(skb_get_tunnel_opt), \
2466 FN(skb_set_tunnel_opt), \
2467 FN(skb_change_proto), \
2468 FN(skb_change_type), \
2469 FN(skb_under_cgroup), \
2470 FN(get_hash_recalc), \
2471 FN(get_current_task), \
2472 FN(probe_write_user), \
2473 FN(current_task_under_cgroup), \
2474 FN(skb_change_tail), \
2475 FN(skb_pull_data), \
2476 FN(csum_update), \
2477 FN(set_hash_invalid), \
2478 FN(get_numa_node_id), \
2479 FN(skb_change_head), \
2480 FN(xdp_adjust_head), \
2481 FN(probe_read_str), \
2482 FN(get_socket_cookie), \
2483 FN(get_socket_uid), \
2484 FN(set_hash), \
2485 FN(setsockopt), \
2486 FN(skb_adjust_room), \
2487 FN(redirect_map), \
2488 FN(sk_redirect_map), \
2489 FN(sock_map_update), \
2490 FN(xdp_adjust_meta), \
2491 FN(perf_event_read_value), \
2492 FN(perf_prog_read_value), \
2493 FN(getsockopt), \
2494 FN(override_return), \
2495 FN(sock_ops_cb_flags_set), \
2496 FN(msg_redirect_map), \
2497 FN(msg_apply_bytes), \
2498 FN(msg_cork_bytes), \
2499 FN(msg_pull_data), \
2500 FN(bind), \
2501 FN(xdp_adjust_tail), \
2502 FN(skb_get_xfrm_state), \
2503 FN(get_stack), \
2504 FN(skb_load_bytes_relative), \
2505 FN(fib_lookup), \
2506 FN(sock_hash_update), \
2507 FN(msg_redirect_hash), \
2508 FN(sk_redirect_hash), \
2509 FN(lwt_push_encap), \
2510 FN(lwt_seg6_store_bytes), \
2511 FN(lwt_seg6_adjust_srh), \
2512 FN(lwt_seg6_action), \
2513 FN(rc_repeat), \
2514 FN(rc_keydown), \
2515 FN(skb_cgroup_id), \
2516 FN(get_current_cgroup_id), \
2517 FN(get_local_storage), \
2518 FN(sk_select_reuseport), \
2519 FN(skb_ancestor_cgroup_id), \
2520 FN(sk_lookup_tcp), \
2521 FN(sk_lookup_udp), \
2522 FN(sk_release), \
2523 FN(map_push_elem), \
2524 FN(map_pop_elem), \
2525 FN(map_peek_elem), \
2526 FN(msg_push_data), \
2527 FN(msg_pop_data), \
2528 FN(rc_pointer_rel), \
2529 FN(spin_lock), \
2530 FN(spin_unlock), \
2531 FN(sk_fullsock), \
2532 FN(tcp_sock), \
2533 FN(skb_ecn_set_ce), \
2534 FN(get_listener_sock),
2535
2536 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2537 * function eBPF program intends to call
2538 */
2539 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2540 enum bpf_func_id {
2541 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2542 __BPF_FUNC_MAX_ID,
2543 };
2544 #undef __BPF_ENUM_FN
2545
2546 /* All flags used by eBPF helper functions, placed here. */
2547
2548 /* BPF_FUNC_skb_store_bytes flags. */
2549 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2550 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2551
2552 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2553 * First 4 bits are for passing the header field size.
2554 */
2555 #define BPF_F_HDR_FIELD_MASK 0xfULL
2556
2557 /* BPF_FUNC_l4_csum_replace flags. */
2558 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2559 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2560 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2561
2562 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2563 #define BPF_F_INGRESS (1ULL << 0)
2564
2565 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2566 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2567
2568 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2569 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2570 #define BPF_F_USER_STACK (1ULL << 8)
2571 /* flags used by BPF_FUNC_get_stackid only. */
2572 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2573 #define BPF_F_REUSE_STACKID (1ULL << 10)
2574 /* flags used by BPF_FUNC_get_stack only. */
2575 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2576
2577 /* BPF_FUNC_skb_set_tunnel_key flags. */
2578 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2579 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2580 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2581
2582 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2583 * BPF_FUNC_perf_event_read_value flags.
2584 */
2585 #define BPF_F_INDEX_MASK 0xffffffffULL
2586 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2587 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2588 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2589
2590 /* Current network namespace */
2591 #define BPF_F_CURRENT_NETNS (-1L)
2592
2593 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2594 enum bpf_adj_room_mode {
2595 BPF_ADJ_ROOM_NET,
2596 };
2597
2598 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2599 enum bpf_hdr_start_off {
2600 BPF_HDR_START_MAC,
2601 BPF_HDR_START_NET,
2602 };
2603
2604 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2605 enum bpf_lwt_encap_mode {
2606 BPF_LWT_ENCAP_SEG6,
2607 BPF_LWT_ENCAP_SEG6_INLINE,
2608 BPF_LWT_ENCAP_IP,
2609 };
2610
2611 #define __bpf_md_ptr(type, name) \
2612 union { \
2613 type name; \
2614 __u64 :64; \
2615 } __attribute__((aligned(8)))
2616
2617 /* user accessible mirror of in-kernel sk_buff.
2618 * new fields can only be added to the end of this structure
2619 */
2620 struct __sk_buff {
2621 __u32 len;
2622 __u32 pkt_type;
2623 __u32 mark;
2624 __u32 queue_mapping;
2625 __u32 protocol;
2626 __u32 vlan_present;
2627 __u32 vlan_tci;
2628 __u32 vlan_proto;
2629 __u32 priority;
2630 __u32 ingress_ifindex;
2631 __u32 ifindex;
2632 __u32 tc_index;
2633 __u32 cb[5];
2634 __u32 hash;
2635 __u32 tc_classid;
2636 __u32 data;
2637 __u32 data_end;
2638 __u32 napi_id;
2639
2640 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2641 __u32 family;
2642 __u32 remote_ip4; /* Stored in network byte order */
2643 __u32 local_ip4; /* Stored in network byte order */
2644 __u32 remote_ip6[4]; /* Stored in network byte order */
2645 __u32 local_ip6[4]; /* Stored in network byte order */
2646 __u32 remote_port; /* Stored in network byte order */
2647 __u32 local_port; /* stored in host byte order */
2648 /* ... here. */
2649
2650 __u32 data_meta;
2651 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
2652 __u64 tstamp;
2653 __u32 wire_len;
2654 __u32 gso_segs;
2655 __bpf_md_ptr(struct bpf_sock *, sk);
2656 };
2657
2658 struct bpf_tunnel_key {
2659 __u32 tunnel_id;
2660 union {
2661 __u32 remote_ipv4;
2662 __u32 remote_ipv6[4];
2663 };
2664 __u8 tunnel_tos;
2665 __u8 tunnel_ttl;
2666 __u16 tunnel_ext; /* Padding, future use. */
2667 __u32 tunnel_label;
2668 };
2669
2670 /* user accessible mirror of in-kernel xfrm_state.
2671 * new fields can only be added to the end of this structure
2672 */
2673 struct bpf_xfrm_state {
2674 __u32 reqid;
2675 __u32 spi; /* Stored in network byte order */
2676 __u16 family;
2677 __u16 ext; /* Padding, future use. */
2678 union {
2679 __u32 remote_ipv4; /* Stored in network byte order */
2680 __u32 remote_ipv6[4]; /* Stored in network byte order */
2681 };
2682 };
2683
2684 /* Generic BPF return codes which all BPF program types may support.
2685 * The values are binary compatible with their TC_ACT_* counter-part to
2686 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2687 * programs.
2688 *
2689 * XDP is handled seprately, see XDP_*.
2690 */
2691 enum bpf_ret_code {
2692 BPF_OK = 0,
2693 /* 1 reserved */
2694 BPF_DROP = 2,
2695 /* 3-6 reserved */
2696 BPF_REDIRECT = 7,
2697 /* >127 are reserved for prog type specific return codes.
2698 *
2699 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
2700 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
2701 * changed and should be routed based on its new L3 header.
2702 * (This is an L3 redirect, as opposed to L2 redirect
2703 * represented by BPF_REDIRECT above).
2704 */
2705 BPF_LWT_REROUTE = 128,
2706 };
2707
2708 struct bpf_sock {
2709 __u32 bound_dev_if;
2710 __u32 family;
2711 __u32 type;
2712 __u32 protocol;
2713 __u32 mark;
2714 __u32 priority;
2715 /* IP address also allows 1 and 2 bytes access */
2716 __u32 src_ip4;
2717 __u32 src_ip6[4];
2718 __u32 src_port; /* host byte order */
2719 __u32 dst_port; /* network byte order */
2720 __u32 dst_ip4;
2721 __u32 dst_ip6[4];
2722 __u32 state;
2723 };
2724
2725 struct bpf_tcp_sock {
2726 __u32 snd_cwnd; /* Sending congestion window */
2727 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
2728 __u32 rtt_min;
2729 __u32 snd_ssthresh; /* Slow start size threshold */
2730 __u32 rcv_nxt; /* What we want to receive next */
2731 __u32 snd_nxt; /* Next sequence we send */
2732 __u32 snd_una; /* First byte we want an ack for */
2733 __u32 mss_cache; /* Cached effective mss, not including SACKS */
2734 __u32 ecn_flags; /* ECN status bits. */
2735 __u32 rate_delivered; /* saved rate sample: packets delivered */
2736 __u32 rate_interval_us; /* saved rate sample: time elapsed */
2737 __u32 packets_out; /* Packets which are "in flight" */
2738 __u32 retrans_out; /* Retransmitted packets out */
2739 __u32 total_retrans; /* Total retransmits for entire connection */
2740 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
2741 * total number of segments in.
2742 */
2743 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
2744 * total number of data segments in.
2745 */
2746 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
2747 * The total number of segments sent.
2748 */
2749 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
2750 * total number of data segments sent.
2751 */
2752 __u32 lost_out; /* Lost packets */
2753 __u32 sacked_out; /* SACK'd packets */
2754 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
2755 * sum(delta(rcv_nxt)), or how many bytes
2756 * were acked.
2757 */
2758 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
2759 * sum(delta(snd_una)), or how many bytes
2760 * were acked.
2761 */
2762 };
2763
2764 struct bpf_sock_tuple {
2765 union {
2766 struct {
2767 __be32 saddr;
2768 __be32 daddr;
2769 __be16 sport;
2770 __be16 dport;
2771 } ipv4;
2772 struct {
2773 __be32 saddr[4];
2774 __be32 daddr[4];
2775 __be16 sport;
2776 __be16 dport;
2777 } ipv6;
2778 };
2779 };
2780
2781 #define XDP_PACKET_HEADROOM 256
2782
2783 /* User return codes for XDP prog type.
2784 * A valid XDP program must return one of these defined values. All other
2785 * return codes are reserved for future use. Unknown return codes will
2786 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2787 */
2788 enum xdp_action {
2789 XDP_ABORTED = 0,
2790 XDP_DROP,
2791 XDP_PASS,
2792 XDP_TX,
2793 XDP_REDIRECT,
2794 };
2795
2796 /* user accessible metadata for XDP packet hook
2797 * new fields must be added to the end of this structure
2798 */
2799 struct xdp_md {
2800 __u32 data;
2801 __u32 data_end;
2802 __u32 data_meta;
2803 /* Below access go through struct xdp_rxq_info */
2804 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2805 __u32 rx_queue_index; /* rxq->queue_index */
2806 };
2807
2808 enum sk_action {
2809 SK_DROP = 0,
2810 SK_PASS,
2811 };
2812
2813 /* user accessible metadata for SK_MSG packet hook, new fields must
2814 * be added to the end of this structure
2815 */
2816 struct sk_msg_md {
2817 __bpf_md_ptr(void *, data);
2818 __bpf_md_ptr(void *, data_end);
2819
2820 __u32 family;
2821 __u32 remote_ip4; /* Stored in network byte order */
2822 __u32 local_ip4; /* Stored in network byte order */
2823 __u32 remote_ip6[4]; /* Stored in network byte order */
2824 __u32 local_ip6[4]; /* Stored in network byte order */
2825 __u32 remote_port; /* Stored in network byte order */
2826 __u32 local_port; /* stored in host byte order */
2827 __u32 size; /* Total size of sk_msg */
2828 };
2829
2830 struct sk_reuseport_md {
2831 /*
2832 * Start of directly accessible data. It begins from
2833 * the tcp/udp header.
2834 */
2835 __bpf_md_ptr(void *, data);
2836 /* End of directly accessible data */
2837 __bpf_md_ptr(void *, data_end);
2838 /*
2839 * Total length of packet (starting from the tcp/udp header).
2840 * Note that the directly accessible bytes (data_end - data)
2841 * could be less than this "len". Those bytes could be
2842 * indirectly read by a helper "bpf_skb_load_bytes()".
2843 */
2844 __u32 len;
2845 /*
2846 * Eth protocol in the mac header (network byte order). e.g.
2847 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2848 */
2849 __u32 eth_protocol;
2850 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2851 __u32 bind_inany; /* Is sock bound to an INANY address? */
2852 __u32 hash; /* A hash of the packet 4 tuples */
2853 };
2854
2855 #define BPF_TAG_SIZE 8
2856
2857 struct bpf_prog_info {
2858 __u32 type;
2859 __u32 id;
2860 __u8 tag[BPF_TAG_SIZE];
2861 __u32 jited_prog_len;
2862 __u32 xlated_prog_len;
2863 __aligned_u64 jited_prog_insns;
2864 __aligned_u64 xlated_prog_insns;
2865 __u64 load_time; /* ns since boottime */
2866 __u32 created_by_uid;
2867 __u32 nr_map_ids;
2868 __aligned_u64 map_ids;
2869 char name[BPF_OBJ_NAME_LEN];
2870 __u32 ifindex;
2871 __u32 gpl_compatible:1;
2872 __u64 netns_dev;
2873 __u64 netns_ino;
2874 __u32 nr_jited_ksyms;
2875 __u32 nr_jited_func_lens;
2876 __aligned_u64 jited_ksyms;
2877 __aligned_u64 jited_func_lens;
2878 __u32 btf_id;
2879 __u32 func_info_rec_size;
2880 __aligned_u64 func_info;
2881 __u32 nr_func_info;
2882 __u32 nr_line_info;
2883 __aligned_u64 line_info;
2884 __aligned_u64 jited_line_info;
2885 __u32 nr_jited_line_info;
2886 __u32 line_info_rec_size;
2887 __u32 jited_line_info_rec_size;
2888 __u32 nr_prog_tags;
2889 __aligned_u64 prog_tags;
2890 __u64 run_time_ns;
2891 __u64 run_cnt;
2892 } __attribute__((aligned(8)));
2893
2894 struct bpf_map_info {
2895 __u32 type;
2896 __u32 id;
2897 __u32 key_size;
2898 __u32 value_size;
2899 __u32 max_entries;
2900 __u32 map_flags;
2901 char name[BPF_OBJ_NAME_LEN];
2902 __u32 ifindex;
2903 __u32 :32;
2904 __u64 netns_dev;
2905 __u64 netns_ino;
2906 __u32 btf_id;
2907 __u32 btf_key_type_id;
2908 __u32 btf_value_type_id;
2909 } __attribute__((aligned(8)));
2910
2911 struct bpf_btf_info {
2912 __aligned_u64 btf;
2913 __u32 btf_size;
2914 __u32 id;
2915 } __attribute__((aligned(8)));
2916
2917 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2918 * by user and intended to be used by socket (e.g. to bind to, depends on
2919 * attach attach type).
2920 */
2921 struct bpf_sock_addr {
2922 __u32 user_family; /* Allows 4-byte read, but no write. */
2923 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2924 * Stored in network byte order.
2925 */
2926 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2927 * Stored in network byte order.
2928 */
2929 __u32 user_port; /* Allows 4-byte read and write.
2930 * Stored in network byte order
2931 */
2932 __u32 family; /* Allows 4-byte read, but no write */
2933 __u32 type; /* Allows 4-byte read, but no write */
2934 __u32 protocol; /* Allows 4-byte read, but no write */
2935 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2936 * Stored in network byte order.
2937 */
2938 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2939 * Stored in network byte order.
2940 */
2941 };
2942
2943 /* User bpf_sock_ops struct to access socket values and specify request ops
2944 * and their replies.
2945 * Some of this fields are in network (bigendian) byte order and may need
2946 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2947 * New fields can only be added at the end of this structure
2948 */
2949 struct bpf_sock_ops {
2950 __u32 op;
2951 union {
2952 __u32 args[4]; /* Optionally passed to bpf program */
2953 __u32 reply; /* Returned by bpf program */
2954 __u32 replylong[4]; /* Optionally returned by bpf prog */
2955 };
2956 __u32 family;
2957 __u32 remote_ip4; /* Stored in network byte order */
2958 __u32 local_ip4; /* Stored in network byte order */
2959 __u32 remote_ip6[4]; /* Stored in network byte order */
2960 __u32 local_ip6[4]; /* Stored in network byte order */
2961 __u32 remote_port; /* Stored in network byte order */
2962 __u32 local_port; /* stored in host byte order */
2963 __u32 is_fullsock; /* Some TCP fields are only valid if
2964 * there is a full socket. If not, the
2965 * fields read as zero.
2966 */
2967 __u32 snd_cwnd;
2968 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2969 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2970 __u32 state;
2971 __u32 rtt_min;
2972 __u32 snd_ssthresh;
2973 __u32 rcv_nxt;
2974 __u32 snd_nxt;
2975 __u32 snd_una;
2976 __u32 mss_cache;
2977 __u32 ecn_flags;
2978 __u32 rate_delivered;
2979 __u32 rate_interval_us;
2980 __u32 packets_out;
2981 __u32 retrans_out;
2982 __u32 total_retrans;
2983 __u32 segs_in;
2984 __u32 data_segs_in;
2985 __u32 segs_out;
2986 __u32 data_segs_out;
2987 __u32 lost_out;
2988 __u32 sacked_out;
2989 __u32 sk_txhash;
2990 __u64 bytes_received;
2991 __u64 bytes_acked;
2992 };
2993
2994 /* Definitions for bpf_sock_ops_cb_flags */
2995 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2996 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2997 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2998 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2999 * supported cb flags
3000 */
3001
3002 /* List of known BPF sock_ops operators.
3003 * New entries can only be added at the end
3004 */
3005 enum {
3006 BPF_SOCK_OPS_VOID,
3007 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3008 * -1 if default value should be used
3009 */
3010 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3011 * window (in packets) or -1 if default
3012 * value should be used
3013 */
3014 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3015 * active connection is initialized
3016 */
3017 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3018 * active connection is
3019 * established
3020 */
3021 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3022 * passive connection is
3023 * established
3024 */
3025 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3026 * needs ECN
3027 */
3028 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3029 * based on the path and may be
3030 * dependent on the congestion control
3031 * algorithm. In general it indicates
3032 * a congestion threshold. RTTs above
3033 * this indicate congestion
3034 */
3035 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3036 * Arg1: value of icsk_retransmits
3037 * Arg2: value of icsk_rto
3038 * Arg3: whether RTO has expired
3039 */
3040 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3041 * Arg1: sequence number of 1st byte
3042 * Arg2: # segments
3043 * Arg3: return value of
3044 * tcp_transmit_skb (0 => success)
3045 */
3046 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3047 * Arg1: old_state
3048 * Arg2: new_state
3049 */
3050 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3051 * socket transition to LISTEN state.
3052 */
3053 };
3054
3055 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3056 * changes between the TCP and BPF versions. Ideally this should never happen.
3057 * If it does, we need to add code to convert them before calling
3058 * the BPF sock_ops function.
3059 */
3060 enum {
3061 BPF_TCP_ESTABLISHED = 1,
3062 BPF_TCP_SYN_SENT,
3063 BPF_TCP_SYN_RECV,
3064 BPF_TCP_FIN_WAIT1,
3065 BPF_TCP_FIN_WAIT2,
3066 BPF_TCP_TIME_WAIT,
3067 BPF_TCP_CLOSE,
3068 BPF_TCP_CLOSE_WAIT,
3069 BPF_TCP_LAST_ACK,
3070 BPF_TCP_LISTEN,
3071 BPF_TCP_CLOSING, /* Now a valid state */
3072 BPF_TCP_NEW_SYN_RECV,
3073
3074 BPF_TCP_MAX_STATES /* Leave at the end! */
3075 };
3076
3077 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3078 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3079
3080 struct bpf_perf_event_value {
3081 __u64 counter;
3082 __u64 enabled;
3083 __u64 running;
3084 };
3085
3086 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3087 #define BPF_DEVCG_ACC_READ (1ULL << 1)
3088 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3089
3090 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3091 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3092
3093 struct bpf_cgroup_dev_ctx {
3094 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3095 __u32 access_type;
3096 __u32 major;
3097 __u32 minor;
3098 };
3099
3100 struct bpf_raw_tracepoint_args {
3101 __u64 args[0];
3102 };
3103
3104 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
3105 * OUTPUT: Do lookup from egress perspective; default is ingress
3106 */
3107 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
3108 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
3109
3110 enum {
3111 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3112 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3113 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3114 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3115 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3116 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3117 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3118 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3119 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3120 };
3121
3122 struct bpf_fib_lookup {
3123 /* input: network family for lookup (AF_INET, AF_INET6)
3124 * output: network family of egress nexthop
3125 */
3126 __u8 family;
3127
3128 /* set if lookup is to consider L4 data - e.g., FIB rules */
3129 __u8 l4_protocol;
3130 __be16 sport;
3131 __be16 dport;
3132
3133 /* total length of packet from network header - used for MTU check */
3134 __u16 tot_len;
3135
3136 /* input: L3 device index for lookup
3137 * output: device index from FIB lookup
3138 */
3139 __u32 ifindex;
3140
3141 union {
3142 /* inputs to lookup */
3143 __u8 tos; /* AF_INET */
3144 __be32 flowinfo; /* AF_INET6, flow_label + priority */
3145
3146 /* output: metric of fib result (IPv4/IPv6 only) */
3147 __u32 rt_metric;
3148 };
3149
3150 union {
3151 __be32 ipv4_src;
3152 __u32 ipv6_src[4]; /* in6_addr; network order */
3153 };
3154
3155 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3156 * network header. output: bpf_fib_lookup sets to gateway address
3157 * if FIB lookup returns gateway route
3158 */
3159 union {
3160 __be32 ipv4_dst;
3161 __u32 ipv6_dst[4]; /* in6_addr; network order */
3162 };
3163
3164 /* output */
3165 __be16 h_vlan_proto;
3166 __be16 h_vlan_TCI;
3167 __u8 smac[6]; /* ETH_ALEN */
3168 __u8 dmac[6]; /* ETH_ALEN */
3169 };
3170
3171 enum bpf_task_fd_type {
3172 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3173 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3174 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3175 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3176 BPF_FD_TYPE_UPROBE, /* filename + offset */
3177 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3178 };
3179
3180 struct bpf_flow_keys {
3181 __u16 nhoff;
3182 __u16 thoff;
3183 __u16 addr_proto; /* ETH_P_* of valid addrs */
3184 __u8 is_frag;
3185 __u8 is_first_frag;
3186 __u8 is_encap;
3187 __u8 ip_proto;
3188 __be16 n_proto;
3189 __be16 sport;
3190 __be16 dport;
3191 union {
3192 struct {
3193 __be32 ipv4_src;
3194 __be32 ipv4_dst;
3195 };
3196 struct {
3197 __u32 ipv6_src[4]; /* in6_addr; network order */
3198 __u32 ipv6_dst[4]; /* in6_addr; network order */
3199 };
3200 };
3201 };
3202
3203 struct bpf_func_info {
3204 __u32 insn_off;
3205 __u32 type_id;
3206 };
3207
3208 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3209 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3210
3211 struct bpf_line_info {
3212 __u32 insn_off;
3213 __u32 file_name_off;
3214 __u32 line_off;
3215 __u32 line_col;
3216 };
3217
3218 struct bpf_spin_lock {
3219 __u32 val;
3220 };
3221 #endif /* _UAPI__LINUX_BPF_H__ */