]> git.proxmox.com Git - mirror_iproute2.git/blob - include/uapi/linux/bpf.h
Update kernel headers and add xdp_diag.h
[mirror_iproute2.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 __LINUX_BPF_H__
9 #define __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
271 /* flags for BPF_MAP_CREATE command */
272 #define BPF_F_NO_PREALLOC (1U << 0)
273 /* Instead of having one common LRU list in the
274 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
275 * which can scale and perform better.
276 * Note, the LRU nodes (including free nodes) cannot be moved
277 * across different LRU lists.
278 */
279 #define BPF_F_NO_COMMON_LRU (1U << 1)
280 /* Specify numa node during map creation */
281 #define BPF_F_NUMA_NODE (1U << 2)
282
283 #define BPF_OBJ_NAME_LEN 16U
284
285 /* Flags for accessing BPF object */
286 #define BPF_F_RDONLY (1U << 3)
287 #define BPF_F_WRONLY (1U << 4)
288
289 /* Flag for stack_map, store build_id+offset instead of pointer */
290 #define BPF_F_STACK_BUILD_ID (1U << 5)
291
292 /* Zero-initialize hash function seed. This should only be used for testing. */
293 #define BPF_F_ZERO_SEED (1U << 6)
294
295 /* flags for BPF_PROG_QUERY */
296 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
297
298 enum bpf_stack_build_id_status {
299 /* user space need an empty entry to identify end of a trace */
300 BPF_STACK_BUILD_ID_EMPTY = 0,
301 /* with valid build_id and offset */
302 BPF_STACK_BUILD_ID_VALID = 1,
303 /* couldn't get build_id, fallback to ip */
304 BPF_STACK_BUILD_ID_IP = 2,
305 };
306
307 #define BPF_BUILD_ID_SIZE 20
308 struct bpf_stack_build_id {
309 __s32 status;
310 unsigned char build_id[BPF_BUILD_ID_SIZE];
311 union {
312 __u64 offset;
313 __u64 ip;
314 };
315 };
316
317 union bpf_attr {
318 struct { /* anonymous struct used by BPF_MAP_CREATE command */
319 __u32 map_type; /* one of enum bpf_map_type */
320 __u32 key_size; /* size of key in bytes */
321 __u32 value_size; /* size of value in bytes */
322 __u32 max_entries; /* max number of entries in a map */
323 __u32 map_flags; /* BPF_MAP_CREATE related
324 * flags defined above.
325 */
326 __u32 inner_map_fd; /* fd pointing to the inner map */
327 __u32 numa_node; /* numa node (effective only if
328 * BPF_F_NUMA_NODE is set).
329 */
330 char map_name[BPF_OBJ_NAME_LEN];
331 __u32 map_ifindex; /* ifindex of netdev to create on */
332 __u32 btf_fd; /* fd pointing to a BTF type data */
333 __u32 btf_key_type_id; /* BTF type_id of the key */
334 __u32 btf_value_type_id; /* BTF type_id of the value */
335 };
336
337 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
338 __u32 map_fd;
339 __aligned_u64 key;
340 union {
341 __aligned_u64 value;
342 __aligned_u64 next_key;
343 };
344 __u64 flags;
345 };
346
347 struct { /* anonymous struct used by BPF_PROG_LOAD command */
348 __u32 prog_type; /* one of enum bpf_prog_type */
349 __u32 insn_cnt;
350 __aligned_u64 insns;
351 __aligned_u64 license;
352 __u32 log_level; /* verbosity level of verifier */
353 __u32 log_size; /* size of user buffer */
354 __aligned_u64 log_buf; /* user supplied buffer */
355 __u32 kern_version; /* not used */
356 __u32 prog_flags;
357 char prog_name[BPF_OBJ_NAME_LEN];
358 __u32 prog_ifindex; /* ifindex of netdev to prep for */
359 /* For some prog types expected attach type must be known at
360 * load time to verify attach type specific parts of prog
361 * (context accesses, allowed helpers, etc).
362 */
363 __u32 expected_attach_type;
364 __u32 prog_btf_fd; /* fd pointing to BTF type data */
365 __u32 func_info_rec_size; /* userspace bpf_func_info size */
366 __aligned_u64 func_info; /* func info */
367 __u32 func_info_cnt; /* number of bpf_func_info records */
368 __u32 line_info_rec_size; /* userspace bpf_line_info size */
369 __aligned_u64 line_info; /* line info */
370 __u32 line_info_cnt; /* number of bpf_line_info records */
371 };
372
373 struct { /* anonymous struct used by BPF_OBJ_* commands */
374 __aligned_u64 pathname;
375 __u32 bpf_fd;
376 __u32 file_flags;
377 };
378
379 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
380 __u32 target_fd; /* container object to attach to */
381 __u32 attach_bpf_fd; /* eBPF program to attach */
382 __u32 attach_type;
383 __u32 attach_flags;
384 };
385
386 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
387 __u32 prog_fd;
388 __u32 retval;
389 __u32 data_size_in; /* input: len of data_in */
390 __u32 data_size_out; /* input/output: len of data_out
391 * returns ENOSPC if data_out
392 * is too small.
393 */
394 __aligned_u64 data_in;
395 __aligned_u64 data_out;
396 __u32 repeat;
397 __u32 duration;
398 } test;
399
400 struct { /* anonymous struct used by BPF_*_GET_*_ID */
401 union {
402 __u32 start_id;
403 __u32 prog_id;
404 __u32 map_id;
405 __u32 btf_id;
406 };
407 __u32 next_id;
408 __u32 open_flags;
409 };
410
411 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
412 __u32 bpf_fd;
413 __u32 info_len;
414 __aligned_u64 info;
415 } info;
416
417 struct { /* anonymous struct used by BPF_PROG_QUERY command */
418 __u32 target_fd; /* container object to query */
419 __u32 attach_type;
420 __u32 query_flags;
421 __u32 attach_flags;
422 __aligned_u64 prog_ids;
423 __u32 prog_cnt;
424 } query;
425
426 struct {
427 __u64 name;
428 __u32 prog_fd;
429 } raw_tracepoint;
430
431 struct { /* anonymous struct for BPF_BTF_LOAD */
432 __aligned_u64 btf;
433 __aligned_u64 btf_log_buf;
434 __u32 btf_size;
435 __u32 btf_log_size;
436 __u32 btf_log_level;
437 };
438
439 struct {
440 __u32 pid; /* input: pid */
441 __u32 fd; /* input: fd */
442 __u32 flags; /* input: flags */
443 __u32 buf_len; /* input/output: buf len */
444 __aligned_u64 buf; /* input/output:
445 * tp_name for tracepoint
446 * symbol for kprobe
447 * filename for uprobe
448 */
449 __u32 prog_id; /* output: prod_id */
450 __u32 fd_type; /* output: BPF_FD_TYPE_* */
451 __u64 probe_offset; /* output: probe_offset */
452 __u64 probe_addr; /* output: probe_addr */
453 } task_fd_query;
454 } __attribute__((aligned(8)));
455
456 /* The description below is an attempt at providing documentation to eBPF
457 * developers about the multiple available eBPF helper functions. It can be
458 * parsed and used to produce a manual page. The workflow is the following,
459 * and requires the rst2man utility:
460 *
461 * $ ./scripts/bpf_helpers_doc.py \
462 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
463 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
464 * $ man /tmp/bpf-helpers.7
465 *
466 * Note that in order to produce this external documentation, some RST
467 * formatting is used in the descriptions to get "bold" and "italics" in
468 * manual pages. Also note that the few trailing white spaces are
469 * intentional, removing them would break paragraphs for rst2man.
470 *
471 * Start of BPF helper function descriptions:
472 *
473 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
474 * Description
475 * Perform a lookup in *map* for an entry associated to *key*.
476 * Return
477 * Map value associated to *key*, or **NULL** if no entry was
478 * found.
479 *
480 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
481 * Description
482 * Add or update the value of the entry associated to *key* in
483 * *map* with *value*. *flags* is one of:
484 *
485 * **BPF_NOEXIST**
486 * The entry for *key* must not exist in the map.
487 * **BPF_EXIST**
488 * The entry for *key* must already exist in the map.
489 * **BPF_ANY**
490 * No condition on the existence of the entry for *key*.
491 *
492 * Flag value **BPF_NOEXIST** cannot be used for maps of types
493 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
494 * elements always exist), the helper would return an error.
495 * Return
496 * 0 on success, or a negative error in case of failure.
497 *
498 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
499 * Description
500 * Delete entry with *key* from *map*.
501 * Return
502 * 0 on success, or a negative error in case of failure.
503 *
504 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
505 * Description
506 * Push an element *value* in *map*. *flags* is one of:
507 *
508 * **BPF_EXIST**
509 * If the queue/stack is full, the oldest element is removed to
510 * make room for this.
511 * Return
512 * 0 on success, or a negative error in case of failure.
513 *
514 * int bpf_probe_read(void *dst, u32 size, const void *src)
515 * Description
516 * For tracing programs, safely attempt to read *size* bytes from
517 * address *src* and store the data in *dst*.
518 * Return
519 * 0 on success, or a negative error in case of failure.
520 *
521 * u64 bpf_ktime_get_ns(void)
522 * Description
523 * Return the time elapsed since system boot, in nanoseconds.
524 * Return
525 * Current *ktime*.
526 *
527 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
528 * Description
529 * This helper is a "printk()-like" facility for debugging. It
530 * prints a message defined by format *fmt* (of size *fmt_size*)
531 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
532 * available. It can take up to three additional **u64**
533 * arguments (as an eBPF helpers, the total number of arguments is
534 * limited to five).
535 *
536 * Each time the helper is called, it appends a line to the trace.
537 * The format of the trace is customizable, and the exact output
538 * one will get depends on the options set in
539 * *\/sys/kernel/debug/tracing/trace_options* (see also the
540 * *README* file under the same directory). However, it usually
541 * defaults to something like:
542 *
543 * ::
544 *
545 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
546 *
547 * In the above:
548 *
549 * * ``telnet`` is the name of the current task.
550 * * ``470`` is the PID of the current task.
551 * * ``001`` is the CPU number on which the task is
552 * running.
553 * * In ``.N..``, each character refers to a set of
554 * options (whether irqs are enabled, scheduling
555 * options, whether hard/softirqs are running, level of
556 * preempt_disabled respectively). **N** means that
557 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
558 * are set.
559 * * ``419421.045894`` is a timestamp.
560 * * ``0x00000001`` is a fake value used by BPF for the
561 * instruction pointer register.
562 * * ``<formatted msg>`` is the message formatted with
563 * *fmt*.
564 *
565 * The conversion specifiers supported by *fmt* are similar, but
566 * more limited than for printk(). They are **%d**, **%i**,
567 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
568 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
569 * of field, padding with zeroes, etc.) is available, and the
570 * helper will return **-EINVAL** (but print nothing) if it
571 * encounters an unknown specifier.
572 *
573 * Also, note that **bpf_trace_printk**\ () is slow, and should
574 * only be used for debugging purposes. For this reason, a notice
575 * bloc (spanning several lines) is printed to kernel logs and
576 * states that the helper should not be used "for production use"
577 * the first time this helper is used (or more precisely, when
578 * **trace_printk**\ () buffers are allocated). For passing values
579 * to user space, perf events should be preferred.
580 * Return
581 * The number of bytes written to the buffer, or a negative error
582 * in case of failure.
583 *
584 * u32 bpf_get_prandom_u32(void)
585 * Description
586 * Get a pseudo-random number.
587 *
588 * From a security point of view, this helper uses its own
589 * pseudo-random internal state, and cannot be used to infer the
590 * seed of other random functions in the kernel. However, it is
591 * essential to note that the generator used by the helper is not
592 * cryptographically secure.
593 * Return
594 * A random 32-bit unsigned value.
595 *
596 * u32 bpf_get_smp_processor_id(void)
597 * Description
598 * Get the SMP (symmetric multiprocessing) processor id. Note that
599 * all programs run with preemption disabled, which means that the
600 * SMP processor id is stable during all the execution of the
601 * program.
602 * Return
603 * The SMP id of the processor running the program.
604 *
605 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
606 * Description
607 * Store *len* bytes from address *from* into the packet
608 * associated to *skb*, at *offset*. *flags* are a combination of
609 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
610 * checksum for the packet after storing the bytes) and
611 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
612 * **->swhash** and *skb*\ **->l4hash** to 0).
613 *
614 * A call to this helper is susceptible to change the underlaying
615 * packet buffer. Therefore, at load time, all checks on pointers
616 * previously done by the verifier are invalidated and must be
617 * performed again, if the helper is used in combination with
618 * direct packet access.
619 * Return
620 * 0 on success, or a negative error in case of failure.
621 *
622 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
623 * Description
624 * Recompute the layer 3 (e.g. IP) checksum for the packet
625 * associated to *skb*. Computation is incremental, so the helper
626 * must know the former value of the header field that was
627 * modified (*from*), the new value of this field (*to*), and the
628 * number of bytes (2 or 4) for this field, stored in *size*.
629 * Alternatively, it is possible to store the difference between
630 * the previous and the new values of the header field in *to*, by
631 * setting *from* and *size* to 0. For both methods, *offset*
632 * indicates the location of the IP checksum within the packet.
633 *
634 * This helper works in combination with **bpf_csum_diff**\ (),
635 * which does not update the checksum in-place, but offers more
636 * flexibility and can handle sizes larger than 2 or 4 for the
637 * checksum to update.
638 *
639 * A call to this helper is susceptible to change the underlaying
640 * packet buffer. Therefore, at load time, all checks on pointers
641 * previously done by the verifier are invalidated and must be
642 * performed again, if the helper is used in combination with
643 * direct packet access.
644 * Return
645 * 0 on success, or a negative error in case of failure.
646 *
647 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
648 * Description
649 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
650 * packet associated to *skb*. Computation is incremental, so the
651 * helper must know the former value of the header field that was
652 * modified (*from*), the new value of this field (*to*), and the
653 * number of bytes (2 or 4) for this field, stored on the lowest
654 * four bits of *flags*. Alternatively, it is possible to store
655 * the difference between the previous and the new values of the
656 * header field in *to*, by setting *from* and the four lowest
657 * bits of *flags* to 0. For both methods, *offset* indicates the
658 * location of the IP checksum within the packet. In addition to
659 * the size of the field, *flags* can be added (bitwise OR) actual
660 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
661 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
662 * for updates resulting in a null checksum the value is set to
663 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
664 * the checksum is to be computed against a pseudo-header.
665 *
666 * This helper works in combination with **bpf_csum_diff**\ (),
667 * which does not update the checksum in-place, but offers more
668 * flexibility and can handle sizes larger than 2 or 4 for the
669 * checksum to update.
670 *
671 * A call to this helper is susceptible to change the underlaying
672 * packet buffer. Therefore, at load time, all checks on pointers
673 * previously done by the verifier are invalidated and must be
674 * performed again, if the helper is used in combination with
675 * direct packet access.
676 * Return
677 * 0 on success, or a negative error in case of failure.
678 *
679 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
680 * Description
681 * This special helper is used to trigger a "tail call", or in
682 * other words, to jump into another eBPF program. The same stack
683 * frame is used (but values on stack and in registers for the
684 * caller are not accessible to the callee). This mechanism allows
685 * for program chaining, either for raising the maximum number of
686 * available eBPF instructions, or to execute given programs in
687 * conditional blocks. For security reasons, there is an upper
688 * limit to the number of successive tail calls that can be
689 * performed.
690 *
691 * Upon call of this helper, the program attempts to jump into a
692 * program referenced at index *index* in *prog_array_map*, a
693 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
694 * *ctx*, a pointer to the context.
695 *
696 * If the call succeeds, the kernel immediately runs the first
697 * instruction of the new program. This is not a function call,
698 * and it never returns to the previous program. If the call
699 * fails, then the helper has no effect, and the caller continues
700 * to run its subsequent instructions. A call can fail if the
701 * destination program for the jump does not exist (i.e. *index*
702 * is superior to the number of entries in *prog_array_map*), or
703 * if the maximum number of tail calls has been reached for this
704 * chain of programs. This limit is defined in the kernel by the
705 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
706 * which is currently set to 32.
707 * Return
708 * 0 on success, or a negative error in case of failure.
709 *
710 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
711 * Description
712 * Clone and redirect the packet associated to *skb* to another
713 * net device of index *ifindex*. Both ingress and egress
714 * interfaces can be used for redirection. The **BPF_F_INGRESS**
715 * value in *flags* is used to make the distinction (ingress path
716 * is selected if the flag is present, egress path otherwise).
717 * This is the only flag supported for now.
718 *
719 * In comparison with **bpf_redirect**\ () helper,
720 * **bpf_clone_redirect**\ () has the associated cost of
721 * duplicating the packet buffer, but this can be executed out of
722 * the eBPF program. Conversely, **bpf_redirect**\ () is more
723 * efficient, but it is handled through an action code where the
724 * redirection happens only after the eBPF program has returned.
725 *
726 * A call to this helper is susceptible to change the underlaying
727 * packet buffer. Therefore, at load time, all checks on pointers
728 * previously done by the verifier are invalidated and must be
729 * performed again, if the helper is used in combination with
730 * direct packet access.
731 * Return
732 * 0 on success, or a negative error in case of failure.
733 *
734 * u64 bpf_get_current_pid_tgid(void)
735 * Return
736 * A 64-bit integer containing the current tgid and pid, and
737 * created as such:
738 * *current_task*\ **->tgid << 32 \|**
739 * *current_task*\ **->pid**.
740 *
741 * u64 bpf_get_current_uid_gid(void)
742 * Return
743 * A 64-bit integer containing the current GID and UID, and
744 * created as such: *current_gid* **<< 32 \|** *current_uid*.
745 *
746 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
747 * Description
748 * Copy the **comm** attribute of the current task into *buf* of
749 * *size_of_buf*. The **comm** attribute contains the name of
750 * the executable (excluding the path) for the current task. The
751 * *size_of_buf* must be strictly positive. On success, the
752 * helper makes sure that the *buf* is NUL-terminated. On failure,
753 * it is filled with zeroes.
754 * Return
755 * 0 on success, or a negative error in case of failure.
756 *
757 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
758 * Description
759 * Retrieve the classid for the current task, i.e. for the net_cls
760 * cgroup to which *skb* belongs.
761 *
762 * This helper can be used on TC egress path, but not on ingress.
763 *
764 * The net_cls cgroup provides an interface to tag network packets
765 * based on a user-provided identifier for all traffic coming from
766 * the tasks belonging to the related cgroup. See also the related
767 * kernel documentation, available from the Linux sources in file
768 * *Documentation/cgroup-v1/net_cls.txt*.
769 *
770 * The Linux kernel has two versions for cgroups: there are
771 * cgroups v1 and cgroups v2. Both are available to users, who can
772 * use a mixture of them, but note that the net_cls cgroup is for
773 * cgroup v1 only. This makes it incompatible with BPF programs
774 * run on cgroups, which is a cgroup-v2-only feature (a socket can
775 * only hold data for one version of cgroups at a time).
776 *
777 * This helper is only available is the kernel was compiled with
778 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
779 * "**y**" or to "**m**".
780 * Return
781 * The classid, or 0 for the default unconfigured classid.
782 *
783 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
784 * Description
785 * Push a *vlan_tci* (VLAN tag control information) of protocol
786 * *vlan_proto* to the packet associated to *skb*, then update
787 * the checksum. Note that if *vlan_proto* is different from
788 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
789 * be **ETH_P_8021Q**.
790 *
791 * A call to this helper is susceptible to change the underlaying
792 * packet buffer. Therefore, at load time, all checks on pointers
793 * previously done by the verifier are invalidated and must be
794 * performed again, if the helper is used in combination with
795 * direct packet access.
796 * Return
797 * 0 on success, or a negative error in case of failure.
798 *
799 * int bpf_skb_vlan_pop(struct sk_buff *skb)
800 * Description
801 * Pop a VLAN header from the packet associated to *skb*.
802 *
803 * A call to this helper is susceptible to change the underlaying
804 * packet buffer. Therefore, at load time, all checks on pointers
805 * previously done by the verifier are invalidated and must be
806 * performed again, if the helper is used in combination with
807 * direct packet access.
808 * Return
809 * 0 on success, or a negative error in case of failure.
810 *
811 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
812 * Description
813 * Get tunnel metadata. This helper takes a pointer *key* to an
814 * empty **struct bpf_tunnel_key** of **size**, that will be
815 * filled with tunnel metadata for the packet associated to *skb*.
816 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
817 * indicates that the tunnel is based on IPv6 protocol instead of
818 * IPv4.
819 *
820 * The **struct bpf_tunnel_key** is an object that generalizes the
821 * principal parameters used by various tunneling protocols into a
822 * single struct. This way, it can be used to easily make a
823 * decision based on the contents of the encapsulation header,
824 * "summarized" in this struct. In particular, it holds the IP
825 * address of the remote end (IPv4 or IPv6, depending on the case)
826 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
827 * this struct exposes the *key*\ **->tunnel_id**, which is
828 * generally mapped to a VNI (Virtual Network Identifier), making
829 * it programmable together with the **bpf_skb_set_tunnel_key**\
830 * () helper.
831 *
832 * Let's imagine that the following code is part of a program
833 * attached to the TC ingress interface, on one end of a GRE
834 * tunnel, and is supposed to filter out all messages coming from
835 * remote ends with IPv4 address other than 10.0.0.1:
836 *
837 * ::
838 *
839 * int ret;
840 * struct bpf_tunnel_key key = {};
841 *
842 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
843 * if (ret < 0)
844 * return TC_ACT_SHOT; // drop packet
845 *
846 * if (key.remote_ipv4 != 0x0a000001)
847 * return TC_ACT_SHOT; // drop packet
848 *
849 * return TC_ACT_OK; // accept packet
850 *
851 * This interface can also be used with all encapsulation devices
852 * that can operate in "collect metadata" mode: instead of having
853 * one network device per specific configuration, the "collect
854 * metadata" mode only requires a single device where the
855 * configuration can be extracted from this helper.
856 *
857 * This can be used together with various tunnels such as VXLan,
858 * Geneve, GRE or IP in IP (IPIP).
859 * Return
860 * 0 on success, or a negative error in case of failure.
861 *
862 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
863 * Description
864 * Populate tunnel metadata for packet associated to *skb.* The
865 * tunnel metadata is set to the contents of *key*, of *size*. The
866 * *flags* can be set to a combination of the following values:
867 *
868 * **BPF_F_TUNINFO_IPV6**
869 * Indicate that the tunnel is based on IPv6 protocol
870 * instead of IPv4.
871 * **BPF_F_ZERO_CSUM_TX**
872 * For IPv4 packets, add a flag to tunnel metadata
873 * indicating that checksum computation should be skipped
874 * and checksum set to zeroes.
875 * **BPF_F_DONT_FRAGMENT**
876 * Add a flag to tunnel metadata indicating that the
877 * packet should not be fragmented.
878 * **BPF_F_SEQ_NUMBER**
879 * Add a flag to tunnel metadata indicating that a
880 * sequence number should be added to tunnel header before
881 * sending the packet. This flag was added for GRE
882 * encapsulation, but might be used with other protocols
883 * as well in the future.
884 *
885 * Here is a typical usage on the transmit path:
886 *
887 * ::
888 *
889 * struct bpf_tunnel_key key;
890 * populate key ...
891 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
892 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
893 *
894 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
895 * helper for additional information.
896 * Return
897 * 0 on success, or a negative error in case of failure.
898 *
899 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
900 * Description
901 * Read the value of a perf event counter. This helper relies on a
902 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
903 * the perf event counter is selected when *map* is updated with
904 * perf event file descriptors. The *map* is an array whose size
905 * is the number of available CPUs, and each cell contains a value
906 * relative to one CPU. The value to retrieve is indicated by
907 * *flags*, that contains the index of the CPU to look up, masked
908 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
909 * **BPF_F_CURRENT_CPU** to indicate that the value for the
910 * current CPU should be retrieved.
911 *
912 * Note that before Linux 4.13, only hardware perf event can be
913 * retrieved.
914 *
915 * Also, be aware that the newer helper
916 * **bpf_perf_event_read_value**\ () is recommended over
917 * **bpf_perf_event_read**\ () in general. The latter has some ABI
918 * quirks where error and counter value are used as a return code
919 * (which is wrong to do since ranges may overlap). This issue is
920 * fixed with **bpf_perf_event_read_value**\ (), which at the same
921 * time provides more features over the **bpf_perf_event_read**\
922 * () interface. Please refer to the description of
923 * **bpf_perf_event_read_value**\ () for details.
924 * Return
925 * The value of the perf event counter read from the map, or a
926 * negative error code in case of failure.
927 *
928 * int bpf_redirect(u32 ifindex, u64 flags)
929 * Description
930 * Redirect the packet to another net device of index *ifindex*.
931 * This helper is somewhat similar to **bpf_clone_redirect**\
932 * (), except that the packet is not cloned, which provides
933 * increased performance.
934 *
935 * Except for XDP, both ingress and egress interfaces can be used
936 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
937 * to make the distinction (ingress path is selected if the flag
938 * is present, egress path otherwise). Currently, XDP only
939 * supports redirection to the egress interface, and accepts no
940 * flag at all.
941 *
942 * The same effect can be attained with the more generic
943 * **bpf_redirect_map**\ (), which requires specific maps to be
944 * used but offers better performance.
945 * Return
946 * For XDP, the helper returns **XDP_REDIRECT** on success or
947 * **XDP_ABORTED** on error. For other program types, the values
948 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
949 * error.
950 *
951 * u32 bpf_get_route_realm(struct sk_buff *skb)
952 * Description
953 * Retrieve the realm or the route, that is to say the
954 * **tclassid** field of the destination for the *skb*. The
955 * indentifier retrieved is a user-provided tag, similar to the
956 * one used with the net_cls cgroup (see description for
957 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
958 * held by a route (a destination entry), not by a task.
959 *
960 * Retrieving this identifier works with the clsact TC egress hook
961 * (see also **tc-bpf(8)**), or alternatively on conventional
962 * classful egress qdiscs, but not on TC ingress path. In case of
963 * clsact TC egress hook, this has the advantage that, internally,
964 * the destination entry has not been dropped yet in the transmit
965 * path. Therefore, the destination entry does not need to be
966 * artificially held via **netif_keep_dst**\ () for a classful
967 * qdisc until the *skb* is freed.
968 *
969 * This helper is available only if the kernel was compiled with
970 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
971 * Return
972 * The realm of the route for the packet associated to *skb*, or 0
973 * if none was found.
974 *
975 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
976 * Description
977 * Write raw *data* blob into a special BPF perf event held by
978 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
979 * event must have the following attributes: **PERF_SAMPLE_RAW**
980 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
981 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
982 *
983 * The *flags* are used to indicate the index in *map* for which
984 * the value must be put, masked with **BPF_F_INDEX_MASK**.
985 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
986 * to indicate that the index of the current CPU core should be
987 * used.
988 *
989 * The value to write, of *size*, is passed through eBPF stack and
990 * pointed by *data*.
991 *
992 * The context of the program *ctx* needs also be passed to the
993 * helper.
994 *
995 * On user space, a program willing to read the values needs to
996 * call **perf_event_open**\ () on the perf event (either for
997 * one or for all CPUs) and to store the file descriptor into the
998 * *map*. This must be done before the eBPF program can send data
999 * into it. An example is available in file
1000 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1001 * tree (the eBPF program counterpart is in
1002 * *samples/bpf/trace_output_kern.c*).
1003 *
1004 * **bpf_perf_event_output**\ () achieves better performance
1005 * than **bpf_trace_printk**\ () for sharing data with user
1006 * space, and is much better suitable for streaming data from eBPF
1007 * programs.
1008 *
1009 * Note that this helper is not restricted to tracing use cases
1010 * and can be used with programs attached to TC or XDP as well,
1011 * where it allows for passing data to user space listeners. Data
1012 * can be:
1013 *
1014 * * Only custom structs,
1015 * * Only the packet payload, or
1016 * * A combination of both.
1017 * Return
1018 * 0 on success, or a negative error in case of failure.
1019 *
1020 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1021 * Description
1022 * This helper was provided as an easy way to load data from a
1023 * packet. It can be used to load *len* bytes from *offset* from
1024 * the packet associated to *skb*, into the buffer pointed by
1025 * *to*.
1026 *
1027 * Since Linux 4.7, usage of this helper has mostly been replaced
1028 * by "direct packet access", enabling packet data to be
1029 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1030 * pointing respectively to the first byte of packet data and to
1031 * the byte after the last byte of packet data. However, it
1032 * remains useful if one wishes to read large quantities of data
1033 * at once from a packet into the eBPF stack.
1034 * Return
1035 * 0 on success, or a negative error in case of failure.
1036 *
1037 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1038 * Description
1039 * Walk a user or a kernel stack and return its id. To achieve
1040 * this, the helper needs *ctx*, which is a pointer to the context
1041 * on which the tracing program is executed, and a pointer to a
1042 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1043 *
1044 * The last argument, *flags*, holds the number of stack frames to
1045 * skip (from 0 to 255), masked with
1046 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1047 * a combination of the following flags:
1048 *
1049 * **BPF_F_USER_STACK**
1050 * Collect a user space stack instead of a kernel stack.
1051 * **BPF_F_FAST_STACK_CMP**
1052 * Compare stacks by hash only.
1053 * **BPF_F_REUSE_STACKID**
1054 * If two different stacks hash into the same *stackid*,
1055 * discard the old one.
1056 *
1057 * The stack id retrieved is a 32 bit long integer handle which
1058 * can be further combined with other data (including other stack
1059 * ids) and used as a key into maps. This can be useful for
1060 * generating a variety of graphs (such as flame graphs or off-cpu
1061 * graphs).
1062 *
1063 * For walking a stack, this helper is an improvement over
1064 * **bpf_probe_read**\ (), which can be used with unrolled loops
1065 * but is not efficient and consumes a lot of eBPF instructions.
1066 * Instead, **bpf_get_stackid**\ () can collect up to
1067 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1068 * this limit can be controlled with the **sysctl** program, and
1069 * that it should be manually increased in order to profile long
1070 * user stacks (such as stacks for Java programs). To do so, use:
1071 *
1072 * ::
1073 *
1074 * # sysctl kernel.perf_event_max_stack=<new value>
1075 * Return
1076 * The positive or null stack id on success, or a negative error
1077 * in case of failure.
1078 *
1079 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1080 * Description
1081 * Compute a checksum difference, from the raw buffer pointed by
1082 * *from*, of length *from_size* (that must be a multiple of 4),
1083 * towards the raw buffer pointed by *to*, of size *to_size*
1084 * (same remark). An optional *seed* can be added to the value
1085 * (this can be cascaded, the seed may come from a previous call
1086 * to the helper).
1087 *
1088 * This is flexible enough to be used in several ways:
1089 *
1090 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1091 * checksum, it can be used when pushing new data.
1092 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1093 * checksum, it can be used when removing data from a packet.
1094 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1095 * can be used to compute a diff. Note that *from_size* and
1096 * *to_size* do not need to be equal.
1097 *
1098 * This helper can be used in combination with
1099 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1100 * which one can feed in the difference computed with
1101 * **bpf_csum_diff**\ ().
1102 * Return
1103 * The checksum result, or a negative error code in case of
1104 * failure.
1105 *
1106 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1107 * Description
1108 * Retrieve tunnel options metadata for the packet associated to
1109 * *skb*, and store the raw tunnel option data to the buffer *opt*
1110 * of *size*.
1111 *
1112 * This helper can be used with encapsulation devices that can
1113 * operate in "collect metadata" mode (please refer to the related
1114 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1115 * more details). A particular example where this can be used is
1116 * in combination with the Geneve encapsulation protocol, where it
1117 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1118 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1119 * the eBPF program. This allows for full customization of these
1120 * headers.
1121 * Return
1122 * The size of the option data retrieved.
1123 *
1124 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1125 * Description
1126 * Set tunnel options metadata for the packet associated to *skb*
1127 * to the option data contained in the raw buffer *opt* of *size*.
1128 *
1129 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1130 * helper for additional information.
1131 * Return
1132 * 0 on success, or a negative error in case of failure.
1133 *
1134 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1135 * Description
1136 * Change the protocol of the *skb* to *proto*. Currently
1137 * supported are transition from IPv4 to IPv6, and from IPv6 to
1138 * IPv4. The helper takes care of the groundwork for the
1139 * transition, including resizing the socket buffer. The eBPF
1140 * program is expected to fill the new headers, if any, via
1141 * **skb_store_bytes**\ () and to recompute the checksums with
1142 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1143 * (). The main case for this helper is to perform NAT64
1144 * operations out of an eBPF program.
1145 *
1146 * Internally, the GSO type is marked as dodgy so that headers are
1147 * checked and segments are recalculated by the GSO/GRO engine.
1148 * The size for GSO target is adapted as well.
1149 *
1150 * All values for *flags* are reserved for future usage, and must
1151 * be left at zero.
1152 *
1153 * A call to this helper is susceptible to change the underlaying
1154 * packet buffer. Therefore, at load time, all checks on pointers
1155 * previously done by the verifier are invalidated and must be
1156 * performed again, if the helper is used in combination with
1157 * direct packet access.
1158 * Return
1159 * 0 on success, or a negative error in case of failure.
1160 *
1161 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1162 * Description
1163 * Change the packet type for the packet associated to *skb*. This
1164 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1165 * the eBPF program does not have a write access to *skb*\
1166 * **->pkt_type** beside this helper. Using a helper here allows
1167 * for graceful handling of errors.
1168 *
1169 * The major use case is to change incoming *skb*s to
1170 * **PACKET_HOST** in a programmatic way instead of having to
1171 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1172 * example.
1173 *
1174 * Note that *type* only allows certain values. At this time, they
1175 * are:
1176 *
1177 * **PACKET_HOST**
1178 * Packet is for us.
1179 * **PACKET_BROADCAST**
1180 * Send packet to all.
1181 * **PACKET_MULTICAST**
1182 * Send packet to group.
1183 * **PACKET_OTHERHOST**
1184 * Send packet to someone else.
1185 * Return
1186 * 0 on success, or a negative error in case of failure.
1187 *
1188 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1189 * Description
1190 * Check whether *skb* is a descendant of the cgroup2 held by
1191 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1192 * Return
1193 * The return value depends on the result of the test, and can be:
1194 *
1195 * * 0, if the *skb* failed the cgroup2 descendant test.
1196 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1197 * * A negative error code, if an error occurred.
1198 *
1199 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1200 * Description
1201 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1202 * not set, in particular if the hash was cleared due to mangling,
1203 * recompute this hash. Later accesses to the hash can be done
1204 * directly with *skb*\ **->hash**.
1205 *
1206 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1207 * prototype with **bpf_skb_change_proto**\ (), or calling
1208 * **bpf_skb_store_bytes**\ () with the
1209 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1210 * the hash and to trigger a new computation for the next call to
1211 * **bpf_get_hash_recalc**\ ().
1212 * Return
1213 * The 32-bit hash.
1214 *
1215 * u64 bpf_get_current_task(void)
1216 * Return
1217 * A pointer to the current task struct.
1218 *
1219 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1220 * Description
1221 * Attempt in a safe way to write *len* bytes from the buffer
1222 * *src* to *dst* in memory. It only works for threads that are in
1223 * user context, and *dst* must be a valid user space address.
1224 *
1225 * This helper should not be used to implement any kind of
1226 * security mechanism because of TOC-TOU attacks, but rather to
1227 * debug, divert, and manipulate execution of semi-cooperative
1228 * processes.
1229 *
1230 * Keep in mind that this feature is meant for experiments, and it
1231 * has a risk of crashing the system and running programs.
1232 * Therefore, when an eBPF program using this helper is attached,
1233 * a warning including PID and process name is printed to kernel
1234 * logs.
1235 * Return
1236 * 0 on success, or a negative error in case of failure.
1237 *
1238 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1239 * Description
1240 * Check whether the probe is being run is the context of a given
1241 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1242 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1243 * Return
1244 * The return value depends on the result of the test, and can be:
1245 *
1246 * * 0, if the *skb* task belongs to the cgroup2.
1247 * * 1, if the *skb* task does not belong to the cgroup2.
1248 * * A negative error code, if an error occurred.
1249 *
1250 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1251 * Description
1252 * Resize (trim or grow) the packet associated to *skb* to the
1253 * new *len*. The *flags* are reserved for future usage, and must
1254 * be left at zero.
1255 *
1256 * The basic idea is that the helper performs the needed work to
1257 * change the size of the packet, then the eBPF program rewrites
1258 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1259 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1260 * and others. This helper is a slow path utility intended for
1261 * replies with control messages. And because it is targeted for
1262 * slow path, the helper itself can afford to be slow: it
1263 * implicitly linearizes, unclones and drops offloads from the
1264 * *skb*.
1265 *
1266 * A call to this helper is susceptible to change the underlaying
1267 * packet buffer. Therefore, at load time, all checks on pointers
1268 * previously done by the verifier are invalidated and must be
1269 * performed again, if the helper is used in combination with
1270 * direct packet access.
1271 * Return
1272 * 0 on success, or a negative error in case of failure.
1273 *
1274 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1275 * Description
1276 * Pull in non-linear data in case the *skb* is non-linear and not
1277 * all of *len* are part of the linear section. Make *len* bytes
1278 * from *skb* readable and writable. If a zero value is passed for
1279 * *len*, then the whole length of the *skb* is pulled.
1280 *
1281 * This helper is only needed for reading and writing with direct
1282 * packet access.
1283 *
1284 * For direct packet access, testing that offsets to access
1285 * are within packet boundaries (test on *skb*\ **->data_end**) is
1286 * susceptible to fail if offsets are invalid, or if the requested
1287 * data is in non-linear parts of the *skb*. On failure the
1288 * program can just bail out, or in the case of a non-linear
1289 * buffer, use a helper to make the data available. The
1290 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1291 * the data. Another one consists in using **bpf_skb_pull_data**
1292 * to pull in once the non-linear parts, then retesting and
1293 * eventually access the data.
1294 *
1295 * At the same time, this also makes sure the *skb* is uncloned,
1296 * which is a necessary condition for direct write. As this needs
1297 * to be an invariant for the write part only, the verifier
1298 * detects writes and adds a prologue that is calling
1299 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1300 * the very beginning in case it is indeed cloned.
1301 *
1302 * A call to this helper is susceptible to change the underlaying
1303 * packet buffer. Therefore, at load time, all checks on pointers
1304 * previously done by the verifier are invalidated and must be
1305 * performed again, if the helper is used in combination with
1306 * direct packet access.
1307 * Return
1308 * 0 on success, or a negative error in case of failure.
1309 *
1310 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1311 * Description
1312 * Add the checksum *csum* into *skb*\ **->csum** in case the
1313 * driver has supplied a checksum for the entire packet into that
1314 * field. Return an error otherwise. This helper is intended to be
1315 * used in combination with **bpf_csum_diff**\ (), in particular
1316 * when the checksum needs to be updated after data has been
1317 * written into the packet through direct packet access.
1318 * Return
1319 * The checksum on success, or a negative error code in case of
1320 * failure.
1321 *
1322 * void bpf_set_hash_invalid(struct sk_buff *skb)
1323 * Description
1324 * Invalidate the current *skb*\ **->hash**. It can be used after
1325 * mangling on headers through direct packet access, in order to
1326 * indicate that the hash is outdated and to trigger a
1327 * recalculation the next time the kernel tries to access this
1328 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1329 *
1330 * int bpf_get_numa_node_id(void)
1331 * Description
1332 * Return the id of the current NUMA node. The primary use case
1333 * for this helper is the selection of sockets for the local NUMA
1334 * node, when the program is attached to sockets using the
1335 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1336 * but the helper is also available to other eBPF program types,
1337 * similarly to **bpf_get_smp_processor_id**\ ().
1338 * Return
1339 * The id of current NUMA node.
1340 *
1341 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1342 * Description
1343 * Grows headroom of packet associated to *skb* and adjusts the
1344 * offset of the MAC header accordingly, adding *len* bytes of
1345 * space. It automatically extends and reallocates memory as
1346 * required.
1347 *
1348 * This helper can be used on a layer 3 *skb* to push a MAC header
1349 * for redirection into a layer 2 device.
1350 *
1351 * All values for *flags* are reserved for future usage, and must
1352 * be left at zero.
1353 *
1354 * A call to this helper is susceptible to change the underlaying
1355 * packet buffer. Therefore, at load time, all checks on pointers
1356 * previously done by the verifier are invalidated and must be
1357 * performed again, if the helper is used in combination with
1358 * direct packet access.
1359 * Return
1360 * 0 on success, or a negative error in case of failure.
1361 *
1362 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1363 * Description
1364 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1365 * it is possible to use a negative value for *delta*. This helper
1366 * can be used to prepare the packet for pushing or popping
1367 * headers.
1368 *
1369 * A call to this helper is susceptible to change the underlaying
1370 * packet buffer. Therefore, at load time, all checks on pointers
1371 * previously done by the verifier are invalidated and must be
1372 * performed again, if the helper is used in combination with
1373 * direct packet access.
1374 * Return
1375 * 0 on success, or a negative error in case of failure.
1376 *
1377 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1378 * Description
1379 * Copy a NUL terminated string from an unsafe address
1380 * *unsafe_ptr* to *dst*. The *size* should include the
1381 * terminating NUL byte. In case the string length is smaller than
1382 * *size*, the target is not padded with further NUL bytes. If the
1383 * string length is larger than *size*, just *size*-1 bytes are
1384 * copied and the last byte is set to NUL.
1385 *
1386 * On success, the length of the copied string is returned. This
1387 * makes this helper useful in tracing programs for reading
1388 * strings, and more importantly to get its length at runtime. See
1389 * the following snippet:
1390 *
1391 * ::
1392 *
1393 * SEC("kprobe/sys_open")
1394 * void bpf_sys_open(struct pt_regs *ctx)
1395 * {
1396 * char buf[PATHLEN]; // PATHLEN is defined to 256
1397 * int res = bpf_probe_read_str(buf, sizeof(buf),
1398 * ctx->di);
1399 *
1400 * // Consume buf, for example push it to
1401 * // userspace via bpf_perf_event_output(); we
1402 * // can use res (the string length) as event
1403 * // size, after checking its boundaries.
1404 * }
1405 *
1406 * In comparison, using **bpf_probe_read()** helper here instead
1407 * to read the string would require to estimate the length at
1408 * compile time, and would often result in copying more memory
1409 * than necessary.
1410 *
1411 * Another useful use case is when parsing individual process
1412 * arguments or individual environment variables navigating
1413 * *current*\ **->mm->arg_start** and *current*\
1414 * **->mm->env_start**: using this helper and the return value,
1415 * one can quickly iterate at the right offset of the memory area.
1416 * Return
1417 * On success, the strictly positive length of the string,
1418 * including the trailing NUL character. On error, a negative
1419 * value.
1420 *
1421 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1422 * Description
1423 * If the **struct sk_buff** pointed by *skb* has a known socket,
1424 * retrieve the cookie (generated by the kernel) of this socket.
1425 * If no cookie has been set yet, generate a new cookie. Once
1426 * generated, the socket cookie remains stable for the life of the
1427 * socket. This helper can be useful for monitoring per socket
1428 * networking traffic statistics as it provides a unique socket
1429 * identifier per namespace.
1430 * Return
1431 * A 8-byte long non-decreasing number on success, or 0 if the
1432 * socket field is missing inside *skb*.
1433 *
1434 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1435 * Description
1436 * Equivalent to bpf_get_socket_cookie() helper that accepts
1437 * *skb*, but gets socket from **struct bpf_sock_addr** contex.
1438 * Return
1439 * A 8-byte long non-decreasing number.
1440 *
1441 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1442 * Description
1443 * Equivalent to bpf_get_socket_cookie() helper that accepts
1444 * *skb*, but gets socket from **struct bpf_sock_ops** contex.
1445 * Return
1446 * A 8-byte long non-decreasing number.
1447 *
1448 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1449 * Return
1450 * The owner UID of the socket associated to *skb*. If the socket
1451 * is **NULL**, or if it is not a full socket (i.e. if it is a
1452 * time-wait or a request socket instead), **overflowuid** value
1453 * is returned (note that **overflowuid** might also be the actual
1454 * UID value for the socket).
1455 *
1456 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1457 * Description
1458 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1459 * to value *hash*.
1460 * Return
1461 * 0
1462 *
1463 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1464 * Description
1465 * Emulate a call to **setsockopt()** on the socket associated to
1466 * *bpf_socket*, which must be a full socket. The *level* at
1467 * which the option resides and the name *optname* of the option
1468 * must be specified, see **setsockopt(2)** for more information.
1469 * The option value of length *optlen* is pointed by *optval*.
1470 *
1471 * This helper actually implements a subset of **setsockopt()**.
1472 * It supports the following *level*\ s:
1473 *
1474 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1475 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1476 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1477 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1478 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1479 * **TCP_BPF_SNDCWND_CLAMP**.
1480 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1481 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1482 * Return
1483 * 0 on success, or a negative error in case of failure.
1484 *
1485 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1486 * Description
1487 * Grow or shrink the room for data in the packet associated to
1488 * *skb* by *len_diff*, and according to the selected *mode*.
1489 *
1490 * There is a single supported mode at this time:
1491 *
1492 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1493 * (room space is added or removed below the layer 3 header).
1494 *
1495 * All values for *flags* are reserved for future usage, and must
1496 * be left at zero.
1497 *
1498 * A call to this helper is susceptible to change the underlaying
1499 * packet buffer. Therefore, at load time, all checks on pointers
1500 * previously done by the verifier are invalidated and must be
1501 * performed again, if the helper is used in combination with
1502 * direct packet access.
1503 * Return
1504 * 0 on success, or a negative error in case of failure.
1505 *
1506 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1507 * Description
1508 * Redirect the packet to the endpoint referenced by *map* at
1509 * index *key*. Depending on its type, this *map* can contain
1510 * references to net devices (for forwarding packets through other
1511 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1512 * but this is only implemented for native XDP (with driver
1513 * support) as of this writing).
1514 *
1515 * All values for *flags* are reserved for future usage, and must
1516 * be left at zero.
1517 *
1518 * When used to redirect packets to net devices, this helper
1519 * provides a high performance increase over **bpf_redirect**\ ().
1520 * This is due to various implementation details of the underlying
1521 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1522 * () tries to send packet as a "bulk" to the device.
1523 * Return
1524 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1525 *
1526 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1527 * Description
1528 * Redirect the packet to the socket referenced by *map* (of type
1529 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1530 * egress interfaces can be used for redirection. The
1531 * **BPF_F_INGRESS** value in *flags* is used to make the
1532 * distinction (ingress path is selected if the flag is present,
1533 * egress path otherwise). This is the only flag supported for now.
1534 * Return
1535 * **SK_PASS** on success, or **SK_DROP** on error.
1536 *
1537 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1538 * Description
1539 * Add an entry to, or update a *map* referencing sockets. The
1540 * *skops* is used as a new value for the entry associated to
1541 * *key*. *flags* is one of:
1542 *
1543 * **BPF_NOEXIST**
1544 * The entry for *key* must not exist in the map.
1545 * **BPF_EXIST**
1546 * The entry for *key* must already exist in the map.
1547 * **BPF_ANY**
1548 * No condition on the existence of the entry for *key*.
1549 *
1550 * If the *map* has eBPF programs (parser and verdict), those will
1551 * be inherited by the socket being added. If the socket is
1552 * already attached to eBPF programs, this results in an error.
1553 * Return
1554 * 0 on success, or a negative error in case of failure.
1555 *
1556 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1557 * Description
1558 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1559 * *delta* (which can be positive or negative). Note that this
1560 * operation modifies the address stored in *xdp_md*\ **->data**,
1561 * so the latter must be loaded only after the helper has been
1562 * called.
1563 *
1564 * The use of *xdp_md*\ **->data_meta** is optional and programs
1565 * are not required to use it. The rationale is that when the
1566 * packet is processed with XDP (e.g. as DoS filter), it is
1567 * possible to push further meta data along with it before passing
1568 * to the stack, and to give the guarantee that an ingress eBPF
1569 * program attached as a TC classifier on the same device can pick
1570 * this up for further post-processing. Since TC works with socket
1571 * buffers, it remains possible to set from XDP the **mark** or
1572 * **priority** pointers, or other pointers for the socket buffer.
1573 * Having this scratch space generic and programmable allows for
1574 * more flexibility as the user is free to store whatever meta
1575 * data they need.
1576 *
1577 * A call to this helper is susceptible to change the underlaying
1578 * packet buffer. Therefore, at load time, all checks on pointers
1579 * previously done by the verifier are invalidated and must be
1580 * performed again, if the helper is used in combination with
1581 * direct packet access.
1582 * Return
1583 * 0 on success, or a negative error in case of failure.
1584 *
1585 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1586 * Description
1587 * Read the value of a perf event counter, and store it into *buf*
1588 * of size *buf_size*. This helper relies on a *map* of type
1589 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1590 * counter is selected when *map* is updated with perf event file
1591 * descriptors. The *map* is an array whose size is the number of
1592 * available CPUs, and each cell contains a value relative to one
1593 * CPU. The value to retrieve is indicated by *flags*, that
1594 * contains the index of the CPU to look up, masked with
1595 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1596 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1597 * current CPU should be retrieved.
1598 *
1599 * This helper behaves in a way close to
1600 * **bpf_perf_event_read**\ () helper, save that instead of
1601 * just returning the value observed, it fills the *buf*
1602 * structure. This allows for additional data to be retrieved: in
1603 * particular, the enabled and running times (in *buf*\
1604 * **->enabled** and *buf*\ **->running**, respectively) are
1605 * copied. In general, **bpf_perf_event_read_value**\ () is
1606 * recommended over **bpf_perf_event_read**\ (), which has some
1607 * ABI issues and provides fewer functionalities.
1608 *
1609 * These values are interesting, because hardware PMU (Performance
1610 * Monitoring Unit) counters are limited resources. When there are
1611 * more PMU based perf events opened than available counters,
1612 * kernel will multiplex these events so each event gets certain
1613 * percentage (but not all) of the PMU time. In case that
1614 * multiplexing happens, the number of samples or counter value
1615 * will not reflect the case compared to when no multiplexing
1616 * occurs. This makes comparison between different runs difficult.
1617 * Typically, the counter value should be normalized before
1618 * comparing to other experiments. The usual normalization is done
1619 * as follows.
1620 *
1621 * ::
1622 *
1623 * normalized_counter = counter * t_enabled / t_running
1624 *
1625 * Where t_enabled is the time enabled for event and t_running is
1626 * the time running for event since last normalization. The
1627 * enabled and running times are accumulated since the perf event
1628 * open. To achieve scaling factor between two invocations of an
1629 * eBPF program, users can can use CPU id as the key (which is
1630 * typical for perf array usage model) to remember the previous
1631 * value and do the calculation inside the eBPF program.
1632 * Return
1633 * 0 on success, or a negative error in case of failure.
1634 *
1635 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1636 * Description
1637 * For en eBPF program attached to a perf event, retrieve the
1638 * value of the event counter associated to *ctx* and store it in
1639 * the structure pointed by *buf* and of size *buf_size*. Enabled
1640 * and running times are also stored in the structure (see
1641 * description of helper **bpf_perf_event_read_value**\ () for
1642 * more details).
1643 * Return
1644 * 0 on success, or a negative error in case of failure.
1645 *
1646 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1647 * Description
1648 * Emulate a call to **getsockopt()** on the socket associated to
1649 * *bpf_socket*, which must be a full socket. The *level* at
1650 * which the option resides and the name *optname* of the option
1651 * must be specified, see **getsockopt(2)** for more information.
1652 * The retrieved value is stored in the structure pointed by
1653 * *opval* and of length *optlen*.
1654 *
1655 * This helper actually implements a subset of **getsockopt()**.
1656 * It supports the following *level*\ s:
1657 *
1658 * * **IPPROTO_TCP**, which supports *optname*
1659 * **TCP_CONGESTION**.
1660 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1661 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1662 * Return
1663 * 0 on success, or a negative error in case of failure.
1664 *
1665 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1666 * Description
1667 * Used for error injection, this helper uses kprobes to override
1668 * the return value of the probed function, and to set it to *rc*.
1669 * The first argument is the context *regs* on which the kprobe
1670 * works.
1671 *
1672 * This helper works by setting setting the PC (program counter)
1673 * to an override function which is run in place of the original
1674 * probed function. This means the probed function is not run at
1675 * all. The replacement function just returns with the required
1676 * value.
1677 *
1678 * This helper has security implications, and thus is subject to
1679 * restrictions. It is only available if the kernel was compiled
1680 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1681 * option, and in this case it only works on functions tagged with
1682 * **ALLOW_ERROR_INJECTION** in the kernel code.
1683 *
1684 * Also, the helper is only available for the architectures having
1685 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1686 * x86 architecture is the only one to support this feature.
1687 * Return
1688 * 0
1689 *
1690 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1691 * Description
1692 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1693 * for the full TCP socket associated to *bpf_sock_ops* to
1694 * *argval*.
1695 *
1696 * The primary use of this field is to determine if there should
1697 * be calls to eBPF programs of type
1698 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1699 * code. A program of the same type can change its value, per
1700 * connection and as necessary, when the connection is
1701 * established. This field is directly accessible for reading, but
1702 * this helper must be used for updates in order to return an
1703 * error if an eBPF program tries to set a callback that is not
1704 * supported in the current kernel.
1705 *
1706 * The supported callback values that *argval* can combine are:
1707 *
1708 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1709 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1710 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1711 *
1712 * Here are some examples of where one could call such eBPF
1713 * program:
1714 *
1715 * * When RTO fires.
1716 * * When a packet is retransmitted.
1717 * * When the connection terminates.
1718 * * When a packet is sent.
1719 * * When a packet is received.
1720 * Return
1721 * Code **-EINVAL** if the socket is not a full TCP socket;
1722 * otherwise, a positive number containing the bits that could not
1723 * be set is returned (which comes down to 0 if all bits were set
1724 * as required).
1725 *
1726 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1727 * Description
1728 * This helper is used in programs implementing policies at the
1729 * socket level. If the message *msg* is allowed to pass (i.e. if
1730 * the verdict eBPF program returns **SK_PASS**), redirect it to
1731 * the socket referenced by *map* (of type
1732 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1733 * egress interfaces can be used for redirection. The
1734 * **BPF_F_INGRESS** value in *flags* is used to make the
1735 * distinction (ingress path is selected if the flag is present,
1736 * egress path otherwise). This is the only flag supported for now.
1737 * Return
1738 * **SK_PASS** on success, or **SK_DROP** on error.
1739 *
1740 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1741 * Description
1742 * For socket policies, apply the verdict of the eBPF program to
1743 * the next *bytes* (number of bytes) of message *msg*.
1744 *
1745 * For example, this helper can be used in the following cases:
1746 *
1747 * * A single **sendmsg**\ () or **sendfile**\ () system call
1748 * contains multiple logical messages that the eBPF program is
1749 * supposed to read and for which it should apply a verdict.
1750 * * An eBPF program only cares to read the first *bytes* of a
1751 * *msg*. If the message has a large payload, then setting up
1752 * and calling the eBPF program repeatedly for all bytes, even
1753 * though the verdict is already known, would create unnecessary
1754 * overhead.
1755 *
1756 * When called from within an eBPF program, the helper sets a
1757 * counter internal to the BPF infrastructure, that is used to
1758 * apply the last verdict to the next *bytes*. If *bytes* is
1759 * smaller than the current data being processed from a
1760 * **sendmsg**\ () or **sendfile**\ () system call, the first
1761 * *bytes* will be sent and the eBPF program will be re-run with
1762 * the pointer for start of data pointing to byte number *bytes*
1763 * **+ 1**. If *bytes* is larger than the current data being
1764 * processed, then the eBPF verdict will be applied to multiple
1765 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1766 * consumed.
1767 *
1768 * Note that if a socket closes with the internal counter holding
1769 * a non-zero value, this is not a problem because data is not
1770 * being buffered for *bytes* and is sent as it is received.
1771 * Return
1772 * 0
1773 *
1774 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1775 * Description
1776 * For socket policies, prevent the execution of the verdict eBPF
1777 * program for message *msg* until *bytes* (byte number) have been
1778 * accumulated.
1779 *
1780 * This can be used when one needs a specific number of bytes
1781 * before a verdict can be assigned, even if the data spans
1782 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1783 * case would be a user calling **sendmsg**\ () repeatedly with
1784 * 1-byte long message segments. Obviously, this is bad for
1785 * performance, but it is still valid. If the eBPF program needs
1786 * *bytes* bytes to validate a header, this helper can be used to
1787 * prevent the eBPF program to be called again until *bytes* have
1788 * been accumulated.
1789 * Return
1790 * 0
1791 *
1792 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1793 * Description
1794 * For socket policies, pull in non-linear data from user space
1795 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1796 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1797 * respectively.
1798 *
1799 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1800 * *msg* it can only parse data that the (**data**, **data_end**)
1801 * pointers have already consumed. For **sendmsg**\ () hooks this
1802 * is likely the first scatterlist element. But for calls relying
1803 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1804 * be the range (**0**, **0**) because the data is shared with
1805 * user space and by default the objective is to avoid allowing
1806 * user space to modify data while (or after) eBPF verdict is
1807 * being decided. This helper can be used to pull in data and to
1808 * set the start and end pointer to given values. Data will be
1809 * copied if necessary (i.e. if data was not linear and if start
1810 * and end pointers do not point to the same chunk).
1811 *
1812 * A call to this helper is susceptible to change the underlaying
1813 * packet buffer. Therefore, at load time, all checks on pointers
1814 * previously done by the verifier are invalidated and must be
1815 * performed again, if the helper is used in combination with
1816 * direct packet access.
1817 *
1818 * All values for *flags* are reserved for future usage, and must
1819 * be left at zero.
1820 * Return
1821 * 0 on success, or a negative error in case of failure.
1822 *
1823 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1824 * Description
1825 * Bind the socket associated to *ctx* to the address pointed by
1826 * *addr*, of length *addr_len*. This allows for making outgoing
1827 * connection from the desired IP address, which can be useful for
1828 * example when all processes inside a cgroup should use one
1829 * single IP address on a host that has multiple IP configured.
1830 *
1831 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1832 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1833 * **AF_INET6**). Looking for a free port to bind to can be
1834 * expensive, therefore binding to port is not permitted by the
1835 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1836 * must be set to zero.
1837 * Return
1838 * 0 on success, or a negative error in case of failure.
1839 *
1840 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1841 * Description
1842 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1843 * only possible to shrink the packet as of this writing,
1844 * therefore *delta* must be a negative integer.
1845 *
1846 * A call to this helper is susceptible to change the underlaying
1847 * packet buffer. Therefore, at load time, all checks on pointers
1848 * previously done by the verifier are invalidated and must be
1849 * performed again, if the helper is used in combination with
1850 * direct packet access.
1851 * Return
1852 * 0 on success, or a negative error in case of failure.
1853 *
1854 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1855 * Description
1856 * Retrieve the XFRM state (IP transform framework, see also
1857 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1858 *
1859 * The retrieved value is stored in the **struct bpf_xfrm_state**
1860 * pointed by *xfrm_state* and of length *size*.
1861 *
1862 * All values for *flags* are reserved for future usage, and must
1863 * be left at zero.
1864 *
1865 * This helper is available only if the kernel was compiled with
1866 * **CONFIG_XFRM** configuration option.
1867 * Return
1868 * 0 on success, or a negative error in case of failure.
1869 *
1870 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1871 * Description
1872 * Return a user or a kernel stack in bpf program provided buffer.
1873 * To achieve this, the helper needs *ctx*, which is a pointer
1874 * to the context on which the tracing program is executed.
1875 * To store the stacktrace, the bpf program provides *buf* with
1876 * a nonnegative *size*.
1877 *
1878 * The last argument, *flags*, holds the number of stack frames to
1879 * skip (from 0 to 255), masked with
1880 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1881 * the following flags:
1882 *
1883 * **BPF_F_USER_STACK**
1884 * Collect a user space stack instead of a kernel stack.
1885 * **BPF_F_USER_BUILD_ID**
1886 * Collect buildid+offset instead of ips for user stack,
1887 * only valid if **BPF_F_USER_STACK** is also specified.
1888 *
1889 * **bpf_get_stack**\ () can collect up to
1890 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1891 * to sufficient large buffer size. Note that
1892 * this limit can be controlled with the **sysctl** program, and
1893 * that it should be manually increased in order to profile long
1894 * user stacks (such as stacks for Java programs). To do so, use:
1895 *
1896 * ::
1897 *
1898 * # sysctl kernel.perf_event_max_stack=<new value>
1899 * Return
1900 * A non-negative value equal to or less than *size* on success,
1901 * or a negative error in case of failure.
1902 *
1903 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1904 * Description
1905 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1906 * it provides an easy way to load *len* bytes from *offset*
1907 * from the packet associated to *skb*, into the buffer pointed
1908 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1909 * a fifth argument *start_header* exists in order to select a
1910 * base offset to start from. *start_header* can be one of:
1911 *
1912 * **BPF_HDR_START_MAC**
1913 * Base offset to load data from is *skb*'s mac header.
1914 * **BPF_HDR_START_NET**
1915 * Base offset to load data from is *skb*'s network header.
1916 *
1917 * In general, "direct packet access" is the preferred method to
1918 * access packet data, however, this helper is in particular useful
1919 * in socket filters where *skb*\ **->data** does not always point
1920 * to the start of the mac header and where "direct packet access"
1921 * is not available.
1922 * Return
1923 * 0 on success, or a negative error in case of failure.
1924 *
1925 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1926 * Description
1927 * Do FIB lookup in kernel tables using parameters in *params*.
1928 * If lookup is successful and result shows packet is to be
1929 * forwarded, the neighbor tables are searched for the nexthop.
1930 * If successful (ie., FIB lookup shows forwarding and nexthop
1931 * is resolved), the nexthop address is returned in ipv4_dst
1932 * or ipv6_dst based on family, smac is set to mac address of
1933 * egress device, dmac is set to nexthop mac address, rt_metric
1934 * is set to metric from route (IPv4/IPv6 only), and ifindex
1935 * is set to the device index of the nexthop from the FIB lookup.
1936 *
1937 * *plen* argument is the size of the passed in struct.
1938 * *flags* argument can be a combination of one or more of the
1939 * following values:
1940 *
1941 * **BPF_FIB_LOOKUP_DIRECT**
1942 * Do a direct table lookup vs full lookup using FIB
1943 * rules.
1944 * **BPF_FIB_LOOKUP_OUTPUT**
1945 * Perform lookup from an egress perspective (default is
1946 * ingress).
1947 *
1948 * *ctx* is either **struct xdp_md** for XDP programs or
1949 * **struct sk_buff** tc cls_act programs.
1950 * Return
1951 * * < 0 if any input argument is invalid
1952 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1953 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1954 * packet is not forwarded or needs assist from full stack
1955 *
1956 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1957 * Description
1958 * Add an entry to, or update a sockhash *map* referencing sockets.
1959 * The *skops* is used as a new value for the entry associated to
1960 * *key*. *flags* is one of:
1961 *
1962 * **BPF_NOEXIST**
1963 * The entry for *key* must not exist in the map.
1964 * **BPF_EXIST**
1965 * The entry for *key* must already exist in the map.
1966 * **BPF_ANY**
1967 * No condition on the existence of the entry for *key*.
1968 *
1969 * If the *map* has eBPF programs (parser and verdict), those will
1970 * be inherited by the socket being added. If the socket is
1971 * already attached to eBPF programs, this results in an error.
1972 * Return
1973 * 0 on success, or a negative error in case of failure.
1974 *
1975 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1976 * Description
1977 * This helper is used in programs implementing policies at the
1978 * socket level. If the message *msg* is allowed to pass (i.e. if
1979 * the verdict eBPF program returns **SK_PASS**), redirect it to
1980 * the socket referenced by *map* (of type
1981 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1982 * egress interfaces can be used for redirection. The
1983 * **BPF_F_INGRESS** value in *flags* is used to make the
1984 * distinction (ingress path is selected if the flag is present,
1985 * egress path otherwise). This is the only flag supported for now.
1986 * Return
1987 * **SK_PASS** on success, or **SK_DROP** on error.
1988 *
1989 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1990 * Description
1991 * This helper is used in programs implementing policies at the
1992 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1993 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1994 * to the socket referenced by *map* (of type
1995 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1996 * egress interfaces can be used for redirection. The
1997 * **BPF_F_INGRESS** value in *flags* is used to make the
1998 * distinction (ingress path is selected if the flag is present,
1999 * egress otherwise). This is the only flag supported for now.
2000 * Return
2001 * **SK_PASS** on success, or **SK_DROP** on error.
2002 *
2003 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2004 * Description
2005 * Encapsulate the packet associated to *skb* within a Layer 3
2006 * protocol header. This header is provided in the buffer at
2007 * address *hdr*, with *len* its size in bytes. *type* indicates
2008 * the protocol of the header and can be one of:
2009 *
2010 * **BPF_LWT_ENCAP_SEG6**
2011 * IPv6 encapsulation with Segment Routing Header
2012 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2013 * the IPv6 header is computed by the kernel.
2014 * **BPF_LWT_ENCAP_SEG6_INLINE**
2015 * Only works if *skb* contains an IPv6 packet. Insert a
2016 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2017 * the IPv6 header.
2018 *
2019 * A call to this helper is susceptible to change the underlaying
2020 * packet buffer. Therefore, at load time, all checks on pointers
2021 * previously done by the verifier are invalidated and must be
2022 * performed again, if the helper is used in combination with
2023 * direct packet access.
2024 * Return
2025 * 0 on success, or a negative error in case of failure.
2026 *
2027 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2028 * Description
2029 * Store *len* bytes from address *from* into the packet
2030 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2031 * inside the outermost IPv6 Segment Routing Header can be
2032 * modified through this helper.
2033 *
2034 * A call to this helper is susceptible to change the underlaying
2035 * packet buffer. Therefore, at load time, all checks on pointers
2036 * previously done by the verifier are invalidated and must be
2037 * performed again, if the helper is used in combination with
2038 * direct packet access.
2039 * Return
2040 * 0 on success, or a negative error in case of failure.
2041 *
2042 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2043 * Description
2044 * Adjust the size allocated to TLVs in the outermost IPv6
2045 * Segment Routing Header contained in the packet associated to
2046 * *skb*, at position *offset* by *delta* bytes. Only offsets
2047 * after the segments are accepted. *delta* can be as well
2048 * positive (growing) as negative (shrinking).
2049 *
2050 * A call to this helper is susceptible to change the underlaying
2051 * packet buffer. Therefore, at load time, all checks on pointers
2052 * previously done by the verifier are invalidated and must be
2053 * performed again, if the helper is used in combination with
2054 * direct packet access.
2055 * Return
2056 * 0 on success, or a negative error in case of failure.
2057 *
2058 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2059 * Description
2060 * Apply an IPv6 Segment Routing action of type *action* to the
2061 * packet associated to *skb*. Each action takes a parameter
2062 * contained at address *param*, and of length *param_len* bytes.
2063 * *action* can be one of:
2064 *
2065 * **SEG6_LOCAL_ACTION_END_X**
2066 * End.X action: Endpoint with Layer-3 cross-connect.
2067 * Type of *param*: **struct in6_addr**.
2068 * **SEG6_LOCAL_ACTION_END_T**
2069 * End.T action: Endpoint with specific IPv6 table lookup.
2070 * Type of *param*: **int**.
2071 * **SEG6_LOCAL_ACTION_END_B6**
2072 * End.B6 action: Endpoint bound to an SRv6 policy.
2073 * Type of param: **struct ipv6_sr_hdr**.
2074 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2075 * End.B6.Encap action: Endpoint bound to an SRv6
2076 * encapsulation policy.
2077 * Type of param: **struct ipv6_sr_hdr**.
2078 *
2079 * A call to this helper is susceptible to change the underlaying
2080 * packet buffer. Therefore, at load time, all checks on pointers
2081 * previously done by the verifier are invalidated and must be
2082 * performed again, if the helper is used in combination with
2083 * direct packet access.
2084 * Return
2085 * 0 on success, or a negative error in case of failure.
2086 *
2087 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2088 * Description
2089 * This helper is used in programs implementing IR decoding, to
2090 * report a successfully decoded key press with *scancode*,
2091 * *toggle* value in the given *protocol*. The scancode will be
2092 * translated to a keycode using the rc keymap, and reported as
2093 * an input key down event. After a period a key up event is
2094 * generated. This period can be extended by calling either
2095 * **bpf_rc_keydown**\ () again with the same values, or calling
2096 * **bpf_rc_repeat**\ ().
2097 *
2098 * Some protocols include a toggle bit, in case the button was
2099 * released and pressed again between consecutive scancodes.
2100 *
2101 * The *ctx* should point to the lirc sample as passed into
2102 * the program.
2103 *
2104 * The *protocol* is the decoded protocol number (see
2105 * **enum rc_proto** for some predefined values).
2106 *
2107 * This helper is only available is the kernel was compiled with
2108 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2109 * "**y**".
2110 * Return
2111 * 0
2112 *
2113 * int bpf_rc_repeat(void *ctx)
2114 * Description
2115 * This helper is used in programs implementing IR decoding, to
2116 * report a successfully decoded repeat key message. This delays
2117 * the generation of a key up event for previously generated
2118 * key down event.
2119 *
2120 * Some IR protocols like NEC have a special IR message for
2121 * repeating last button, for when a button is held down.
2122 *
2123 * The *ctx* should point to the lirc sample as passed into
2124 * the program.
2125 *
2126 * This helper is only available is the kernel was compiled with
2127 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2128 * "**y**".
2129 * Return
2130 * 0
2131 *
2132 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2133 * Description
2134 * Return the cgroup v2 id of the socket associated with the *skb*.
2135 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2136 * helper for cgroup v1 by providing a tag resp. identifier that
2137 * can be matched on or used for map lookups e.g. to implement
2138 * policy. The cgroup v2 id of a given path in the hierarchy is
2139 * exposed in user space through the f_handle API in order to get
2140 * to the same 64-bit id.
2141 *
2142 * This helper can be used on TC egress path, but not on ingress,
2143 * and is available only if the kernel was compiled with the
2144 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2145 * Return
2146 * The id is returned or 0 in case the id could not be retrieved.
2147 *
2148 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2149 * Description
2150 * Return id of cgroup v2 that is ancestor of cgroup associated
2151 * with the *skb* at the *ancestor_level*. The root cgroup is at
2152 * *ancestor_level* zero and each step down the hierarchy
2153 * increments the level. If *ancestor_level* == level of cgroup
2154 * associated with *skb*, then return value will be same as that
2155 * of **bpf_skb_cgroup_id**\ ().
2156 *
2157 * The helper is useful to implement policies based on cgroups
2158 * that are upper in hierarchy than immediate cgroup associated
2159 * with *skb*.
2160 *
2161 * The format of returned id and helper limitations are same as in
2162 * **bpf_skb_cgroup_id**\ ().
2163 * Return
2164 * The id is returned or 0 in case the id could not be retrieved.
2165 *
2166 * u64 bpf_get_current_cgroup_id(void)
2167 * Return
2168 * A 64-bit integer containing the current cgroup id based
2169 * on the cgroup within which the current task is running.
2170 *
2171 * void* get_local_storage(void *map, u64 flags)
2172 * Description
2173 * Get the pointer to the local storage area.
2174 * The type and the size of the local storage is defined
2175 * by the *map* argument.
2176 * The *flags* meaning is specific for each map type,
2177 * and has to be 0 for cgroup local storage.
2178 *
2179 * Depending on the BPF program type, a local storage area
2180 * can be shared between multiple instances of the BPF program,
2181 * running simultaneously.
2182 *
2183 * A user should care about the synchronization by himself.
2184 * For example, by using the **BPF_STX_XADD** instruction to alter
2185 * the shared data.
2186 * Return
2187 * A pointer to the local storage area.
2188 *
2189 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2190 * Description
2191 * Select a **SO_REUSEPORT** socket from a
2192 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2193 * It checks the selected socket is matching the incoming
2194 * request in the socket buffer.
2195 * Return
2196 * 0 on success, or a negative error in case of failure.
2197 *
2198 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2199 * Description
2200 * Look for TCP socket matching *tuple*, optionally in a child
2201 * network namespace *netns*. The return value must be checked,
2202 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2203 *
2204 * The *ctx* should point to the context of the program, such as
2205 * the skb or socket (depending on the hook in use). This is used
2206 * to determine the base network namespace for the lookup.
2207 *
2208 * *tuple_size* must be one of:
2209 *
2210 * **sizeof**\ (*tuple*\ **->ipv4**)
2211 * Look for an IPv4 socket.
2212 * **sizeof**\ (*tuple*\ **->ipv6**)
2213 * Look for an IPv6 socket.
2214 *
2215 * If the *netns* is a negative signed 32-bit integer, then the
2216 * socket lookup table in the netns associated with the *ctx* will
2217 * will be used. For the TC hooks, this is the netns of the device
2218 * in the skb. For socket hooks, this is the netns of the socket.
2219 * If *netns* is any other signed 32-bit value greater than or
2220 * equal to zero then it specifies the ID of the netns relative to
2221 * the netns associated with the *ctx*. *netns* values beyond the
2222 * range of 32-bit integers are reserved for future use.
2223 *
2224 * All values for *flags* are reserved for future usage, and must
2225 * be left at zero.
2226 *
2227 * This helper is available only if the kernel was compiled with
2228 * **CONFIG_NET** configuration option.
2229 * Return
2230 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2231 * For sockets with reuseport option, the **struct bpf_sock**
2232 * result is from **reuse->socks**\ [] using the hash of the tuple.
2233 *
2234 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2235 * Description
2236 * Look for UDP socket matching *tuple*, optionally in a child
2237 * network namespace *netns*. The return value must be checked,
2238 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2239 *
2240 * The *ctx* should point to the context of the program, such as
2241 * the skb or socket (depending on the hook in use). This is used
2242 * to determine the base network namespace for the lookup.
2243 *
2244 * *tuple_size* must be one of:
2245 *
2246 * **sizeof**\ (*tuple*\ **->ipv4**)
2247 * Look for an IPv4 socket.
2248 * **sizeof**\ (*tuple*\ **->ipv6**)
2249 * Look for an IPv6 socket.
2250 *
2251 * If the *netns* is a negative signed 32-bit integer, then the
2252 * socket lookup table in the netns associated with the *ctx* will
2253 * will be used. For the TC hooks, this is the netns of the device
2254 * in the skb. For socket hooks, this is the netns of the socket.
2255 * If *netns* is any other signed 32-bit value greater than or
2256 * equal to zero then it specifies the ID of the netns relative to
2257 * the netns associated with the *ctx*. *netns* values beyond the
2258 * range of 32-bit integers are reserved for future use.
2259 *
2260 * All values for *flags* are reserved for future usage, and must
2261 * be left at zero.
2262 *
2263 * This helper is available only if the kernel was compiled with
2264 * **CONFIG_NET** configuration option.
2265 * Return
2266 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2267 * For sockets with reuseport option, the **struct bpf_sock**
2268 * result is from **reuse->socks**\ [] using the hash of the tuple.
2269 *
2270 * int bpf_sk_release(struct bpf_sock *sock)
2271 * Description
2272 * Release the reference held by *sock*. *sock* must be a
2273 * non-**NULL** pointer that was returned from
2274 * **bpf_sk_lookup_xxx**\ ().
2275 * Return
2276 * 0 on success, or a negative error in case of failure.
2277 *
2278 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2279 * Description
2280 * Pop an element from *map*.
2281 * Return
2282 * 0 on success, or a negative error in case of failure.
2283 *
2284 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2285 * Description
2286 * Get an element from *map* without removing it.
2287 * Return
2288 * 0 on success, or a negative error in case of failure.
2289 *
2290 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2291 * Description
2292 * For socket policies, insert *len* bytes into *msg* at offset
2293 * *start*.
2294 *
2295 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2296 * *msg* it may want to insert metadata or options into the *msg*.
2297 * This can later be read and used by any of the lower layer BPF
2298 * hooks.
2299 *
2300 * This helper may fail if under memory pressure (a malloc
2301 * fails) in these cases BPF programs will get an appropriate
2302 * error and BPF programs will need to handle them.
2303 * Return
2304 * 0 on success, or a negative error in case of failure.
2305 *
2306 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2307 * Description
2308 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2309 * This may result in **ENOMEM** errors under certain situations if
2310 * an allocation and copy are required due to a full ring buffer.
2311 * However, the helper will try to avoid doing the allocation
2312 * if possible. Other errors can occur if input parameters are
2313 * invalid either due to *start* byte not being valid part of *msg*
2314 * payload and/or *pop* value being to large.
2315 * Return
2316 * 0 on success, or a negative error in case of failure.
2317 *
2318 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2319 * Description
2320 * This helper is used in programs implementing IR decoding, to
2321 * report a successfully decoded pointer movement.
2322 *
2323 * The *ctx* should point to the lirc sample as passed into
2324 * the program.
2325 *
2326 * This helper is only available is the kernel was compiled with
2327 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2328 * "**y**".
2329 * Return
2330 * 0
2331 */
2332 #define __BPF_FUNC_MAPPER(FN) \
2333 FN(unspec), \
2334 FN(map_lookup_elem), \
2335 FN(map_update_elem), \
2336 FN(map_delete_elem), \
2337 FN(probe_read), \
2338 FN(ktime_get_ns), \
2339 FN(trace_printk), \
2340 FN(get_prandom_u32), \
2341 FN(get_smp_processor_id), \
2342 FN(skb_store_bytes), \
2343 FN(l3_csum_replace), \
2344 FN(l4_csum_replace), \
2345 FN(tail_call), \
2346 FN(clone_redirect), \
2347 FN(get_current_pid_tgid), \
2348 FN(get_current_uid_gid), \
2349 FN(get_current_comm), \
2350 FN(get_cgroup_classid), \
2351 FN(skb_vlan_push), \
2352 FN(skb_vlan_pop), \
2353 FN(skb_get_tunnel_key), \
2354 FN(skb_set_tunnel_key), \
2355 FN(perf_event_read), \
2356 FN(redirect), \
2357 FN(get_route_realm), \
2358 FN(perf_event_output), \
2359 FN(skb_load_bytes), \
2360 FN(get_stackid), \
2361 FN(csum_diff), \
2362 FN(skb_get_tunnel_opt), \
2363 FN(skb_set_tunnel_opt), \
2364 FN(skb_change_proto), \
2365 FN(skb_change_type), \
2366 FN(skb_under_cgroup), \
2367 FN(get_hash_recalc), \
2368 FN(get_current_task), \
2369 FN(probe_write_user), \
2370 FN(current_task_under_cgroup), \
2371 FN(skb_change_tail), \
2372 FN(skb_pull_data), \
2373 FN(csum_update), \
2374 FN(set_hash_invalid), \
2375 FN(get_numa_node_id), \
2376 FN(skb_change_head), \
2377 FN(xdp_adjust_head), \
2378 FN(probe_read_str), \
2379 FN(get_socket_cookie), \
2380 FN(get_socket_uid), \
2381 FN(set_hash), \
2382 FN(setsockopt), \
2383 FN(skb_adjust_room), \
2384 FN(redirect_map), \
2385 FN(sk_redirect_map), \
2386 FN(sock_map_update), \
2387 FN(xdp_adjust_meta), \
2388 FN(perf_event_read_value), \
2389 FN(perf_prog_read_value), \
2390 FN(getsockopt), \
2391 FN(override_return), \
2392 FN(sock_ops_cb_flags_set), \
2393 FN(msg_redirect_map), \
2394 FN(msg_apply_bytes), \
2395 FN(msg_cork_bytes), \
2396 FN(msg_pull_data), \
2397 FN(bind), \
2398 FN(xdp_adjust_tail), \
2399 FN(skb_get_xfrm_state), \
2400 FN(get_stack), \
2401 FN(skb_load_bytes_relative), \
2402 FN(fib_lookup), \
2403 FN(sock_hash_update), \
2404 FN(msg_redirect_hash), \
2405 FN(sk_redirect_hash), \
2406 FN(lwt_push_encap), \
2407 FN(lwt_seg6_store_bytes), \
2408 FN(lwt_seg6_adjust_srh), \
2409 FN(lwt_seg6_action), \
2410 FN(rc_repeat), \
2411 FN(rc_keydown), \
2412 FN(skb_cgroup_id), \
2413 FN(get_current_cgroup_id), \
2414 FN(get_local_storage), \
2415 FN(sk_select_reuseport), \
2416 FN(skb_ancestor_cgroup_id), \
2417 FN(sk_lookup_tcp), \
2418 FN(sk_lookup_udp), \
2419 FN(sk_release), \
2420 FN(map_push_elem), \
2421 FN(map_pop_elem), \
2422 FN(map_peek_elem), \
2423 FN(msg_push_data), \
2424 FN(msg_pop_data), \
2425 FN(rc_pointer_rel),
2426
2427 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2428 * function eBPF program intends to call
2429 */
2430 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2431 enum bpf_func_id {
2432 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2433 __BPF_FUNC_MAX_ID,
2434 };
2435 #undef __BPF_ENUM_FN
2436
2437 /* All flags used by eBPF helper functions, placed here. */
2438
2439 /* BPF_FUNC_skb_store_bytes flags. */
2440 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2441 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2442
2443 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2444 * First 4 bits are for passing the header field size.
2445 */
2446 #define BPF_F_HDR_FIELD_MASK 0xfULL
2447
2448 /* BPF_FUNC_l4_csum_replace flags. */
2449 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2450 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2451 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2452
2453 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2454 #define BPF_F_INGRESS (1ULL << 0)
2455
2456 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2457 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2458
2459 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2460 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2461 #define BPF_F_USER_STACK (1ULL << 8)
2462 /* flags used by BPF_FUNC_get_stackid only. */
2463 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2464 #define BPF_F_REUSE_STACKID (1ULL << 10)
2465 /* flags used by BPF_FUNC_get_stack only. */
2466 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2467
2468 /* BPF_FUNC_skb_set_tunnel_key flags. */
2469 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2470 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2471 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2472
2473 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2474 * BPF_FUNC_perf_event_read_value flags.
2475 */
2476 #define BPF_F_INDEX_MASK 0xffffffffULL
2477 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2478 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2479 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2480
2481 /* Current network namespace */
2482 #define BPF_F_CURRENT_NETNS (-1L)
2483
2484 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2485 enum bpf_adj_room_mode {
2486 BPF_ADJ_ROOM_NET,
2487 };
2488
2489 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2490 enum bpf_hdr_start_off {
2491 BPF_HDR_START_MAC,
2492 BPF_HDR_START_NET,
2493 };
2494
2495 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2496 enum bpf_lwt_encap_mode {
2497 BPF_LWT_ENCAP_SEG6,
2498 BPF_LWT_ENCAP_SEG6_INLINE
2499 };
2500
2501 #define __bpf_md_ptr(type, name) \
2502 union { \
2503 type name; \
2504 __u64 :64; \
2505 } __attribute__((aligned(8)))
2506
2507 /* user accessible mirror of in-kernel sk_buff.
2508 * new fields can only be added to the end of this structure
2509 */
2510 struct __sk_buff {
2511 __u32 len;
2512 __u32 pkt_type;
2513 __u32 mark;
2514 __u32 queue_mapping;
2515 __u32 protocol;
2516 __u32 vlan_present;
2517 __u32 vlan_tci;
2518 __u32 vlan_proto;
2519 __u32 priority;
2520 __u32 ingress_ifindex;
2521 __u32 ifindex;
2522 __u32 tc_index;
2523 __u32 cb[5];
2524 __u32 hash;
2525 __u32 tc_classid;
2526 __u32 data;
2527 __u32 data_end;
2528 __u32 napi_id;
2529
2530 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2531 __u32 family;
2532 __u32 remote_ip4; /* Stored in network byte order */
2533 __u32 local_ip4; /* Stored in network byte order */
2534 __u32 remote_ip6[4]; /* Stored in network byte order */
2535 __u32 local_ip6[4]; /* Stored in network byte order */
2536 __u32 remote_port; /* Stored in network byte order */
2537 __u32 local_port; /* stored in host byte order */
2538 /* ... here. */
2539
2540 __u32 data_meta;
2541 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
2542 __u64 tstamp;
2543 __u32 wire_len;
2544 __u32 gso_segs;
2545 };
2546
2547 struct bpf_tunnel_key {
2548 __u32 tunnel_id;
2549 union {
2550 __u32 remote_ipv4;
2551 __u32 remote_ipv6[4];
2552 };
2553 __u8 tunnel_tos;
2554 __u8 tunnel_ttl;
2555 __u16 tunnel_ext; /* Padding, future use. */
2556 __u32 tunnel_label;
2557 };
2558
2559 /* user accessible mirror of in-kernel xfrm_state.
2560 * new fields can only be added to the end of this structure
2561 */
2562 struct bpf_xfrm_state {
2563 __u32 reqid;
2564 __u32 spi; /* Stored in network byte order */
2565 __u16 family;
2566 __u16 ext; /* Padding, future use. */
2567 union {
2568 __u32 remote_ipv4; /* Stored in network byte order */
2569 __u32 remote_ipv6[4]; /* Stored in network byte order */
2570 };
2571 };
2572
2573 /* Generic BPF return codes which all BPF program types may support.
2574 * The values are binary compatible with their TC_ACT_* counter-part to
2575 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2576 * programs.
2577 *
2578 * XDP is handled seprately, see XDP_*.
2579 */
2580 enum bpf_ret_code {
2581 BPF_OK = 0,
2582 /* 1 reserved */
2583 BPF_DROP = 2,
2584 /* 3-6 reserved */
2585 BPF_REDIRECT = 7,
2586 /* >127 are reserved for prog type specific return codes */
2587 };
2588
2589 struct bpf_sock {
2590 __u32 bound_dev_if;
2591 __u32 family;
2592 __u32 type;
2593 __u32 protocol;
2594 __u32 mark;
2595 __u32 priority;
2596 __u32 src_ip4; /* Allows 1,2,4-byte read.
2597 * Stored in network byte order.
2598 */
2599 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2600 * Stored in network byte order.
2601 */
2602 __u32 src_port; /* Allows 4-byte read.
2603 * Stored in host byte order
2604 */
2605 };
2606
2607 struct bpf_sock_tuple {
2608 union {
2609 struct {
2610 __be32 saddr;
2611 __be32 daddr;
2612 __be16 sport;
2613 __be16 dport;
2614 } ipv4;
2615 struct {
2616 __be32 saddr[4];
2617 __be32 daddr[4];
2618 __be16 sport;
2619 __be16 dport;
2620 } ipv6;
2621 };
2622 };
2623
2624 #define XDP_PACKET_HEADROOM 256
2625
2626 /* User return codes for XDP prog type.
2627 * A valid XDP program must return one of these defined values. All other
2628 * return codes are reserved for future use. Unknown return codes will
2629 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2630 */
2631 enum xdp_action {
2632 XDP_ABORTED = 0,
2633 XDP_DROP,
2634 XDP_PASS,
2635 XDP_TX,
2636 XDP_REDIRECT,
2637 };
2638
2639 /* user accessible metadata for XDP packet hook
2640 * new fields must be added to the end of this structure
2641 */
2642 struct xdp_md {
2643 __u32 data;
2644 __u32 data_end;
2645 __u32 data_meta;
2646 /* Below access go through struct xdp_rxq_info */
2647 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2648 __u32 rx_queue_index; /* rxq->queue_index */
2649 };
2650
2651 enum sk_action {
2652 SK_DROP = 0,
2653 SK_PASS,
2654 };
2655
2656 /* user accessible metadata for SK_MSG packet hook, new fields must
2657 * be added to the end of this structure
2658 */
2659 struct sk_msg_md {
2660 __bpf_md_ptr(void *, data);
2661 __bpf_md_ptr(void *, data_end);
2662
2663 __u32 family;
2664 __u32 remote_ip4; /* Stored in network byte order */
2665 __u32 local_ip4; /* Stored in network byte order */
2666 __u32 remote_ip6[4]; /* Stored in network byte order */
2667 __u32 local_ip6[4]; /* Stored in network byte order */
2668 __u32 remote_port; /* Stored in network byte order */
2669 __u32 local_port; /* stored in host byte order */
2670 __u32 size; /* Total size of sk_msg */
2671 };
2672
2673 struct sk_reuseport_md {
2674 /*
2675 * Start of directly accessible data. It begins from
2676 * the tcp/udp header.
2677 */
2678 __bpf_md_ptr(void *, data);
2679 /* End of directly accessible data */
2680 __bpf_md_ptr(void *, data_end);
2681 /*
2682 * Total length of packet (starting from the tcp/udp header).
2683 * Note that the directly accessible bytes (data_end - data)
2684 * could be less than this "len". Those bytes could be
2685 * indirectly read by a helper "bpf_skb_load_bytes()".
2686 */
2687 __u32 len;
2688 /*
2689 * Eth protocol in the mac header (network byte order). e.g.
2690 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2691 */
2692 __u32 eth_protocol;
2693 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2694 __u32 bind_inany; /* Is sock bound to an INANY address? */
2695 __u32 hash; /* A hash of the packet 4 tuples */
2696 };
2697
2698 #define BPF_TAG_SIZE 8
2699
2700 struct bpf_prog_info {
2701 __u32 type;
2702 __u32 id;
2703 __u8 tag[BPF_TAG_SIZE];
2704 __u32 jited_prog_len;
2705 __u32 xlated_prog_len;
2706 __aligned_u64 jited_prog_insns;
2707 __aligned_u64 xlated_prog_insns;
2708 __u64 load_time; /* ns since boottime */
2709 __u32 created_by_uid;
2710 __u32 nr_map_ids;
2711 __aligned_u64 map_ids;
2712 char name[BPF_OBJ_NAME_LEN];
2713 __u32 ifindex;
2714 __u32 gpl_compatible:1;
2715 __u64 netns_dev;
2716 __u64 netns_ino;
2717 __u32 nr_jited_ksyms;
2718 __u32 nr_jited_func_lens;
2719 __aligned_u64 jited_ksyms;
2720 __aligned_u64 jited_func_lens;
2721 __u32 btf_id;
2722 __u32 func_info_rec_size;
2723 __aligned_u64 func_info;
2724 __u32 nr_func_info;
2725 __u32 nr_line_info;
2726 __aligned_u64 line_info;
2727 __aligned_u64 jited_line_info;
2728 __u32 nr_jited_line_info;
2729 __u32 line_info_rec_size;
2730 __u32 jited_line_info_rec_size;
2731 __u32 nr_prog_tags;
2732 __aligned_u64 prog_tags;
2733 } __attribute__((aligned(8)));
2734
2735 struct bpf_map_info {
2736 __u32 type;
2737 __u32 id;
2738 __u32 key_size;
2739 __u32 value_size;
2740 __u32 max_entries;
2741 __u32 map_flags;
2742 char name[BPF_OBJ_NAME_LEN];
2743 __u32 ifindex;
2744 __u32 :32;
2745 __u64 netns_dev;
2746 __u64 netns_ino;
2747 __u32 btf_id;
2748 __u32 btf_key_type_id;
2749 __u32 btf_value_type_id;
2750 } __attribute__((aligned(8)));
2751
2752 struct bpf_btf_info {
2753 __aligned_u64 btf;
2754 __u32 btf_size;
2755 __u32 id;
2756 } __attribute__((aligned(8)));
2757
2758 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2759 * by user and intended to be used by socket (e.g. to bind to, depends on
2760 * attach attach type).
2761 */
2762 struct bpf_sock_addr {
2763 __u32 user_family; /* Allows 4-byte read, but no write. */
2764 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2765 * Stored in network byte order.
2766 */
2767 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2768 * Stored in network byte order.
2769 */
2770 __u32 user_port; /* Allows 4-byte read and write.
2771 * Stored in network byte order
2772 */
2773 __u32 family; /* Allows 4-byte read, but no write */
2774 __u32 type; /* Allows 4-byte read, but no write */
2775 __u32 protocol; /* Allows 4-byte read, but no write */
2776 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2777 * Stored in network byte order.
2778 */
2779 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2780 * Stored in network byte order.
2781 */
2782 };
2783
2784 /* User bpf_sock_ops struct to access socket values and specify request ops
2785 * and their replies.
2786 * Some of this fields are in network (bigendian) byte order and may need
2787 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2788 * New fields can only be added at the end of this structure
2789 */
2790 struct bpf_sock_ops {
2791 __u32 op;
2792 union {
2793 __u32 args[4]; /* Optionally passed to bpf program */
2794 __u32 reply; /* Returned by bpf program */
2795 __u32 replylong[4]; /* Optionally returned by bpf prog */
2796 };
2797 __u32 family;
2798 __u32 remote_ip4; /* Stored in network byte order */
2799 __u32 local_ip4; /* Stored in network byte order */
2800 __u32 remote_ip6[4]; /* Stored in network byte order */
2801 __u32 local_ip6[4]; /* Stored in network byte order */
2802 __u32 remote_port; /* Stored in network byte order */
2803 __u32 local_port; /* stored in host byte order */
2804 __u32 is_fullsock; /* Some TCP fields are only valid if
2805 * there is a full socket. If not, the
2806 * fields read as zero.
2807 */
2808 __u32 snd_cwnd;
2809 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2810 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2811 __u32 state;
2812 __u32 rtt_min;
2813 __u32 snd_ssthresh;
2814 __u32 rcv_nxt;
2815 __u32 snd_nxt;
2816 __u32 snd_una;
2817 __u32 mss_cache;
2818 __u32 ecn_flags;
2819 __u32 rate_delivered;
2820 __u32 rate_interval_us;
2821 __u32 packets_out;
2822 __u32 retrans_out;
2823 __u32 total_retrans;
2824 __u32 segs_in;
2825 __u32 data_segs_in;
2826 __u32 segs_out;
2827 __u32 data_segs_out;
2828 __u32 lost_out;
2829 __u32 sacked_out;
2830 __u32 sk_txhash;
2831 __u64 bytes_received;
2832 __u64 bytes_acked;
2833 };
2834
2835 /* Definitions for bpf_sock_ops_cb_flags */
2836 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2837 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2838 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2839 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2840 * supported cb flags
2841 */
2842
2843 /* List of known BPF sock_ops operators.
2844 * New entries can only be added at the end
2845 */
2846 enum {
2847 BPF_SOCK_OPS_VOID,
2848 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2849 * -1 if default value should be used
2850 */
2851 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2852 * window (in packets) or -1 if default
2853 * value should be used
2854 */
2855 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2856 * active connection is initialized
2857 */
2858 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2859 * active connection is
2860 * established
2861 */
2862 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2863 * passive connection is
2864 * established
2865 */
2866 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2867 * needs ECN
2868 */
2869 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2870 * based on the path and may be
2871 * dependent on the congestion control
2872 * algorithm. In general it indicates
2873 * a congestion threshold. RTTs above
2874 * this indicate congestion
2875 */
2876 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2877 * Arg1: value of icsk_retransmits
2878 * Arg2: value of icsk_rto
2879 * Arg3: whether RTO has expired
2880 */
2881 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2882 * Arg1: sequence number of 1st byte
2883 * Arg2: # segments
2884 * Arg3: return value of
2885 * tcp_transmit_skb (0 => success)
2886 */
2887 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2888 * Arg1: old_state
2889 * Arg2: new_state
2890 */
2891 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
2892 * socket transition to LISTEN state.
2893 */
2894 };
2895
2896 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2897 * changes between the TCP and BPF versions. Ideally this should never happen.
2898 * If it does, we need to add code to convert them before calling
2899 * the BPF sock_ops function.
2900 */
2901 enum {
2902 BPF_TCP_ESTABLISHED = 1,
2903 BPF_TCP_SYN_SENT,
2904 BPF_TCP_SYN_RECV,
2905 BPF_TCP_FIN_WAIT1,
2906 BPF_TCP_FIN_WAIT2,
2907 BPF_TCP_TIME_WAIT,
2908 BPF_TCP_CLOSE,
2909 BPF_TCP_CLOSE_WAIT,
2910 BPF_TCP_LAST_ACK,
2911 BPF_TCP_LISTEN,
2912 BPF_TCP_CLOSING, /* Now a valid state */
2913 BPF_TCP_NEW_SYN_RECV,
2914
2915 BPF_TCP_MAX_STATES /* Leave at the end! */
2916 };
2917
2918 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
2919 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
2920
2921 struct bpf_perf_event_value {
2922 __u64 counter;
2923 __u64 enabled;
2924 __u64 running;
2925 };
2926
2927 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2928 #define BPF_DEVCG_ACC_READ (1ULL << 1)
2929 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2930
2931 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2932 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2933
2934 struct bpf_cgroup_dev_ctx {
2935 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2936 __u32 access_type;
2937 __u32 major;
2938 __u32 minor;
2939 };
2940
2941 struct bpf_raw_tracepoint_args {
2942 __u64 args[0];
2943 };
2944
2945 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
2946 * OUTPUT: Do lookup from egress perspective; default is ingress
2947 */
2948 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
2949 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
2950
2951 enum {
2952 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
2953 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
2954 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
2955 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
2956 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
2957 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
2958 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
2959 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
2960 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
2961 };
2962
2963 struct bpf_fib_lookup {
2964 /* input: network family for lookup (AF_INET, AF_INET6)
2965 * output: network family of egress nexthop
2966 */
2967 __u8 family;
2968
2969 /* set if lookup is to consider L4 data - e.g., FIB rules */
2970 __u8 l4_protocol;
2971 __be16 sport;
2972 __be16 dport;
2973
2974 /* total length of packet from network header - used for MTU check */
2975 __u16 tot_len;
2976
2977 /* input: L3 device index for lookup
2978 * output: device index from FIB lookup
2979 */
2980 __u32 ifindex;
2981
2982 union {
2983 /* inputs to lookup */
2984 __u8 tos; /* AF_INET */
2985 __be32 flowinfo; /* AF_INET6, flow_label + priority */
2986
2987 /* output: metric of fib result (IPv4/IPv6 only) */
2988 __u32 rt_metric;
2989 };
2990
2991 union {
2992 __be32 ipv4_src;
2993 __u32 ipv6_src[4]; /* in6_addr; network order */
2994 };
2995
2996 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2997 * network header. output: bpf_fib_lookup sets to gateway address
2998 * if FIB lookup returns gateway route
2999 */
3000 union {
3001 __be32 ipv4_dst;
3002 __u32 ipv6_dst[4]; /* in6_addr; network order */
3003 };
3004
3005 /* output */
3006 __be16 h_vlan_proto;
3007 __be16 h_vlan_TCI;
3008 __u8 smac[6]; /* ETH_ALEN */
3009 __u8 dmac[6]; /* ETH_ALEN */
3010 };
3011
3012 enum bpf_task_fd_type {
3013 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3014 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3015 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3016 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3017 BPF_FD_TYPE_UPROBE, /* filename + offset */
3018 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3019 };
3020
3021 struct bpf_flow_keys {
3022 __u16 nhoff;
3023 __u16 thoff;
3024 __u16 addr_proto; /* ETH_P_* of valid addrs */
3025 __u8 is_frag;
3026 __u8 is_first_frag;
3027 __u8 is_encap;
3028 __u8 ip_proto;
3029 __be16 n_proto;
3030 __be16 sport;
3031 __be16 dport;
3032 union {
3033 struct {
3034 __be32 ipv4_src;
3035 __be32 ipv4_dst;
3036 };
3037 struct {
3038 __u32 ipv6_src[4]; /* in6_addr; network order */
3039 __u32 ipv6_dst[4]; /* in6_addr; network order */
3040 };
3041 };
3042 };
3043
3044 struct bpf_func_info {
3045 __u32 insn_off;
3046 __u32 type_id;
3047 };
3048
3049 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3050 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3051
3052 struct bpf_line_info {
3053 __u32 insn_off;
3054 __u32 file_name_off;
3055 __u32 line_off;
3056 __u32 line_col;
3057 };
3058
3059 #endif /* __LINUX_BPF_H__ */