]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/blob - include/uapi/linux/bpf.h
f9187b41dff62cf81c03606fe9eba6f9f19d7a1e
[mirror_ubuntu-eoan-kernel.git] / include / uapi / linux / bpf.h
1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 */
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
10
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
13
14 /* Extended instruction set based on top of classic BPF */
15
16 /* instruction classes */
17 #define BPF_ALU64 0x07 /* alu mode in double word width */
18
19 /* ld/ldx fields */
20 #define BPF_DW 0x18 /* double word (64-bit) */
21 #define BPF_XADD 0xc0 /* exclusive add */
22
23 /* alu/jmp fields */
24 #define BPF_MOV 0xb0 /* mov reg to reg */
25 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
26
27 /* change endianness of a register */
28 #define BPF_END 0xd0 /* flags for endianness conversion: */
29 #define BPF_TO_LE 0x00 /* convert to little-endian */
30 #define BPF_TO_BE 0x08 /* convert to big-endian */
31 #define BPF_FROM_LE BPF_TO_LE
32 #define BPF_FROM_BE BPF_TO_BE
33
34 /* jmp encodings */
35 #define BPF_JNE 0x50 /* jump != */
36 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
37 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
38 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
39 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
40 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
41 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
42 #define BPF_CALL 0x80 /* function call */
43 #define BPF_EXIT 0x90 /* function return */
44
45 /* Register numbers */
46 enum {
47 BPF_REG_0 = 0,
48 BPF_REG_1,
49 BPF_REG_2,
50 BPF_REG_3,
51 BPF_REG_4,
52 BPF_REG_5,
53 BPF_REG_6,
54 BPF_REG_7,
55 BPF_REG_8,
56 BPF_REG_9,
57 BPF_REG_10,
58 __MAX_BPF_REG,
59 };
60
61 /* BPF has 10 general purpose 64-bit registers and stack frame. */
62 #define MAX_BPF_REG __MAX_BPF_REG
63
64 struct bpf_insn {
65 __u8 code; /* opcode */
66 __u8 dst_reg:4; /* dest register */
67 __u8 src_reg:4; /* source register */
68 __s16 off; /* signed offset */
69 __s32 imm; /* signed immediate constant */
70 };
71
72 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
73 struct bpf_lpm_trie_key {
74 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
75 __u8 data[0]; /* Arbitrary size */
76 };
77
78 struct bpf_cgroup_storage_key {
79 __u64 cgroup_inode_id; /* cgroup inode id */
80 __u32 attach_type; /* program attach type */
81 };
82
83 /* BPF syscall commands, see bpf(2) man-page for details. */
84 enum bpf_cmd {
85 BPF_MAP_CREATE,
86 BPF_MAP_LOOKUP_ELEM,
87 BPF_MAP_UPDATE_ELEM,
88 BPF_MAP_DELETE_ELEM,
89 BPF_MAP_GET_NEXT_KEY,
90 BPF_PROG_LOAD,
91 BPF_OBJ_PIN,
92 BPF_OBJ_GET,
93 BPF_PROG_ATTACH,
94 BPF_PROG_DETACH,
95 BPF_PROG_TEST_RUN,
96 BPF_PROG_GET_NEXT_ID,
97 BPF_MAP_GET_NEXT_ID,
98 BPF_PROG_GET_FD_BY_ID,
99 BPF_MAP_GET_FD_BY_ID,
100 BPF_OBJ_GET_INFO_BY_FD,
101 BPF_PROG_QUERY,
102 BPF_RAW_TRACEPOINT_OPEN,
103 BPF_BTF_LOAD,
104 BPF_BTF_GET_FD_BY_ID,
105 BPF_TASK_FD_QUERY,
106 };
107
108 enum bpf_map_type {
109 BPF_MAP_TYPE_UNSPEC,
110 BPF_MAP_TYPE_HASH,
111 BPF_MAP_TYPE_ARRAY,
112 BPF_MAP_TYPE_PROG_ARRAY,
113 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
114 BPF_MAP_TYPE_PERCPU_HASH,
115 BPF_MAP_TYPE_PERCPU_ARRAY,
116 BPF_MAP_TYPE_STACK_TRACE,
117 BPF_MAP_TYPE_CGROUP_ARRAY,
118 BPF_MAP_TYPE_LRU_HASH,
119 BPF_MAP_TYPE_LRU_PERCPU_HASH,
120 BPF_MAP_TYPE_LPM_TRIE,
121 BPF_MAP_TYPE_ARRAY_OF_MAPS,
122 BPF_MAP_TYPE_HASH_OF_MAPS,
123 BPF_MAP_TYPE_DEVMAP,
124 BPF_MAP_TYPE_SOCKMAP,
125 BPF_MAP_TYPE_CPUMAP,
126 BPF_MAP_TYPE_XSKMAP,
127 BPF_MAP_TYPE_SOCKHASH,
128 BPF_MAP_TYPE_CGROUP_STORAGE,
129 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
130 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
131 };
132
133 enum bpf_prog_type {
134 BPF_PROG_TYPE_UNSPEC,
135 BPF_PROG_TYPE_SOCKET_FILTER,
136 BPF_PROG_TYPE_KPROBE,
137 BPF_PROG_TYPE_SCHED_CLS,
138 BPF_PROG_TYPE_SCHED_ACT,
139 BPF_PROG_TYPE_TRACEPOINT,
140 BPF_PROG_TYPE_XDP,
141 BPF_PROG_TYPE_PERF_EVENT,
142 BPF_PROG_TYPE_CGROUP_SKB,
143 BPF_PROG_TYPE_CGROUP_SOCK,
144 BPF_PROG_TYPE_LWT_IN,
145 BPF_PROG_TYPE_LWT_OUT,
146 BPF_PROG_TYPE_LWT_XMIT,
147 BPF_PROG_TYPE_SOCK_OPS,
148 BPF_PROG_TYPE_SK_SKB,
149 BPF_PROG_TYPE_CGROUP_DEVICE,
150 BPF_PROG_TYPE_SK_MSG,
151 BPF_PROG_TYPE_RAW_TRACEPOINT,
152 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
153 BPF_PROG_TYPE_LWT_SEG6LOCAL,
154 BPF_PROG_TYPE_LIRC_MODE2,
155 BPF_PROG_TYPE_SK_REUSEPORT,
156 BPF_PROG_TYPE_FLOW_DISSECTOR,
157 };
158
159 enum bpf_attach_type {
160 BPF_CGROUP_INET_INGRESS,
161 BPF_CGROUP_INET_EGRESS,
162 BPF_CGROUP_INET_SOCK_CREATE,
163 BPF_CGROUP_SOCK_OPS,
164 BPF_SK_SKB_STREAM_PARSER,
165 BPF_SK_SKB_STREAM_VERDICT,
166 BPF_CGROUP_DEVICE,
167 BPF_SK_MSG_VERDICT,
168 BPF_CGROUP_INET4_BIND,
169 BPF_CGROUP_INET6_BIND,
170 BPF_CGROUP_INET4_CONNECT,
171 BPF_CGROUP_INET6_CONNECT,
172 BPF_CGROUP_INET4_POST_BIND,
173 BPF_CGROUP_INET6_POST_BIND,
174 BPF_CGROUP_UDP4_SENDMSG,
175 BPF_CGROUP_UDP6_SENDMSG,
176 BPF_LIRC_MODE2,
177 BPF_FLOW_DISSECTOR,
178 __MAX_BPF_ATTACH_TYPE
179 };
180
181 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
182
183 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
184 *
185 * NONE(default): No further bpf programs allowed in the subtree.
186 *
187 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
188 * the program in this cgroup yields to sub-cgroup program.
189 *
190 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
191 * that cgroup program gets run in addition to the program in this cgroup.
192 *
193 * Only one program is allowed to be attached to a cgroup with
194 * NONE or BPF_F_ALLOW_OVERRIDE flag.
195 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
196 * release old program and attach the new one. Attach flags has to match.
197 *
198 * Multiple programs are allowed to be attached to a cgroup with
199 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
200 * (those that were attached first, run first)
201 * The programs of sub-cgroup are executed first, then programs of
202 * this cgroup and then programs of parent cgroup.
203 * When children program makes decision (like picking TCP CA or sock bind)
204 * parent program has a chance to override it.
205 *
206 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
207 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
208 * Ex1:
209 * cgrp1 (MULTI progs A, B) ->
210 * cgrp2 (OVERRIDE prog C) ->
211 * cgrp3 (MULTI prog D) ->
212 * cgrp4 (OVERRIDE prog E) ->
213 * cgrp5 (NONE prog F)
214 * the event in cgrp5 triggers execution of F,D,A,B in that order.
215 * if prog F is detached, the execution is E,D,A,B
216 * if prog F and D are detached, the execution is E,A,B
217 * if prog F, E and D are detached, the execution is C,A,B
218 *
219 * All eligible programs are executed regardless of return code from
220 * earlier programs.
221 */
222 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
223 #define BPF_F_ALLOW_MULTI (1U << 1)
224
225 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
226 * verifier will perform strict alignment checking as if the kernel
227 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
228 * and NET_IP_ALIGN defined to 2.
229 */
230 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
231
232 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
233 #define BPF_PSEUDO_MAP_FD 1
234
235 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
236 * offset to another bpf function
237 */
238 #define BPF_PSEUDO_CALL 1
239
240 /* flags for BPF_MAP_UPDATE_ELEM command */
241 #define BPF_ANY 0 /* create new element or update existing */
242 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
243 #define BPF_EXIST 2 /* update existing element */
244
245 /* flags for BPF_MAP_CREATE command */
246 #define BPF_F_NO_PREALLOC (1U << 0)
247 /* Instead of having one common LRU list in the
248 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
249 * which can scale and perform better.
250 * Note, the LRU nodes (including free nodes) cannot be moved
251 * across different LRU lists.
252 */
253 #define BPF_F_NO_COMMON_LRU (1U << 1)
254 /* Specify numa node during map creation */
255 #define BPF_F_NUMA_NODE (1U << 2)
256
257 /* flags for BPF_PROG_QUERY */
258 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
259
260 #define BPF_OBJ_NAME_LEN 16U
261
262 /* Flags for accessing BPF object */
263 #define BPF_F_RDONLY (1U << 3)
264 #define BPF_F_WRONLY (1U << 4)
265
266 /* Flag for stack_map, store build_id+offset instead of pointer */
267 #define BPF_F_STACK_BUILD_ID (1U << 5)
268
269 enum bpf_stack_build_id_status {
270 /* user space need an empty entry to identify end of a trace */
271 BPF_STACK_BUILD_ID_EMPTY = 0,
272 /* with valid build_id and offset */
273 BPF_STACK_BUILD_ID_VALID = 1,
274 /* couldn't get build_id, fallback to ip */
275 BPF_STACK_BUILD_ID_IP = 2,
276 };
277
278 #define BPF_BUILD_ID_SIZE 20
279 struct bpf_stack_build_id {
280 __s32 status;
281 unsigned char build_id[BPF_BUILD_ID_SIZE];
282 union {
283 __u64 offset;
284 __u64 ip;
285 };
286 };
287
288 union bpf_attr {
289 struct { /* anonymous struct used by BPF_MAP_CREATE command */
290 __u32 map_type; /* one of enum bpf_map_type */
291 __u32 key_size; /* size of key in bytes */
292 __u32 value_size; /* size of value in bytes */
293 __u32 max_entries; /* max number of entries in a map */
294 __u32 map_flags; /* BPF_MAP_CREATE related
295 * flags defined above.
296 */
297 __u32 inner_map_fd; /* fd pointing to the inner map */
298 __u32 numa_node; /* numa node (effective only if
299 * BPF_F_NUMA_NODE is set).
300 */
301 char map_name[BPF_OBJ_NAME_LEN];
302 __u32 map_ifindex; /* ifindex of netdev to create on */
303 __u32 btf_fd; /* fd pointing to a BTF type data */
304 __u32 btf_key_type_id; /* BTF type_id of the key */
305 __u32 btf_value_type_id; /* BTF type_id of the value */
306 };
307
308 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
309 __u32 map_fd;
310 __aligned_u64 key;
311 union {
312 __aligned_u64 value;
313 __aligned_u64 next_key;
314 };
315 __u64 flags;
316 };
317
318 struct { /* anonymous struct used by BPF_PROG_LOAD command */
319 __u32 prog_type; /* one of enum bpf_prog_type */
320 __u32 insn_cnt;
321 __aligned_u64 insns;
322 __aligned_u64 license;
323 __u32 log_level; /* verbosity level of verifier */
324 __u32 log_size; /* size of user buffer */
325 __aligned_u64 log_buf; /* user supplied buffer */
326 __u32 kern_version; /* checked when prog_type=kprobe */
327 __u32 prog_flags;
328 char prog_name[BPF_OBJ_NAME_LEN];
329 __u32 prog_ifindex; /* ifindex of netdev to prep for */
330 /* For some prog types expected attach type must be known at
331 * load time to verify attach type specific parts of prog
332 * (context accesses, allowed helpers, etc).
333 */
334 __u32 expected_attach_type;
335 };
336
337 struct { /* anonymous struct used by BPF_OBJ_* commands */
338 __aligned_u64 pathname;
339 __u32 bpf_fd;
340 __u32 file_flags;
341 };
342
343 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
344 __u32 target_fd; /* container object to attach to */
345 __u32 attach_bpf_fd; /* eBPF program to attach */
346 __u32 attach_type;
347 __u32 attach_flags;
348 };
349
350 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
351 __u32 prog_fd;
352 __u32 retval;
353 __u32 data_size_in;
354 __u32 data_size_out;
355 __aligned_u64 data_in;
356 __aligned_u64 data_out;
357 __u32 repeat;
358 __u32 duration;
359 } test;
360
361 struct { /* anonymous struct used by BPF_*_GET_*_ID */
362 union {
363 __u32 start_id;
364 __u32 prog_id;
365 __u32 map_id;
366 __u32 btf_id;
367 };
368 __u32 next_id;
369 __u32 open_flags;
370 };
371
372 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
373 __u32 bpf_fd;
374 __u32 info_len;
375 __aligned_u64 info;
376 } info;
377
378 struct { /* anonymous struct used by BPF_PROG_QUERY command */
379 __u32 target_fd; /* container object to query */
380 __u32 attach_type;
381 __u32 query_flags;
382 __u32 attach_flags;
383 __aligned_u64 prog_ids;
384 __u32 prog_cnt;
385 } query;
386
387 struct {
388 __u64 name;
389 __u32 prog_fd;
390 } raw_tracepoint;
391
392 struct { /* anonymous struct for BPF_BTF_LOAD */
393 __aligned_u64 btf;
394 __aligned_u64 btf_log_buf;
395 __u32 btf_size;
396 __u32 btf_log_size;
397 __u32 btf_log_level;
398 };
399
400 struct {
401 __u32 pid; /* input: pid */
402 __u32 fd; /* input: fd */
403 __u32 flags; /* input: flags */
404 __u32 buf_len; /* input/output: buf len */
405 __aligned_u64 buf; /* input/output:
406 * tp_name for tracepoint
407 * symbol for kprobe
408 * filename for uprobe
409 */
410 __u32 prog_id; /* output: prod_id */
411 __u32 fd_type; /* output: BPF_FD_TYPE_* */
412 __u64 probe_offset; /* output: probe_offset */
413 __u64 probe_addr; /* output: probe_addr */
414 } task_fd_query;
415 } __attribute__((aligned(8)));
416
417 /* The description below is an attempt at providing documentation to eBPF
418 * developers about the multiple available eBPF helper functions. It can be
419 * parsed and used to produce a manual page. The workflow is the following,
420 * and requires the rst2man utility:
421 *
422 * $ ./scripts/bpf_helpers_doc.py \
423 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
424 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
425 * $ man /tmp/bpf-helpers.7
426 *
427 * Note that in order to produce this external documentation, some RST
428 * formatting is used in the descriptions to get "bold" and "italics" in
429 * manual pages. Also note that the few trailing white spaces are
430 * intentional, removing them would break paragraphs for rst2man.
431 *
432 * Start of BPF helper function descriptions:
433 *
434 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
435 * Description
436 * Perform a lookup in *map* for an entry associated to *key*.
437 * Return
438 * Map value associated to *key*, or **NULL** if no entry was
439 * found.
440 *
441 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
442 * Description
443 * Add or update the value of the entry associated to *key* in
444 * *map* with *value*. *flags* is one of:
445 *
446 * **BPF_NOEXIST**
447 * The entry for *key* must not exist in the map.
448 * **BPF_EXIST**
449 * The entry for *key* must already exist in the map.
450 * **BPF_ANY**
451 * No condition on the existence of the entry for *key*.
452 *
453 * Flag value **BPF_NOEXIST** cannot be used for maps of types
454 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
455 * elements always exist), the helper would return an error.
456 * Return
457 * 0 on success, or a negative error in case of failure.
458 *
459 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
460 * Description
461 * Delete entry with *key* from *map*.
462 * Return
463 * 0 on success, or a negative error in case of failure.
464 *
465 * int bpf_probe_read(void *dst, u32 size, const void *src)
466 * Description
467 * For tracing programs, safely attempt to read *size* bytes from
468 * address *src* and store the data in *dst*.
469 * Return
470 * 0 on success, or a negative error in case of failure.
471 *
472 * u64 bpf_ktime_get_ns(void)
473 * Description
474 * Return the time elapsed since system boot, in nanoseconds.
475 * Return
476 * Current *ktime*.
477 *
478 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
479 * Description
480 * This helper is a "printk()-like" facility for debugging. It
481 * prints a message defined by format *fmt* (of size *fmt_size*)
482 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
483 * available. It can take up to three additional **u64**
484 * arguments (as an eBPF helpers, the total number of arguments is
485 * limited to five).
486 *
487 * Each time the helper is called, it appends a line to the trace.
488 * The format of the trace is customizable, and the exact output
489 * one will get depends on the options set in
490 * *\/sys/kernel/debug/tracing/trace_options* (see also the
491 * *README* file under the same directory). However, it usually
492 * defaults to something like:
493 *
494 * ::
495 *
496 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
497 *
498 * In the above:
499 *
500 * * ``telnet`` is the name of the current task.
501 * * ``470`` is the PID of the current task.
502 * * ``001`` is the CPU number on which the task is
503 * running.
504 * * In ``.N..``, each character refers to a set of
505 * options (whether irqs are enabled, scheduling
506 * options, whether hard/softirqs are running, level of
507 * preempt_disabled respectively). **N** means that
508 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
509 * are set.
510 * * ``419421.045894`` is a timestamp.
511 * * ``0x00000001`` is a fake value used by BPF for the
512 * instruction pointer register.
513 * * ``<formatted msg>`` is the message formatted with
514 * *fmt*.
515 *
516 * The conversion specifiers supported by *fmt* are similar, but
517 * more limited than for printk(). They are **%d**, **%i**,
518 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
519 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
520 * of field, padding with zeroes, etc.) is available, and the
521 * helper will return **-EINVAL** (but print nothing) if it
522 * encounters an unknown specifier.
523 *
524 * Also, note that **bpf_trace_printk**\ () is slow, and should
525 * only be used for debugging purposes. For this reason, a notice
526 * bloc (spanning several lines) is printed to kernel logs and
527 * states that the helper should not be used "for production use"
528 * the first time this helper is used (or more precisely, when
529 * **trace_printk**\ () buffers are allocated). For passing values
530 * to user space, perf events should be preferred.
531 * Return
532 * The number of bytes written to the buffer, or a negative error
533 * in case of failure.
534 *
535 * u32 bpf_get_prandom_u32(void)
536 * Description
537 * Get a pseudo-random number.
538 *
539 * From a security point of view, this helper uses its own
540 * pseudo-random internal state, and cannot be used to infer the
541 * seed of other random functions in the kernel. However, it is
542 * essential to note that the generator used by the helper is not
543 * cryptographically secure.
544 * Return
545 * A random 32-bit unsigned value.
546 *
547 * u32 bpf_get_smp_processor_id(void)
548 * Description
549 * Get the SMP (symmetric multiprocessing) processor id. Note that
550 * all programs run with preemption disabled, which means that the
551 * SMP processor id is stable during all the execution of the
552 * program.
553 * Return
554 * The SMP id of the processor running the program.
555 *
556 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
557 * Description
558 * Store *len* bytes from address *from* into the packet
559 * associated to *skb*, at *offset*. *flags* are a combination of
560 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
561 * checksum for the packet after storing the bytes) and
562 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
563 * **->swhash** and *skb*\ **->l4hash** to 0).
564 *
565 * A call to this helper is susceptible to change the underlaying
566 * packet buffer. Therefore, at load time, all checks on pointers
567 * previously done by the verifier are invalidated and must be
568 * performed again, if the helper is used in combination with
569 * direct packet access.
570 * Return
571 * 0 on success, or a negative error in case of failure.
572 *
573 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
574 * Description
575 * Recompute the layer 3 (e.g. IP) checksum for the packet
576 * associated to *skb*. Computation is incremental, so the helper
577 * must know the former value of the header field that was
578 * modified (*from*), the new value of this field (*to*), and the
579 * number of bytes (2 or 4) for this field, stored in *size*.
580 * Alternatively, it is possible to store the difference between
581 * the previous and the new values of the header field in *to*, by
582 * setting *from* and *size* to 0. For both methods, *offset*
583 * indicates the location of the IP checksum within the packet.
584 *
585 * This helper works in combination with **bpf_csum_diff**\ (),
586 * which does not update the checksum in-place, but offers more
587 * flexibility and can handle sizes larger than 2 or 4 for the
588 * checksum to update.
589 *
590 * A call to this helper is susceptible to change the underlaying
591 * packet buffer. Therefore, at load time, all checks on pointers
592 * previously done by the verifier are invalidated and must be
593 * performed again, if the helper is used in combination with
594 * direct packet access.
595 * Return
596 * 0 on success, or a negative error in case of failure.
597 *
598 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
599 * Description
600 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
601 * packet associated to *skb*. Computation is incremental, so the
602 * helper must know the former value of the header field that was
603 * modified (*from*), the new value of this field (*to*), and the
604 * number of bytes (2 or 4) for this field, stored on the lowest
605 * four bits of *flags*. Alternatively, it is possible to store
606 * the difference between the previous and the new values of the
607 * header field in *to*, by setting *from* and the four lowest
608 * bits of *flags* to 0. For both methods, *offset* indicates the
609 * location of the IP checksum within the packet. In addition to
610 * the size of the field, *flags* can be added (bitwise OR) actual
611 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
612 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
613 * for updates resulting in a null checksum the value is set to
614 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
615 * the checksum is to be computed against a pseudo-header.
616 *
617 * This helper works in combination with **bpf_csum_diff**\ (),
618 * which does not update the checksum in-place, but offers more
619 * flexibility and can handle sizes larger than 2 or 4 for the
620 * checksum to update.
621 *
622 * A call to this helper is susceptible to change the underlaying
623 * packet buffer. Therefore, at load time, all checks on pointers
624 * previously done by the verifier are invalidated and must be
625 * performed again, if the helper is used in combination with
626 * direct packet access.
627 * Return
628 * 0 on success, or a negative error in case of failure.
629 *
630 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
631 * Description
632 * This special helper is used to trigger a "tail call", or in
633 * other words, to jump into another eBPF program. The same stack
634 * frame is used (but values on stack and in registers for the
635 * caller are not accessible to the callee). This mechanism allows
636 * for program chaining, either for raising the maximum number of
637 * available eBPF instructions, or to execute given programs in
638 * conditional blocks. For security reasons, there is an upper
639 * limit to the number of successive tail calls that can be
640 * performed.
641 *
642 * Upon call of this helper, the program attempts to jump into a
643 * program referenced at index *index* in *prog_array_map*, a
644 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
645 * *ctx*, a pointer to the context.
646 *
647 * If the call succeeds, the kernel immediately runs the first
648 * instruction of the new program. This is not a function call,
649 * and it never returns to the previous program. If the call
650 * fails, then the helper has no effect, and the caller continues
651 * to run its subsequent instructions. A call can fail if the
652 * destination program for the jump does not exist (i.e. *index*
653 * is superior to the number of entries in *prog_array_map*), or
654 * if the maximum number of tail calls has been reached for this
655 * chain of programs. This limit is defined in the kernel by the
656 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
657 * which is currently set to 32.
658 * Return
659 * 0 on success, or a negative error in case of failure.
660 *
661 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
662 * Description
663 * Clone and redirect the packet associated to *skb* to another
664 * net device of index *ifindex*. Both ingress and egress
665 * interfaces can be used for redirection. The **BPF_F_INGRESS**
666 * value in *flags* is used to make the distinction (ingress path
667 * is selected if the flag is present, egress path otherwise).
668 * This is the only flag supported for now.
669 *
670 * In comparison with **bpf_redirect**\ () helper,
671 * **bpf_clone_redirect**\ () has the associated cost of
672 * duplicating the packet buffer, but this can be executed out of
673 * the eBPF program. Conversely, **bpf_redirect**\ () is more
674 * efficient, but it is handled through an action code where the
675 * redirection happens only after the eBPF program has returned.
676 *
677 * A call to this helper is susceptible to change the underlaying
678 * packet buffer. Therefore, at load time, all checks on pointers
679 * previously done by the verifier are invalidated and must be
680 * performed again, if the helper is used in combination with
681 * direct packet access.
682 * Return
683 * 0 on success, or a negative error in case of failure.
684 *
685 * u64 bpf_get_current_pid_tgid(void)
686 * Return
687 * A 64-bit integer containing the current tgid and pid, and
688 * created as such:
689 * *current_task*\ **->tgid << 32 \|**
690 * *current_task*\ **->pid**.
691 *
692 * u64 bpf_get_current_uid_gid(void)
693 * Return
694 * A 64-bit integer containing the current GID and UID, and
695 * created as such: *current_gid* **<< 32 \|** *current_uid*.
696 *
697 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
698 * Description
699 * Copy the **comm** attribute of the current task into *buf* of
700 * *size_of_buf*. The **comm** attribute contains the name of
701 * the executable (excluding the path) for the current task. The
702 * *size_of_buf* must be strictly positive. On success, the
703 * helper makes sure that the *buf* is NUL-terminated. On failure,
704 * it is filled with zeroes.
705 * Return
706 * 0 on success, or a negative error in case of failure.
707 *
708 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
709 * Description
710 * Retrieve the classid for the current task, i.e. for the net_cls
711 * cgroup to which *skb* belongs.
712 *
713 * This helper can be used on TC egress path, but not on ingress.
714 *
715 * The net_cls cgroup provides an interface to tag network packets
716 * based on a user-provided identifier for all traffic coming from
717 * the tasks belonging to the related cgroup. See also the related
718 * kernel documentation, available from the Linux sources in file
719 * *Documentation/cgroup-v1/net_cls.txt*.
720 *
721 * The Linux kernel has two versions for cgroups: there are
722 * cgroups v1 and cgroups v2. Both are available to users, who can
723 * use a mixture of them, but note that the net_cls cgroup is for
724 * cgroup v1 only. This makes it incompatible with BPF programs
725 * run on cgroups, which is a cgroup-v2-only feature (a socket can
726 * only hold data for one version of cgroups at a time).
727 *
728 * This helper is only available is the kernel was compiled with
729 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
730 * "**y**" or to "**m**".
731 * Return
732 * The classid, or 0 for the default unconfigured classid.
733 *
734 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
735 * Description
736 * Push a *vlan_tci* (VLAN tag control information) of protocol
737 * *vlan_proto* to the packet associated to *skb*, then update
738 * the checksum. Note that if *vlan_proto* is different from
739 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
740 * be **ETH_P_8021Q**.
741 *
742 * A call to this helper is susceptible to change the underlaying
743 * packet buffer. Therefore, at load time, all checks on pointers
744 * previously done by the verifier are invalidated and must be
745 * performed again, if the helper is used in combination with
746 * direct packet access.
747 * Return
748 * 0 on success, or a negative error in case of failure.
749 *
750 * int bpf_skb_vlan_pop(struct sk_buff *skb)
751 * Description
752 * Pop a VLAN header from the packet associated to *skb*.
753 *
754 * A call to this helper is susceptible to change the underlaying
755 * packet buffer. Therefore, at load time, all checks on pointers
756 * previously done by the verifier are invalidated and must be
757 * performed again, if the helper is used in combination with
758 * direct packet access.
759 * Return
760 * 0 on success, or a negative error in case of failure.
761 *
762 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
763 * Description
764 * Get tunnel metadata. This helper takes a pointer *key* to an
765 * empty **struct bpf_tunnel_key** of **size**, that will be
766 * filled with tunnel metadata for the packet associated to *skb*.
767 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
768 * indicates that the tunnel is based on IPv6 protocol instead of
769 * IPv4.
770 *
771 * The **struct bpf_tunnel_key** is an object that generalizes the
772 * principal parameters used by various tunneling protocols into a
773 * single struct. This way, it can be used to easily make a
774 * decision based on the contents of the encapsulation header,
775 * "summarized" in this struct. In particular, it holds the IP
776 * address of the remote end (IPv4 or IPv6, depending on the case)
777 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
778 * this struct exposes the *key*\ **->tunnel_id**, which is
779 * generally mapped to a VNI (Virtual Network Identifier), making
780 * it programmable together with the **bpf_skb_set_tunnel_key**\
781 * () helper.
782 *
783 * Let's imagine that the following code is part of a program
784 * attached to the TC ingress interface, on one end of a GRE
785 * tunnel, and is supposed to filter out all messages coming from
786 * remote ends with IPv4 address other than 10.0.0.1:
787 *
788 * ::
789 *
790 * int ret;
791 * struct bpf_tunnel_key key = {};
792 *
793 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
794 * if (ret < 0)
795 * return TC_ACT_SHOT; // drop packet
796 *
797 * if (key.remote_ipv4 != 0x0a000001)
798 * return TC_ACT_SHOT; // drop packet
799 *
800 * return TC_ACT_OK; // accept packet
801 *
802 * This interface can also be used with all encapsulation devices
803 * that can operate in "collect metadata" mode: instead of having
804 * one network device per specific configuration, the "collect
805 * metadata" mode only requires a single device where the
806 * configuration can be extracted from this helper.
807 *
808 * This can be used together with various tunnels such as VXLan,
809 * Geneve, GRE or IP in IP (IPIP).
810 * Return
811 * 0 on success, or a negative error in case of failure.
812 *
813 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
814 * Description
815 * Populate tunnel metadata for packet associated to *skb.* The
816 * tunnel metadata is set to the contents of *key*, of *size*. The
817 * *flags* can be set to a combination of the following values:
818 *
819 * **BPF_F_TUNINFO_IPV6**
820 * Indicate that the tunnel is based on IPv6 protocol
821 * instead of IPv4.
822 * **BPF_F_ZERO_CSUM_TX**
823 * For IPv4 packets, add a flag to tunnel metadata
824 * indicating that checksum computation should be skipped
825 * and checksum set to zeroes.
826 * **BPF_F_DONT_FRAGMENT**
827 * Add a flag to tunnel metadata indicating that the
828 * packet should not be fragmented.
829 * **BPF_F_SEQ_NUMBER**
830 * Add a flag to tunnel metadata indicating that a
831 * sequence number should be added to tunnel header before
832 * sending the packet. This flag was added for GRE
833 * encapsulation, but might be used with other protocols
834 * as well in the future.
835 *
836 * Here is a typical usage on the transmit path:
837 *
838 * ::
839 *
840 * struct bpf_tunnel_key key;
841 * populate key ...
842 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
843 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
844 *
845 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
846 * helper for additional information.
847 * Return
848 * 0 on success, or a negative error in case of failure.
849 *
850 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
851 * Description
852 * Read the value of a perf event counter. This helper relies on a
853 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
854 * the perf event counter is selected when *map* is updated with
855 * perf event file descriptors. The *map* is an array whose size
856 * is the number of available CPUs, and each cell contains a value
857 * relative to one CPU. The value to retrieve is indicated by
858 * *flags*, that contains the index of the CPU to look up, masked
859 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
860 * **BPF_F_CURRENT_CPU** to indicate that the value for the
861 * current CPU should be retrieved.
862 *
863 * Note that before Linux 4.13, only hardware perf event can be
864 * retrieved.
865 *
866 * Also, be aware that the newer helper
867 * **bpf_perf_event_read_value**\ () is recommended over
868 * **bpf_perf_event_read**\ () in general. The latter has some ABI
869 * quirks where error and counter value are used as a return code
870 * (which is wrong to do since ranges may overlap). This issue is
871 * fixed with **bpf_perf_event_read_value**\ (), which at the same
872 * time provides more features over the **bpf_perf_event_read**\
873 * () interface. Please refer to the description of
874 * **bpf_perf_event_read_value**\ () for details.
875 * Return
876 * The value of the perf event counter read from the map, or a
877 * negative error code in case of failure.
878 *
879 * int bpf_redirect(u32 ifindex, u64 flags)
880 * Description
881 * Redirect the packet to another net device of index *ifindex*.
882 * This helper is somewhat similar to **bpf_clone_redirect**\
883 * (), except that the packet is not cloned, which provides
884 * increased performance.
885 *
886 * Except for XDP, both ingress and egress interfaces can be used
887 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
888 * to make the distinction (ingress path is selected if the flag
889 * is present, egress path otherwise). Currently, XDP only
890 * supports redirection to the egress interface, and accepts no
891 * flag at all.
892 *
893 * The same effect can be attained with the more generic
894 * **bpf_redirect_map**\ (), which requires specific maps to be
895 * used but offers better performance.
896 * Return
897 * For XDP, the helper returns **XDP_REDIRECT** on success or
898 * **XDP_ABORTED** on error. For other program types, the values
899 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
900 * error.
901 *
902 * u32 bpf_get_route_realm(struct sk_buff *skb)
903 * Description
904 * Retrieve the realm or the route, that is to say the
905 * **tclassid** field of the destination for the *skb*. The
906 * indentifier retrieved is a user-provided tag, similar to the
907 * one used with the net_cls cgroup (see description for
908 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
909 * held by a route (a destination entry), not by a task.
910 *
911 * Retrieving this identifier works with the clsact TC egress hook
912 * (see also **tc-bpf(8)**), or alternatively on conventional
913 * classful egress qdiscs, but not on TC ingress path. In case of
914 * clsact TC egress hook, this has the advantage that, internally,
915 * the destination entry has not been dropped yet in the transmit
916 * path. Therefore, the destination entry does not need to be
917 * artificially held via **netif_keep_dst**\ () for a classful
918 * qdisc until the *skb* is freed.
919 *
920 * This helper is available only if the kernel was compiled with
921 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
922 * Return
923 * The realm of the route for the packet associated to *skb*, or 0
924 * if none was found.
925 *
926 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
927 * Description
928 * Write raw *data* blob into a special BPF perf event held by
929 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
930 * event must have the following attributes: **PERF_SAMPLE_RAW**
931 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
932 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
933 *
934 * The *flags* are used to indicate the index in *map* for which
935 * the value must be put, masked with **BPF_F_INDEX_MASK**.
936 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
937 * to indicate that the index of the current CPU core should be
938 * used.
939 *
940 * The value to write, of *size*, is passed through eBPF stack and
941 * pointed by *data*.
942 *
943 * The context of the program *ctx* needs also be passed to the
944 * helper.
945 *
946 * On user space, a program willing to read the values needs to
947 * call **perf_event_open**\ () on the perf event (either for
948 * one or for all CPUs) and to store the file descriptor into the
949 * *map*. This must be done before the eBPF program can send data
950 * into it. An example is available in file
951 * *samples/bpf/trace_output_user.c* in the Linux kernel source
952 * tree (the eBPF program counterpart is in
953 * *samples/bpf/trace_output_kern.c*).
954 *
955 * **bpf_perf_event_output**\ () achieves better performance
956 * than **bpf_trace_printk**\ () for sharing data with user
957 * space, and is much better suitable for streaming data from eBPF
958 * programs.
959 *
960 * Note that this helper is not restricted to tracing use cases
961 * and can be used with programs attached to TC or XDP as well,
962 * where it allows for passing data to user space listeners. Data
963 * can be:
964 *
965 * * Only custom structs,
966 * * Only the packet payload, or
967 * * A combination of both.
968 * Return
969 * 0 on success, or a negative error in case of failure.
970 *
971 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
972 * Description
973 * This helper was provided as an easy way to load data from a
974 * packet. It can be used to load *len* bytes from *offset* from
975 * the packet associated to *skb*, into the buffer pointed by
976 * *to*.
977 *
978 * Since Linux 4.7, usage of this helper has mostly been replaced
979 * by "direct packet access", enabling packet data to be
980 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
981 * pointing respectively to the first byte of packet data and to
982 * the byte after the last byte of packet data. However, it
983 * remains useful if one wishes to read large quantities of data
984 * at once from a packet into the eBPF stack.
985 * Return
986 * 0 on success, or a negative error in case of failure.
987 *
988 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
989 * Description
990 * Walk a user or a kernel stack and return its id. To achieve
991 * this, the helper needs *ctx*, which is a pointer to the context
992 * on which the tracing program is executed, and a pointer to a
993 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
994 *
995 * The last argument, *flags*, holds the number of stack frames to
996 * skip (from 0 to 255), masked with
997 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
998 * a combination of the following flags:
999 *
1000 * **BPF_F_USER_STACK**
1001 * Collect a user space stack instead of a kernel stack.
1002 * **BPF_F_FAST_STACK_CMP**
1003 * Compare stacks by hash only.
1004 * **BPF_F_REUSE_STACKID**
1005 * If two different stacks hash into the same *stackid*,
1006 * discard the old one.
1007 *
1008 * The stack id retrieved is a 32 bit long integer handle which
1009 * can be further combined with other data (including other stack
1010 * ids) and used as a key into maps. This can be useful for
1011 * generating a variety of graphs (such as flame graphs or off-cpu
1012 * graphs).
1013 *
1014 * For walking a stack, this helper is an improvement over
1015 * **bpf_probe_read**\ (), which can be used with unrolled loops
1016 * but is not efficient and consumes a lot of eBPF instructions.
1017 * Instead, **bpf_get_stackid**\ () can collect up to
1018 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1019 * this limit can be controlled with the **sysctl** program, and
1020 * that it should be manually increased in order to profile long
1021 * user stacks (such as stacks for Java programs). To do so, use:
1022 *
1023 * ::
1024 *
1025 * # sysctl kernel.perf_event_max_stack=<new value>
1026 * Return
1027 * The positive or null stack id on success, or a negative error
1028 * in case of failure.
1029 *
1030 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1031 * Description
1032 * Compute a checksum difference, from the raw buffer pointed by
1033 * *from*, of length *from_size* (that must be a multiple of 4),
1034 * towards the raw buffer pointed by *to*, of size *to_size*
1035 * (same remark). An optional *seed* can be added to the value
1036 * (this can be cascaded, the seed may come from a previous call
1037 * to the helper).
1038 *
1039 * This is flexible enough to be used in several ways:
1040 *
1041 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1042 * checksum, it can be used when pushing new data.
1043 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1044 * checksum, it can be used when removing data from a packet.
1045 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1046 * can be used to compute a diff. Note that *from_size* and
1047 * *to_size* do not need to be equal.
1048 *
1049 * This helper can be used in combination with
1050 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1051 * which one can feed in the difference computed with
1052 * **bpf_csum_diff**\ ().
1053 * Return
1054 * The checksum result, or a negative error code in case of
1055 * failure.
1056 *
1057 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1058 * Description
1059 * Retrieve tunnel options metadata for the packet associated to
1060 * *skb*, and store the raw tunnel option data to the buffer *opt*
1061 * of *size*.
1062 *
1063 * This helper can be used with encapsulation devices that can
1064 * operate in "collect metadata" mode (please refer to the related
1065 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1066 * more details). A particular example where this can be used is
1067 * in combination with the Geneve encapsulation protocol, where it
1068 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1069 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1070 * the eBPF program. This allows for full customization of these
1071 * headers.
1072 * Return
1073 * The size of the option data retrieved.
1074 *
1075 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1076 * Description
1077 * Set tunnel options metadata for the packet associated to *skb*
1078 * to the option data contained in the raw buffer *opt* of *size*.
1079 *
1080 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1081 * helper for additional information.
1082 * Return
1083 * 0 on success, or a negative error in case of failure.
1084 *
1085 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1086 * Description
1087 * Change the protocol of the *skb* to *proto*. Currently
1088 * supported are transition from IPv4 to IPv6, and from IPv6 to
1089 * IPv4. The helper takes care of the groundwork for the
1090 * transition, including resizing the socket buffer. The eBPF
1091 * program is expected to fill the new headers, if any, via
1092 * **skb_store_bytes**\ () and to recompute the checksums with
1093 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1094 * (). The main case for this helper is to perform NAT64
1095 * operations out of an eBPF program.
1096 *
1097 * Internally, the GSO type is marked as dodgy so that headers are
1098 * checked and segments are recalculated by the GSO/GRO engine.
1099 * The size for GSO target is adapted as well.
1100 *
1101 * All values for *flags* are reserved for future usage, and must
1102 * be left at zero.
1103 *
1104 * A call to this helper is susceptible to change the underlaying
1105 * packet buffer. Therefore, at load time, all checks on pointers
1106 * previously done by the verifier are invalidated and must be
1107 * performed again, if the helper is used in combination with
1108 * direct packet access.
1109 * Return
1110 * 0 on success, or a negative error in case of failure.
1111 *
1112 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1113 * Description
1114 * Change the packet type for the packet associated to *skb*. This
1115 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1116 * the eBPF program does not have a write access to *skb*\
1117 * **->pkt_type** beside this helper. Using a helper here allows
1118 * for graceful handling of errors.
1119 *
1120 * The major use case is to change incoming *skb*s to
1121 * **PACKET_HOST** in a programmatic way instead of having to
1122 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1123 * example.
1124 *
1125 * Note that *type* only allows certain values. At this time, they
1126 * are:
1127 *
1128 * **PACKET_HOST**
1129 * Packet is for us.
1130 * **PACKET_BROADCAST**
1131 * Send packet to all.
1132 * **PACKET_MULTICAST**
1133 * Send packet to group.
1134 * **PACKET_OTHERHOST**
1135 * Send packet to someone else.
1136 * Return
1137 * 0 on success, or a negative error in case of failure.
1138 *
1139 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1140 * Description
1141 * Check whether *skb* is a descendant of the cgroup2 held by
1142 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1143 * Return
1144 * The return value depends on the result of the test, and can be:
1145 *
1146 * * 0, if the *skb* failed the cgroup2 descendant test.
1147 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1148 * * A negative error code, if an error occurred.
1149 *
1150 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1151 * Description
1152 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1153 * not set, in particular if the hash was cleared due to mangling,
1154 * recompute this hash. Later accesses to the hash can be done
1155 * directly with *skb*\ **->hash**.
1156 *
1157 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1158 * prototype with **bpf_skb_change_proto**\ (), or calling
1159 * **bpf_skb_store_bytes**\ () with the
1160 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1161 * the hash and to trigger a new computation for the next call to
1162 * **bpf_get_hash_recalc**\ ().
1163 * Return
1164 * The 32-bit hash.
1165 *
1166 * u64 bpf_get_current_task(void)
1167 * Return
1168 * A pointer to the current task struct.
1169 *
1170 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1171 * Description
1172 * Attempt in a safe way to write *len* bytes from the buffer
1173 * *src* to *dst* in memory. It only works for threads that are in
1174 * user context, and *dst* must be a valid user space address.
1175 *
1176 * This helper should not be used to implement any kind of
1177 * security mechanism because of TOC-TOU attacks, but rather to
1178 * debug, divert, and manipulate execution of semi-cooperative
1179 * processes.
1180 *
1181 * Keep in mind that this feature is meant for experiments, and it
1182 * has a risk of crashing the system and running programs.
1183 * Therefore, when an eBPF program using this helper is attached,
1184 * a warning including PID and process name is printed to kernel
1185 * logs.
1186 * Return
1187 * 0 on success, or a negative error in case of failure.
1188 *
1189 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1190 * Description
1191 * Check whether the probe is being run is the context of a given
1192 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1193 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1194 * Return
1195 * The return value depends on the result of the test, and can be:
1196 *
1197 * * 0, if the *skb* task belongs to the cgroup2.
1198 * * 1, if the *skb* task does not belong to the cgroup2.
1199 * * A negative error code, if an error occurred.
1200 *
1201 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1202 * Description
1203 * Resize (trim or grow) the packet associated to *skb* to the
1204 * new *len*. The *flags* are reserved for future usage, and must
1205 * be left at zero.
1206 *
1207 * The basic idea is that the helper performs the needed work to
1208 * change the size of the packet, then the eBPF program rewrites
1209 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1210 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1211 * and others. This helper is a slow path utility intended for
1212 * replies with control messages. And because it is targeted for
1213 * slow path, the helper itself can afford to be slow: it
1214 * implicitly linearizes, unclones and drops offloads from the
1215 * *skb*.
1216 *
1217 * A call to this helper is susceptible to change the underlaying
1218 * packet buffer. Therefore, at load time, all checks on pointers
1219 * previously done by the verifier are invalidated and must be
1220 * performed again, if the helper is used in combination with
1221 * direct packet access.
1222 * Return
1223 * 0 on success, or a negative error in case of failure.
1224 *
1225 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1226 * Description
1227 * Pull in non-linear data in case the *skb* is non-linear and not
1228 * all of *len* are part of the linear section. Make *len* bytes
1229 * from *skb* readable and writable. If a zero value is passed for
1230 * *len*, then the whole length of the *skb* is pulled.
1231 *
1232 * This helper is only needed for reading and writing with direct
1233 * packet access.
1234 *
1235 * For direct packet access, testing that offsets to access
1236 * are within packet boundaries (test on *skb*\ **->data_end**) is
1237 * susceptible to fail if offsets are invalid, or if the requested
1238 * data is in non-linear parts of the *skb*. On failure the
1239 * program can just bail out, or in the case of a non-linear
1240 * buffer, use a helper to make the data available. The
1241 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1242 * the data. Another one consists in using **bpf_skb_pull_data**
1243 * to pull in once the non-linear parts, then retesting and
1244 * eventually access the data.
1245 *
1246 * At the same time, this also makes sure the *skb* is uncloned,
1247 * which is a necessary condition for direct write. As this needs
1248 * to be an invariant for the write part only, the verifier
1249 * detects writes and adds a prologue that is calling
1250 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1251 * the very beginning in case it is indeed cloned.
1252 *
1253 * A call to this helper is susceptible to change the underlaying
1254 * packet buffer. Therefore, at load time, all checks on pointers
1255 * previously done by the verifier are invalidated and must be
1256 * performed again, if the helper is used in combination with
1257 * direct packet access.
1258 * Return
1259 * 0 on success, or a negative error in case of failure.
1260 *
1261 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1262 * Description
1263 * Add the checksum *csum* into *skb*\ **->csum** in case the
1264 * driver has supplied a checksum for the entire packet into that
1265 * field. Return an error otherwise. This helper is intended to be
1266 * used in combination with **bpf_csum_diff**\ (), in particular
1267 * when the checksum needs to be updated after data has been
1268 * written into the packet through direct packet access.
1269 * Return
1270 * The checksum on success, or a negative error code in case of
1271 * failure.
1272 *
1273 * void bpf_set_hash_invalid(struct sk_buff *skb)
1274 * Description
1275 * Invalidate the current *skb*\ **->hash**. It can be used after
1276 * mangling on headers through direct packet access, in order to
1277 * indicate that the hash is outdated and to trigger a
1278 * recalculation the next time the kernel tries to access this
1279 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1280 *
1281 * int bpf_get_numa_node_id(void)
1282 * Description
1283 * Return the id of the current NUMA node. The primary use case
1284 * for this helper is the selection of sockets for the local NUMA
1285 * node, when the program is attached to sockets using the
1286 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1287 * but the helper is also available to other eBPF program types,
1288 * similarly to **bpf_get_smp_processor_id**\ ().
1289 * Return
1290 * The id of current NUMA node.
1291 *
1292 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1293 * Description
1294 * Grows headroom of packet associated to *skb* and adjusts the
1295 * offset of the MAC header accordingly, adding *len* bytes of
1296 * space. It automatically extends and reallocates memory as
1297 * required.
1298 *
1299 * This helper can be used on a layer 3 *skb* to push a MAC header
1300 * for redirection into a layer 2 device.
1301 *
1302 * All values for *flags* are reserved for future usage, and must
1303 * be left at zero.
1304 *
1305 * A call to this helper is susceptible to change the underlaying
1306 * packet buffer. Therefore, at load time, all checks on pointers
1307 * previously done by the verifier are invalidated and must be
1308 * performed again, if the helper is used in combination with
1309 * direct packet access.
1310 * Return
1311 * 0 on success, or a negative error in case of failure.
1312 *
1313 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1314 * Description
1315 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1316 * it is possible to use a negative value for *delta*. This helper
1317 * can be used to prepare the packet for pushing or popping
1318 * headers.
1319 *
1320 * A call to this helper is susceptible to change the underlaying
1321 * packet buffer. Therefore, at load time, all checks on pointers
1322 * previously done by the verifier are invalidated and must be
1323 * performed again, if the helper is used in combination with
1324 * direct packet access.
1325 * Return
1326 * 0 on success, or a negative error in case of failure.
1327 *
1328 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1329 * Description
1330 * Copy a NUL terminated string from an unsafe address
1331 * *unsafe_ptr* to *dst*. The *size* should include the
1332 * terminating NUL byte. In case the string length is smaller than
1333 * *size*, the target is not padded with further NUL bytes. If the
1334 * string length is larger than *size*, just *size*-1 bytes are
1335 * copied and the last byte is set to NUL.
1336 *
1337 * On success, the length of the copied string is returned. This
1338 * makes this helper useful in tracing programs for reading
1339 * strings, and more importantly to get its length at runtime. See
1340 * the following snippet:
1341 *
1342 * ::
1343 *
1344 * SEC("kprobe/sys_open")
1345 * void bpf_sys_open(struct pt_regs *ctx)
1346 * {
1347 * char buf[PATHLEN]; // PATHLEN is defined to 256
1348 * int res = bpf_probe_read_str(buf, sizeof(buf),
1349 * ctx->di);
1350 *
1351 * // Consume buf, for example push it to
1352 * // userspace via bpf_perf_event_output(); we
1353 * // can use res (the string length) as event
1354 * // size, after checking its boundaries.
1355 * }
1356 *
1357 * In comparison, using **bpf_probe_read()** helper here instead
1358 * to read the string would require to estimate the length at
1359 * compile time, and would often result in copying more memory
1360 * than necessary.
1361 *
1362 * Another useful use case is when parsing individual process
1363 * arguments or individual environment variables navigating
1364 * *current*\ **->mm->arg_start** and *current*\
1365 * **->mm->env_start**: using this helper and the return value,
1366 * one can quickly iterate at the right offset of the memory area.
1367 * Return
1368 * On success, the strictly positive length of the string,
1369 * including the trailing NUL character. On error, a negative
1370 * value.
1371 *
1372 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1373 * Description
1374 * If the **struct sk_buff** pointed by *skb* has a known socket,
1375 * retrieve the cookie (generated by the kernel) of this socket.
1376 * If no cookie has been set yet, generate a new cookie. Once
1377 * generated, the socket cookie remains stable for the life of the
1378 * socket. This helper can be useful for monitoring per socket
1379 * networking traffic statistics as it provides a unique socket
1380 * identifier per namespace.
1381 * Return
1382 * A 8-byte long non-decreasing number on success, or 0 if the
1383 * socket field is missing inside *skb*.
1384 *
1385 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1386 * Description
1387 * Equivalent to bpf_get_socket_cookie() helper that accepts
1388 * *skb*, but gets socket from **struct bpf_sock_addr** contex.
1389 * Return
1390 * A 8-byte long non-decreasing number.
1391 *
1392 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1393 * Description
1394 * Equivalent to bpf_get_socket_cookie() helper that accepts
1395 * *skb*, but gets socket from **struct bpf_sock_ops** contex.
1396 * Return
1397 * A 8-byte long non-decreasing number.
1398 *
1399 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1400 * Return
1401 * The owner UID of the socket associated to *skb*. If the socket
1402 * is **NULL**, or if it is not a full socket (i.e. if it is a
1403 * time-wait or a request socket instead), **overflowuid** value
1404 * is returned (note that **overflowuid** might also be the actual
1405 * UID value for the socket).
1406 *
1407 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1408 * Description
1409 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1410 * to value *hash*.
1411 * Return
1412 * 0
1413 *
1414 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1415 * Description
1416 * Emulate a call to **setsockopt()** on the socket associated to
1417 * *bpf_socket*, which must be a full socket. The *level* at
1418 * which the option resides and the name *optname* of the option
1419 * must be specified, see **setsockopt(2)** for more information.
1420 * The option value of length *optlen* is pointed by *optval*.
1421 *
1422 * This helper actually implements a subset of **setsockopt()**.
1423 * It supports the following *level*\ s:
1424 *
1425 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1426 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1427 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1428 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1429 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1430 * **TCP_BPF_SNDCWND_CLAMP**.
1431 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1432 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1433 * Return
1434 * 0 on success, or a negative error in case of failure.
1435 *
1436 * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags)
1437 * Description
1438 * Grow or shrink the room for data in the packet associated to
1439 * *skb* by *len_diff*, and according to the selected *mode*.
1440 *
1441 * There is a single supported mode at this time:
1442 *
1443 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1444 * (room space is added or removed below the layer 3 header).
1445 *
1446 * All values for *flags* are reserved for future usage, and must
1447 * be left at zero.
1448 *
1449 * A call to this helper is susceptible to change the underlaying
1450 * packet buffer. Therefore, at load time, all checks on pointers
1451 * previously done by the verifier are invalidated and must be
1452 * performed again, if the helper is used in combination with
1453 * direct packet access.
1454 * Return
1455 * 0 on success, or a negative error in case of failure.
1456 *
1457 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1458 * Description
1459 * Redirect the packet to the endpoint referenced by *map* at
1460 * index *key*. Depending on its type, this *map* can contain
1461 * references to net devices (for forwarding packets through other
1462 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1463 * but this is only implemented for native XDP (with driver
1464 * support) as of this writing).
1465 *
1466 * All values for *flags* are reserved for future usage, and must
1467 * be left at zero.
1468 *
1469 * When used to redirect packets to net devices, this helper
1470 * provides a high performance increase over **bpf_redirect**\ ().
1471 * This is due to various implementation details of the underlying
1472 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1473 * () tries to send packet as a "bulk" to the device.
1474 * Return
1475 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1476 *
1477 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1478 * Description
1479 * Redirect the packet to the socket referenced by *map* (of type
1480 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1481 * egress interfaces can be used for redirection. The
1482 * **BPF_F_INGRESS** value in *flags* is used to make the
1483 * distinction (ingress path is selected if the flag is present,
1484 * egress path otherwise). This is the only flag supported for now.
1485 * Return
1486 * **SK_PASS** on success, or **SK_DROP** on error.
1487 *
1488 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1489 * Description
1490 * Add an entry to, or update a *map* referencing sockets. The
1491 * *skops* is used as a new value for the entry associated to
1492 * *key*. *flags* is one of:
1493 *
1494 * **BPF_NOEXIST**
1495 * The entry for *key* must not exist in the map.
1496 * **BPF_EXIST**
1497 * The entry for *key* must already exist in the map.
1498 * **BPF_ANY**
1499 * No condition on the existence of the entry for *key*.
1500 *
1501 * If the *map* has eBPF programs (parser and verdict), those will
1502 * be inherited by the socket being added. If the socket is
1503 * already attached to eBPF programs, this results in an error.
1504 * Return
1505 * 0 on success, or a negative error in case of failure.
1506 *
1507 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1508 * Description
1509 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1510 * *delta* (which can be positive or negative). Note that this
1511 * operation modifies the address stored in *xdp_md*\ **->data**,
1512 * so the latter must be loaded only after the helper has been
1513 * called.
1514 *
1515 * The use of *xdp_md*\ **->data_meta** is optional and programs
1516 * are not required to use it. The rationale is that when the
1517 * packet is processed with XDP (e.g. as DoS filter), it is
1518 * possible to push further meta data along with it before passing
1519 * to the stack, and to give the guarantee that an ingress eBPF
1520 * program attached as a TC classifier on the same device can pick
1521 * this up for further post-processing. Since TC works with socket
1522 * buffers, it remains possible to set from XDP the **mark** or
1523 * **priority** pointers, or other pointers for the socket buffer.
1524 * Having this scratch space generic and programmable allows for
1525 * more flexibility as the user is free to store whatever meta
1526 * data they need.
1527 *
1528 * A call to this helper is susceptible to change the underlaying
1529 * packet buffer. Therefore, at load time, all checks on pointers
1530 * previously done by the verifier are invalidated and must be
1531 * performed again, if the helper is used in combination with
1532 * direct packet access.
1533 * Return
1534 * 0 on success, or a negative error in case of failure.
1535 *
1536 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1537 * Description
1538 * Read the value of a perf event counter, and store it into *buf*
1539 * of size *buf_size*. This helper relies on a *map* of type
1540 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1541 * counter is selected when *map* is updated with perf event file
1542 * descriptors. The *map* is an array whose size is the number of
1543 * available CPUs, and each cell contains a value relative to one
1544 * CPU. The value to retrieve is indicated by *flags*, that
1545 * contains the index of the CPU to look up, masked with
1546 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1547 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1548 * current CPU should be retrieved.
1549 *
1550 * This helper behaves in a way close to
1551 * **bpf_perf_event_read**\ () helper, save that instead of
1552 * just returning the value observed, it fills the *buf*
1553 * structure. This allows for additional data to be retrieved: in
1554 * particular, the enabled and running times (in *buf*\
1555 * **->enabled** and *buf*\ **->running**, respectively) are
1556 * copied. In general, **bpf_perf_event_read_value**\ () is
1557 * recommended over **bpf_perf_event_read**\ (), which has some
1558 * ABI issues and provides fewer functionalities.
1559 *
1560 * These values are interesting, because hardware PMU (Performance
1561 * Monitoring Unit) counters are limited resources. When there are
1562 * more PMU based perf events opened than available counters,
1563 * kernel will multiplex these events so each event gets certain
1564 * percentage (but not all) of the PMU time. In case that
1565 * multiplexing happens, the number of samples or counter value
1566 * will not reflect the case compared to when no multiplexing
1567 * occurs. This makes comparison between different runs difficult.
1568 * Typically, the counter value should be normalized before
1569 * comparing to other experiments. The usual normalization is done
1570 * as follows.
1571 *
1572 * ::
1573 *
1574 * normalized_counter = counter * t_enabled / t_running
1575 *
1576 * Where t_enabled is the time enabled for event and t_running is
1577 * the time running for event since last normalization. The
1578 * enabled and running times are accumulated since the perf event
1579 * open. To achieve scaling factor between two invocations of an
1580 * eBPF program, users can can use CPU id as the key (which is
1581 * typical for perf array usage model) to remember the previous
1582 * value and do the calculation inside the eBPF program.
1583 * Return
1584 * 0 on success, or a negative error in case of failure.
1585 *
1586 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1587 * Description
1588 * For en eBPF program attached to a perf event, retrieve the
1589 * value of the event counter associated to *ctx* and store it in
1590 * the structure pointed by *buf* and of size *buf_size*. Enabled
1591 * and running times are also stored in the structure (see
1592 * description of helper **bpf_perf_event_read_value**\ () for
1593 * more details).
1594 * Return
1595 * 0 on success, or a negative error in case of failure.
1596 *
1597 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1598 * Description
1599 * Emulate a call to **getsockopt()** on the socket associated to
1600 * *bpf_socket*, which must be a full socket. The *level* at
1601 * which the option resides and the name *optname* of the option
1602 * must be specified, see **getsockopt(2)** for more information.
1603 * The retrieved value is stored in the structure pointed by
1604 * *opval* and of length *optlen*.
1605 *
1606 * This helper actually implements a subset of **getsockopt()**.
1607 * It supports the following *level*\ s:
1608 *
1609 * * **IPPROTO_TCP**, which supports *optname*
1610 * **TCP_CONGESTION**.
1611 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1612 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1613 * Return
1614 * 0 on success, or a negative error in case of failure.
1615 *
1616 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1617 * Description
1618 * Used for error injection, this helper uses kprobes to override
1619 * the return value of the probed function, and to set it to *rc*.
1620 * The first argument is the context *regs* on which the kprobe
1621 * works.
1622 *
1623 * This helper works by setting setting the PC (program counter)
1624 * to an override function which is run in place of the original
1625 * probed function. This means the probed function is not run at
1626 * all. The replacement function just returns with the required
1627 * value.
1628 *
1629 * This helper has security implications, and thus is subject to
1630 * restrictions. It is only available if the kernel was compiled
1631 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1632 * option, and in this case it only works on functions tagged with
1633 * **ALLOW_ERROR_INJECTION** in the kernel code.
1634 *
1635 * Also, the helper is only available for the architectures having
1636 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1637 * x86 architecture is the only one to support this feature.
1638 * Return
1639 * 0
1640 *
1641 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1642 * Description
1643 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1644 * for the full TCP socket associated to *bpf_sock_ops* to
1645 * *argval*.
1646 *
1647 * The primary use of this field is to determine if there should
1648 * be calls to eBPF programs of type
1649 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1650 * code. A program of the same type can change its value, per
1651 * connection and as necessary, when the connection is
1652 * established. This field is directly accessible for reading, but
1653 * this helper must be used for updates in order to return an
1654 * error if an eBPF program tries to set a callback that is not
1655 * supported in the current kernel.
1656 *
1657 * The supported callback values that *argval* can combine are:
1658 *
1659 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1660 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1661 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1662 *
1663 * Here are some examples of where one could call such eBPF
1664 * program:
1665 *
1666 * * When RTO fires.
1667 * * When a packet is retransmitted.
1668 * * When the connection terminates.
1669 * * When a packet is sent.
1670 * * When a packet is received.
1671 * Return
1672 * Code **-EINVAL** if the socket is not a full TCP socket;
1673 * otherwise, a positive number containing the bits that could not
1674 * be set is returned (which comes down to 0 if all bits were set
1675 * as required).
1676 *
1677 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1678 * Description
1679 * This helper is used in programs implementing policies at the
1680 * socket level. If the message *msg* is allowed to pass (i.e. if
1681 * the verdict eBPF program returns **SK_PASS**), redirect it to
1682 * the socket referenced by *map* (of type
1683 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1684 * egress interfaces can be used for redirection. The
1685 * **BPF_F_INGRESS** value in *flags* is used to make the
1686 * distinction (ingress path is selected if the flag is present,
1687 * egress path otherwise). This is the only flag supported for now.
1688 * Return
1689 * **SK_PASS** on success, or **SK_DROP** on error.
1690 *
1691 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1692 * Description
1693 * For socket policies, apply the verdict of the eBPF program to
1694 * the next *bytes* (number of bytes) of message *msg*.
1695 *
1696 * For example, this helper can be used in the following cases:
1697 *
1698 * * A single **sendmsg**\ () or **sendfile**\ () system call
1699 * contains multiple logical messages that the eBPF program is
1700 * supposed to read and for which it should apply a verdict.
1701 * * An eBPF program only cares to read the first *bytes* of a
1702 * *msg*. If the message has a large payload, then setting up
1703 * and calling the eBPF program repeatedly for all bytes, even
1704 * though the verdict is already known, would create unnecessary
1705 * overhead.
1706 *
1707 * When called from within an eBPF program, the helper sets a
1708 * counter internal to the BPF infrastructure, that is used to
1709 * apply the last verdict to the next *bytes*. If *bytes* is
1710 * smaller than the current data being processed from a
1711 * **sendmsg**\ () or **sendfile**\ () system call, the first
1712 * *bytes* will be sent and the eBPF program will be re-run with
1713 * the pointer for start of data pointing to byte number *bytes*
1714 * **+ 1**. If *bytes* is larger than the current data being
1715 * processed, then the eBPF verdict will be applied to multiple
1716 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1717 * consumed.
1718 *
1719 * Note that if a socket closes with the internal counter holding
1720 * a non-zero value, this is not a problem because data is not
1721 * being buffered for *bytes* and is sent as it is received.
1722 * Return
1723 * 0
1724 *
1725 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1726 * Description
1727 * For socket policies, prevent the execution of the verdict eBPF
1728 * program for message *msg* until *bytes* (byte number) have been
1729 * accumulated.
1730 *
1731 * This can be used when one needs a specific number of bytes
1732 * before a verdict can be assigned, even if the data spans
1733 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1734 * case would be a user calling **sendmsg**\ () repeatedly with
1735 * 1-byte long message segments. Obviously, this is bad for
1736 * performance, but it is still valid. If the eBPF program needs
1737 * *bytes* bytes to validate a header, this helper can be used to
1738 * prevent the eBPF program to be called again until *bytes* have
1739 * been accumulated.
1740 * Return
1741 * 0
1742 *
1743 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1744 * Description
1745 * For socket policies, pull in non-linear data from user space
1746 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1747 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1748 * respectively.
1749 *
1750 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1751 * *msg* it can only parse data that the (**data**, **data_end**)
1752 * pointers have already consumed. For **sendmsg**\ () hooks this
1753 * is likely the first scatterlist element. But for calls relying
1754 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1755 * be the range (**0**, **0**) because the data is shared with
1756 * user space and by default the objective is to avoid allowing
1757 * user space to modify data while (or after) eBPF verdict is
1758 * being decided. This helper can be used to pull in data and to
1759 * set the start and end pointer to given values. Data will be
1760 * copied if necessary (i.e. if data was not linear and if start
1761 * and end pointers do not point to the same chunk).
1762 *
1763 * A call to this helper is susceptible to change the underlaying
1764 * packet buffer. Therefore, at load time, all checks on pointers
1765 * previously done by the verifier are invalidated and must be
1766 * performed again, if the helper is used in combination with
1767 * direct packet access.
1768 *
1769 * All values for *flags* are reserved for future usage, and must
1770 * be left at zero.
1771 * Return
1772 * 0 on success, or a negative error in case of failure.
1773 *
1774 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1775 * Description
1776 * Bind the socket associated to *ctx* to the address pointed by
1777 * *addr*, of length *addr_len*. This allows for making outgoing
1778 * connection from the desired IP address, which can be useful for
1779 * example when all processes inside a cgroup should use one
1780 * single IP address on a host that has multiple IP configured.
1781 *
1782 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1783 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1784 * **AF_INET6**). Looking for a free port to bind to can be
1785 * expensive, therefore binding to port is not permitted by the
1786 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1787 * must be set to zero.
1788 * Return
1789 * 0 on success, or a negative error in case of failure.
1790 *
1791 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1792 * Description
1793 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1794 * only possible to shrink the packet as of this writing,
1795 * therefore *delta* must be a negative integer.
1796 *
1797 * A call to this helper is susceptible to change the underlaying
1798 * packet buffer. Therefore, at load time, all checks on pointers
1799 * previously done by the verifier are invalidated and must be
1800 * performed again, if the helper is used in combination with
1801 * direct packet access.
1802 * Return
1803 * 0 on success, or a negative error in case of failure.
1804 *
1805 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1806 * Description
1807 * Retrieve the XFRM state (IP transform framework, see also
1808 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1809 *
1810 * The retrieved value is stored in the **struct bpf_xfrm_state**
1811 * pointed by *xfrm_state* and of length *size*.
1812 *
1813 * All values for *flags* are reserved for future usage, and must
1814 * be left at zero.
1815 *
1816 * This helper is available only if the kernel was compiled with
1817 * **CONFIG_XFRM** configuration option.
1818 * Return
1819 * 0 on success, or a negative error in case of failure.
1820 *
1821 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1822 * Description
1823 * Return a user or a kernel stack in bpf program provided buffer.
1824 * To achieve this, the helper needs *ctx*, which is a pointer
1825 * to the context on which the tracing program is executed.
1826 * To store the stacktrace, the bpf program provides *buf* with
1827 * a nonnegative *size*.
1828 *
1829 * The last argument, *flags*, holds the number of stack frames to
1830 * skip (from 0 to 255), masked with
1831 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1832 * the following flags:
1833 *
1834 * **BPF_F_USER_STACK**
1835 * Collect a user space stack instead of a kernel stack.
1836 * **BPF_F_USER_BUILD_ID**
1837 * Collect buildid+offset instead of ips for user stack,
1838 * only valid if **BPF_F_USER_STACK** is also specified.
1839 *
1840 * **bpf_get_stack**\ () can collect up to
1841 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1842 * to sufficient large buffer size. Note that
1843 * this limit can be controlled with the **sysctl** program, and
1844 * that it should be manually increased in order to profile long
1845 * user stacks (such as stacks for Java programs). To do so, use:
1846 *
1847 * ::
1848 *
1849 * # sysctl kernel.perf_event_max_stack=<new value>
1850 * Return
1851 * A non-negative value equal to or less than *size* on success,
1852 * or a negative error in case of failure.
1853 *
1854 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1855 * Description
1856 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1857 * it provides an easy way to load *len* bytes from *offset*
1858 * from the packet associated to *skb*, into the buffer pointed
1859 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1860 * a fifth argument *start_header* exists in order to select a
1861 * base offset to start from. *start_header* can be one of:
1862 *
1863 * **BPF_HDR_START_MAC**
1864 * Base offset to load data from is *skb*'s mac header.
1865 * **BPF_HDR_START_NET**
1866 * Base offset to load data from is *skb*'s network header.
1867 *
1868 * In general, "direct packet access" is the preferred method to
1869 * access packet data, however, this helper is in particular useful
1870 * in socket filters where *skb*\ **->data** does not always point
1871 * to the start of the mac header and where "direct packet access"
1872 * is not available.
1873 * Return
1874 * 0 on success, or a negative error in case of failure.
1875 *
1876 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1877 * Description
1878 * Do FIB lookup in kernel tables using parameters in *params*.
1879 * If lookup is successful and result shows packet is to be
1880 * forwarded, the neighbor tables are searched for the nexthop.
1881 * If successful (ie., FIB lookup shows forwarding and nexthop
1882 * is resolved), the nexthop address is returned in ipv4_dst
1883 * or ipv6_dst based on family, smac is set to mac address of
1884 * egress device, dmac is set to nexthop mac address, rt_metric
1885 * is set to metric from route (IPv4/IPv6 only), and ifindex
1886 * is set to the device index of the nexthop from the FIB lookup.
1887 *
1888 * *plen* argument is the size of the passed in struct.
1889 * *flags* argument can be a combination of one or more of the
1890 * following values:
1891 *
1892 * **BPF_FIB_LOOKUP_DIRECT**
1893 * Do a direct table lookup vs full lookup using FIB
1894 * rules.
1895 * **BPF_FIB_LOOKUP_OUTPUT**
1896 * Perform lookup from an egress perspective (default is
1897 * ingress).
1898 *
1899 * *ctx* is either **struct xdp_md** for XDP programs or
1900 * **struct sk_buff** tc cls_act programs.
1901 * Return
1902 * * < 0 if any input argument is invalid
1903 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1904 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1905 * packet is not forwarded or needs assist from full stack
1906 *
1907 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1908 * Description
1909 * Add an entry to, or update a sockhash *map* referencing sockets.
1910 * The *skops* is used as a new value for the entry associated to
1911 * *key*. *flags* is one of:
1912 *
1913 * **BPF_NOEXIST**
1914 * The entry for *key* must not exist in the map.
1915 * **BPF_EXIST**
1916 * The entry for *key* must already exist in the map.
1917 * **BPF_ANY**
1918 * No condition on the existence of the entry for *key*.
1919 *
1920 * If the *map* has eBPF programs (parser and verdict), those will
1921 * be inherited by the socket being added. If the socket is
1922 * already attached to eBPF programs, this results in an error.
1923 * Return
1924 * 0 on success, or a negative error in case of failure.
1925 *
1926 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1927 * Description
1928 * This helper is used in programs implementing policies at the
1929 * socket level. If the message *msg* is allowed to pass (i.e. if
1930 * the verdict eBPF program returns **SK_PASS**), redirect it to
1931 * the socket referenced by *map* (of type
1932 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1933 * egress interfaces can be used for redirection. The
1934 * **BPF_F_INGRESS** value in *flags* is used to make the
1935 * distinction (ingress path is selected if the flag is present,
1936 * egress path otherwise). This is the only flag supported for now.
1937 * Return
1938 * **SK_PASS** on success, or **SK_DROP** on error.
1939 *
1940 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1941 * Description
1942 * This helper is used in programs implementing policies at the
1943 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1944 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1945 * to the socket referenced by *map* (of type
1946 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1947 * egress interfaces can be used for redirection. The
1948 * **BPF_F_INGRESS** value in *flags* is used to make the
1949 * distinction (ingress path is selected if the flag is present,
1950 * egress otherwise). This is the only flag supported for now.
1951 * Return
1952 * **SK_PASS** on success, or **SK_DROP** on error.
1953 *
1954 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
1955 * Description
1956 * Encapsulate the packet associated to *skb* within a Layer 3
1957 * protocol header. This header is provided in the buffer at
1958 * address *hdr*, with *len* its size in bytes. *type* indicates
1959 * the protocol of the header and can be one of:
1960 *
1961 * **BPF_LWT_ENCAP_SEG6**
1962 * IPv6 encapsulation with Segment Routing Header
1963 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
1964 * the IPv6 header is computed by the kernel.
1965 * **BPF_LWT_ENCAP_SEG6_INLINE**
1966 * Only works if *skb* contains an IPv6 packet. Insert a
1967 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
1968 * the IPv6 header.
1969 *
1970 * A call to this helper is susceptible to change the underlaying
1971 * packet buffer. Therefore, at load time, all checks on pointers
1972 * previously done by the verifier are invalidated and must be
1973 * performed again, if the helper is used in combination with
1974 * direct packet access.
1975 * Return
1976 * 0 on success, or a negative error in case of failure.
1977 *
1978 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
1979 * Description
1980 * Store *len* bytes from address *from* into the packet
1981 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
1982 * inside the outermost IPv6 Segment Routing Header can be
1983 * modified through this helper.
1984 *
1985 * A call to this helper is susceptible to change the underlaying
1986 * packet buffer. Therefore, at load time, all checks on pointers
1987 * previously done by the verifier are invalidated and must be
1988 * performed again, if the helper is used in combination with
1989 * direct packet access.
1990 * Return
1991 * 0 on success, or a negative error in case of failure.
1992 *
1993 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
1994 * Description
1995 * Adjust the size allocated to TLVs in the outermost IPv6
1996 * Segment Routing Header contained in the packet associated to
1997 * *skb*, at position *offset* by *delta* bytes. Only offsets
1998 * after the segments are accepted. *delta* can be as well
1999 * positive (growing) as negative (shrinking).
2000 *
2001 * A call to this helper is susceptible to change the underlaying
2002 * packet buffer. Therefore, at load time, all checks on pointers
2003 * previously done by the verifier are invalidated and must be
2004 * performed again, if the helper is used in combination with
2005 * direct packet access.
2006 * Return
2007 * 0 on success, or a negative error in case of failure.
2008 *
2009 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2010 * Description
2011 * Apply an IPv6 Segment Routing action of type *action* to the
2012 * packet associated to *skb*. Each action takes a parameter
2013 * contained at address *param*, and of length *param_len* bytes.
2014 * *action* can be one of:
2015 *
2016 * **SEG6_LOCAL_ACTION_END_X**
2017 * End.X action: Endpoint with Layer-3 cross-connect.
2018 * Type of *param*: **struct in6_addr**.
2019 * **SEG6_LOCAL_ACTION_END_T**
2020 * End.T action: Endpoint with specific IPv6 table lookup.
2021 * Type of *param*: **int**.
2022 * **SEG6_LOCAL_ACTION_END_B6**
2023 * End.B6 action: Endpoint bound to an SRv6 policy.
2024 * Type of param: **struct ipv6_sr_hdr**.
2025 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2026 * End.B6.Encap action: Endpoint bound to an SRv6
2027 * encapsulation policy.
2028 * Type of param: **struct ipv6_sr_hdr**.
2029 *
2030 * A call to this helper is susceptible to change the underlaying
2031 * packet buffer. Therefore, at load time, all checks on pointers
2032 * previously done by the verifier are invalidated and must be
2033 * performed again, if the helper is used in combination with
2034 * direct packet access.
2035 * Return
2036 * 0 on success, or a negative error in case of failure.
2037 *
2038 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2039 * Description
2040 * This helper is used in programs implementing IR decoding, to
2041 * report a successfully decoded key press with *scancode*,
2042 * *toggle* value in the given *protocol*. The scancode will be
2043 * translated to a keycode using the rc keymap, and reported as
2044 * an input key down event. After a period a key up event is
2045 * generated. This period can be extended by calling either
2046 * **bpf_rc_keydown** () again with the same values, or calling
2047 * **bpf_rc_repeat** ().
2048 *
2049 * Some protocols include a toggle bit, in case the button was
2050 * released and pressed again between consecutive scancodes.
2051 *
2052 * The *ctx* should point to the lirc sample as passed into
2053 * the program.
2054 *
2055 * The *protocol* is the decoded protocol number (see
2056 * **enum rc_proto** for some predefined values).
2057 *
2058 * This helper is only available is the kernel was compiled with
2059 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2060 * "**y**".
2061 * Return
2062 * 0
2063 *
2064 * int bpf_rc_repeat(void *ctx)
2065 * Description
2066 * This helper is used in programs implementing IR decoding, to
2067 * report a successfully decoded repeat key message. This delays
2068 * the generation of a key up event for previously generated
2069 * key down event.
2070 *
2071 * Some IR protocols like NEC have a special IR message for
2072 * repeating last button, for when a button is held down.
2073 *
2074 * The *ctx* should point to the lirc sample as passed into
2075 * the program.
2076 *
2077 * This helper is only available is the kernel was compiled with
2078 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2079 * "**y**".
2080 * Return
2081 * 0
2082 *
2083 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2084 * Description
2085 * Return the cgroup v2 id of the socket associated with the *skb*.
2086 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2087 * helper for cgroup v1 by providing a tag resp. identifier that
2088 * can be matched on or used for map lookups e.g. to implement
2089 * policy. The cgroup v2 id of a given path in the hierarchy is
2090 * exposed in user space through the f_handle API in order to get
2091 * to the same 64-bit id.
2092 *
2093 * This helper can be used on TC egress path, but not on ingress,
2094 * and is available only if the kernel was compiled with the
2095 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2096 * Return
2097 * The id is returned or 0 in case the id could not be retrieved.
2098 *
2099 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2100 * Description
2101 * Return id of cgroup v2 that is ancestor of cgroup associated
2102 * with the *skb* at the *ancestor_level*. The root cgroup is at
2103 * *ancestor_level* zero and each step down the hierarchy
2104 * increments the level. If *ancestor_level* == level of cgroup
2105 * associated with *skb*, then return value will be same as that
2106 * of **bpf_skb_cgroup_id**\ ().
2107 *
2108 * The helper is useful to implement policies based on cgroups
2109 * that are upper in hierarchy than immediate cgroup associated
2110 * with *skb*.
2111 *
2112 * The format of returned id and helper limitations are same as in
2113 * **bpf_skb_cgroup_id**\ ().
2114 * Return
2115 * The id is returned or 0 in case the id could not be retrieved.
2116 *
2117 * u64 bpf_get_current_cgroup_id(void)
2118 * Return
2119 * A 64-bit integer containing the current cgroup id based
2120 * on the cgroup within which the current task is running.
2121 *
2122 * void* get_local_storage(void *map, u64 flags)
2123 * Description
2124 * Get the pointer to the local storage area.
2125 * The type and the size of the local storage is defined
2126 * by the *map* argument.
2127 * The *flags* meaning is specific for each map type,
2128 * and has to be 0 for cgroup local storage.
2129 *
2130 * Depending on the bpf program type, a local storage area
2131 * can be shared between multiple instances of the bpf program,
2132 * running simultaneously.
2133 *
2134 * A user should care about the synchronization by himself.
2135 * For example, by using the BPF_STX_XADD instruction to alter
2136 * the shared data.
2137 * Return
2138 * Pointer to the local storage area.
2139 *
2140 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2141 * Description
2142 * Select a SO_REUSEPORT sk from a BPF_MAP_TYPE_REUSEPORT_ARRAY map
2143 * It checks the selected sk is matching the incoming
2144 * request in the skb.
2145 * Return
2146 * 0 on success, or a negative error in case of failure.
2147 *
2148 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2149 * Description
2150 * Look for TCP socket matching *tuple*, optionally in a child
2151 * network namespace *netns*. The return value must be checked,
2152 * and if non-NULL, released via **bpf_sk_release**\ ().
2153 *
2154 * The *ctx* should point to the context of the program, such as
2155 * the skb or socket (depending on the hook in use). This is used
2156 * to determine the base network namespace for the lookup.
2157 *
2158 * *tuple_size* must be one of:
2159 *
2160 * **sizeof**\ (*tuple*\ **->ipv4**)
2161 * Look for an IPv4 socket.
2162 * **sizeof**\ (*tuple*\ **->ipv6**)
2163 * Look for an IPv6 socket.
2164 *
2165 * If the *netns* is zero, then the socket lookup table in the
2166 * netns associated with the *ctx* will be used. For the TC hooks,
2167 * this in the netns of the device in the skb. For socket hooks,
2168 * this in the netns of the socket. If *netns* is non-zero, then
2169 * it specifies the ID of the netns relative to the netns
2170 * associated with the *ctx*.
2171 *
2172 * All values for *flags* are reserved for future usage, and must
2173 * be left at zero.
2174 *
2175 * This helper is available only if the kernel was compiled with
2176 * **CONFIG_NET** configuration option.
2177 * Return
2178 * Pointer to *struct bpf_sock*, or NULL in case of failure.
2179 *
2180 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2181 * Description
2182 * Look for UDP socket matching *tuple*, optionally in a child
2183 * network namespace *netns*. The return value must be checked,
2184 * and if non-NULL, released via **bpf_sk_release**\ ().
2185 *
2186 * The *ctx* should point to the context of the program, such as
2187 * the skb or socket (depending on the hook in use). This is used
2188 * to determine the base network namespace for the lookup.
2189 *
2190 * *tuple_size* must be one of:
2191 *
2192 * **sizeof**\ (*tuple*\ **->ipv4**)
2193 * Look for an IPv4 socket.
2194 * **sizeof**\ (*tuple*\ **->ipv6**)
2195 * Look for an IPv6 socket.
2196 *
2197 * If the *netns* is zero, then the socket lookup table in the
2198 * netns associated with the *ctx* will be used. For the TC hooks,
2199 * this in the netns of the device in the skb. For socket hooks,
2200 * this in the netns of the socket. If *netns* is non-zero, then
2201 * it specifies the ID of the netns relative to the netns
2202 * associated with the *ctx*.
2203 *
2204 * All values for *flags* are reserved for future usage, and must
2205 * be left at zero.
2206 *
2207 * This helper is available only if the kernel was compiled with
2208 * **CONFIG_NET** configuration option.
2209 * Return
2210 * Pointer to *struct bpf_sock*, or NULL in case of failure.
2211 *
2212 * int bpf_sk_release(struct bpf_sock *sk)
2213 * Description
2214 * Release the reference held by *sock*. *sock* must be a non-NULL
2215 * pointer that was returned from bpf_sk_lookup_xxx\ ().
2216 * Return
2217 * 0 on success, or a negative error in case of failure.
2218 */
2219 #define __BPF_FUNC_MAPPER(FN) \
2220 FN(unspec), \
2221 FN(map_lookup_elem), \
2222 FN(map_update_elem), \
2223 FN(map_delete_elem), \
2224 FN(probe_read), \
2225 FN(ktime_get_ns), \
2226 FN(trace_printk), \
2227 FN(get_prandom_u32), \
2228 FN(get_smp_processor_id), \
2229 FN(skb_store_bytes), \
2230 FN(l3_csum_replace), \
2231 FN(l4_csum_replace), \
2232 FN(tail_call), \
2233 FN(clone_redirect), \
2234 FN(get_current_pid_tgid), \
2235 FN(get_current_uid_gid), \
2236 FN(get_current_comm), \
2237 FN(get_cgroup_classid), \
2238 FN(skb_vlan_push), \
2239 FN(skb_vlan_pop), \
2240 FN(skb_get_tunnel_key), \
2241 FN(skb_set_tunnel_key), \
2242 FN(perf_event_read), \
2243 FN(redirect), \
2244 FN(get_route_realm), \
2245 FN(perf_event_output), \
2246 FN(skb_load_bytes), \
2247 FN(get_stackid), \
2248 FN(csum_diff), \
2249 FN(skb_get_tunnel_opt), \
2250 FN(skb_set_tunnel_opt), \
2251 FN(skb_change_proto), \
2252 FN(skb_change_type), \
2253 FN(skb_under_cgroup), \
2254 FN(get_hash_recalc), \
2255 FN(get_current_task), \
2256 FN(probe_write_user), \
2257 FN(current_task_under_cgroup), \
2258 FN(skb_change_tail), \
2259 FN(skb_pull_data), \
2260 FN(csum_update), \
2261 FN(set_hash_invalid), \
2262 FN(get_numa_node_id), \
2263 FN(skb_change_head), \
2264 FN(xdp_adjust_head), \
2265 FN(probe_read_str), \
2266 FN(get_socket_cookie), \
2267 FN(get_socket_uid), \
2268 FN(set_hash), \
2269 FN(setsockopt), \
2270 FN(skb_adjust_room), \
2271 FN(redirect_map), \
2272 FN(sk_redirect_map), \
2273 FN(sock_map_update), \
2274 FN(xdp_adjust_meta), \
2275 FN(perf_event_read_value), \
2276 FN(perf_prog_read_value), \
2277 FN(getsockopt), \
2278 FN(override_return), \
2279 FN(sock_ops_cb_flags_set), \
2280 FN(msg_redirect_map), \
2281 FN(msg_apply_bytes), \
2282 FN(msg_cork_bytes), \
2283 FN(msg_pull_data), \
2284 FN(bind), \
2285 FN(xdp_adjust_tail), \
2286 FN(skb_get_xfrm_state), \
2287 FN(get_stack), \
2288 FN(skb_load_bytes_relative), \
2289 FN(fib_lookup), \
2290 FN(sock_hash_update), \
2291 FN(msg_redirect_hash), \
2292 FN(sk_redirect_hash), \
2293 FN(lwt_push_encap), \
2294 FN(lwt_seg6_store_bytes), \
2295 FN(lwt_seg6_adjust_srh), \
2296 FN(lwt_seg6_action), \
2297 FN(rc_repeat), \
2298 FN(rc_keydown), \
2299 FN(skb_cgroup_id), \
2300 FN(get_current_cgroup_id), \
2301 FN(get_local_storage), \
2302 FN(sk_select_reuseport), \
2303 FN(skb_ancestor_cgroup_id), \
2304 FN(sk_lookup_tcp), \
2305 FN(sk_lookup_udp), \
2306 FN(sk_release),
2307
2308 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2309 * function eBPF program intends to call
2310 */
2311 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2312 enum bpf_func_id {
2313 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2314 __BPF_FUNC_MAX_ID,
2315 };
2316 #undef __BPF_ENUM_FN
2317
2318 /* All flags used by eBPF helper functions, placed here. */
2319
2320 /* BPF_FUNC_skb_store_bytes flags. */
2321 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2322 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2323
2324 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2325 * First 4 bits are for passing the header field size.
2326 */
2327 #define BPF_F_HDR_FIELD_MASK 0xfULL
2328
2329 /* BPF_FUNC_l4_csum_replace flags. */
2330 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2331 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2332 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2333
2334 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2335 #define BPF_F_INGRESS (1ULL << 0)
2336
2337 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2338 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2339
2340 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2341 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2342 #define BPF_F_USER_STACK (1ULL << 8)
2343 /* flags used by BPF_FUNC_get_stackid only. */
2344 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2345 #define BPF_F_REUSE_STACKID (1ULL << 10)
2346 /* flags used by BPF_FUNC_get_stack only. */
2347 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2348
2349 /* BPF_FUNC_skb_set_tunnel_key flags. */
2350 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2351 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2352 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2353
2354 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2355 * BPF_FUNC_perf_event_read_value flags.
2356 */
2357 #define BPF_F_INDEX_MASK 0xffffffffULL
2358 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2359 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2360 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2361
2362 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2363 enum bpf_adj_room_mode {
2364 BPF_ADJ_ROOM_NET,
2365 };
2366
2367 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2368 enum bpf_hdr_start_off {
2369 BPF_HDR_START_MAC,
2370 BPF_HDR_START_NET,
2371 };
2372
2373 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2374 enum bpf_lwt_encap_mode {
2375 BPF_LWT_ENCAP_SEG6,
2376 BPF_LWT_ENCAP_SEG6_INLINE
2377 };
2378
2379 /* user accessible mirror of in-kernel sk_buff.
2380 * new fields can only be added to the end of this structure
2381 */
2382 struct __sk_buff {
2383 __u32 len;
2384 __u32 pkt_type;
2385 __u32 mark;
2386 __u32 queue_mapping;
2387 __u32 protocol;
2388 __u32 vlan_present;
2389 __u32 vlan_tci;
2390 __u32 vlan_proto;
2391 __u32 priority;
2392 __u32 ingress_ifindex;
2393 __u32 ifindex;
2394 __u32 tc_index;
2395 __u32 cb[5];
2396 __u32 hash;
2397 __u32 tc_classid;
2398 __u32 data;
2399 __u32 data_end;
2400 __u32 napi_id;
2401
2402 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2403 __u32 family;
2404 __u32 remote_ip4; /* Stored in network byte order */
2405 __u32 local_ip4; /* Stored in network byte order */
2406 __u32 remote_ip6[4]; /* Stored in network byte order */
2407 __u32 local_ip6[4]; /* Stored in network byte order */
2408 __u32 remote_port; /* Stored in network byte order */
2409 __u32 local_port; /* stored in host byte order */
2410 /* ... here. */
2411
2412 __u32 data_meta;
2413 struct bpf_flow_keys *flow_keys;
2414 };
2415
2416 struct bpf_tunnel_key {
2417 __u32 tunnel_id;
2418 union {
2419 __u32 remote_ipv4;
2420 __u32 remote_ipv6[4];
2421 };
2422 __u8 tunnel_tos;
2423 __u8 tunnel_ttl;
2424 __u16 tunnel_ext; /* Padding, future use. */
2425 __u32 tunnel_label;
2426 };
2427
2428 /* user accessible mirror of in-kernel xfrm_state.
2429 * new fields can only be added to the end of this structure
2430 */
2431 struct bpf_xfrm_state {
2432 __u32 reqid;
2433 __u32 spi; /* Stored in network byte order */
2434 __u16 family;
2435 __u16 ext; /* Padding, future use. */
2436 union {
2437 __u32 remote_ipv4; /* Stored in network byte order */
2438 __u32 remote_ipv6[4]; /* Stored in network byte order */
2439 };
2440 };
2441
2442 /* Generic BPF return codes which all BPF program types may support.
2443 * The values are binary compatible with their TC_ACT_* counter-part to
2444 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2445 * programs.
2446 *
2447 * XDP is handled seprately, see XDP_*.
2448 */
2449 enum bpf_ret_code {
2450 BPF_OK = 0,
2451 /* 1 reserved */
2452 BPF_DROP = 2,
2453 /* 3-6 reserved */
2454 BPF_REDIRECT = 7,
2455 /* >127 are reserved for prog type specific return codes */
2456 };
2457
2458 struct bpf_sock {
2459 __u32 bound_dev_if;
2460 __u32 family;
2461 __u32 type;
2462 __u32 protocol;
2463 __u32 mark;
2464 __u32 priority;
2465 __u32 src_ip4; /* Allows 1,2,4-byte read.
2466 * Stored in network byte order.
2467 */
2468 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2469 * Stored in network byte order.
2470 */
2471 __u32 src_port; /* Allows 4-byte read.
2472 * Stored in host byte order
2473 */
2474 };
2475
2476 struct bpf_sock_tuple {
2477 union {
2478 struct {
2479 __be32 saddr;
2480 __be32 daddr;
2481 __be16 sport;
2482 __be16 dport;
2483 } ipv4;
2484 struct {
2485 __be32 saddr[4];
2486 __be32 daddr[4];
2487 __be16 sport;
2488 __be16 dport;
2489 } ipv6;
2490 };
2491 };
2492
2493 #define XDP_PACKET_HEADROOM 256
2494
2495 /* User return codes for XDP prog type.
2496 * A valid XDP program must return one of these defined values. All other
2497 * return codes are reserved for future use. Unknown return codes will
2498 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2499 */
2500 enum xdp_action {
2501 XDP_ABORTED = 0,
2502 XDP_DROP,
2503 XDP_PASS,
2504 XDP_TX,
2505 XDP_REDIRECT,
2506 };
2507
2508 /* user accessible metadata for XDP packet hook
2509 * new fields must be added to the end of this structure
2510 */
2511 struct xdp_md {
2512 __u32 data;
2513 __u32 data_end;
2514 __u32 data_meta;
2515 /* Below access go through struct xdp_rxq_info */
2516 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2517 __u32 rx_queue_index; /* rxq->queue_index */
2518 };
2519
2520 enum sk_action {
2521 SK_DROP = 0,
2522 SK_PASS,
2523 };
2524
2525 /* user accessible metadata for SK_MSG packet hook, new fields must
2526 * be added to the end of this structure
2527 */
2528 struct sk_msg_md {
2529 void *data;
2530 void *data_end;
2531
2532 __u32 family;
2533 __u32 remote_ip4; /* Stored in network byte order */
2534 __u32 local_ip4; /* Stored in network byte order */
2535 __u32 remote_ip6[4]; /* Stored in network byte order */
2536 __u32 local_ip6[4]; /* Stored in network byte order */
2537 __u32 remote_port; /* Stored in network byte order */
2538 __u32 local_port; /* stored in host byte order */
2539 };
2540
2541 struct sk_reuseport_md {
2542 /*
2543 * Start of directly accessible data. It begins from
2544 * the tcp/udp header.
2545 */
2546 void *data;
2547 void *data_end; /* End of directly accessible data */
2548 /*
2549 * Total length of packet (starting from the tcp/udp header).
2550 * Note that the directly accessible bytes (data_end - data)
2551 * could be less than this "len". Those bytes could be
2552 * indirectly read by a helper "bpf_skb_load_bytes()".
2553 */
2554 __u32 len;
2555 /*
2556 * Eth protocol in the mac header (network byte order). e.g.
2557 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2558 */
2559 __u32 eth_protocol;
2560 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2561 __u32 bind_inany; /* Is sock bound to an INANY address? */
2562 __u32 hash; /* A hash of the packet 4 tuples */
2563 };
2564
2565 #define BPF_TAG_SIZE 8
2566
2567 struct bpf_prog_info {
2568 __u32 type;
2569 __u32 id;
2570 __u8 tag[BPF_TAG_SIZE];
2571 __u32 jited_prog_len;
2572 __u32 xlated_prog_len;
2573 __aligned_u64 jited_prog_insns;
2574 __aligned_u64 xlated_prog_insns;
2575 __u64 load_time; /* ns since boottime */
2576 __u32 created_by_uid;
2577 __u32 nr_map_ids;
2578 __aligned_u64 map_ids;
2579 char name[BPF_OBJ_NAME_LEN];
2580 __u32 ifindex;
2581 __u32 gpl_compatible:1;
2582 __u64 netns_dev;
2583 __u64 netns_ino;
2584 __u32 nr_jited_ksyms;
2585 __u32 nr_jited_func_lens;
2586 __aligned_u64 jited_ksyms;
2587 __aligned_u64 jited_func_lens;
2588 } __attribute__((aligned(8)));
2589
2590 struct bpf_map_info {
2591 __u32 type;
2592 __u32 id;
2593 __u32 key_size;
2594 __u32 value_size;
2595 __u32 max_entries;
2596 __u32 map_flags;
2597 char name[BPF_OBJ_NAME_LEN];
2598 __u32 ifindex;
2599 __u32 :32;
2600 __u64 netns_dev;
2601 __u64 netns_ino;
2602 __u32 btf_id;
2603 __u32 btf_key_type_id;
2604 __u32 btf_value_type_id;
2605 } __attribute__((aligned(8)));
2606
2607 struct bpf_btf_info {
2608 __aligned_u64 btf;
2609 __u32 btf_size;
2610 __u32 id;
2611 } __attribute__((aligned(8)));
2612
2613 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2614 * by user and intended to be used by socket (e.g. to bind to, depends on
2615 * attach attach type).
2616 */
2617 struct bpf_sock_addr {
2618 __u32 user_family; /* Allows 4-byte read, but no write. */
2619 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2620 * Stored in network byte order.
2621 */
2622 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2623 * Stored in network byte order.
2624 */
2625 __u32 user_port; /* Allows 4-byte read and write.
2626 * Stored in network byte order
2627 */
2628 __u32 family; /* Allows 4-byte read, but no write */
2629 __u32 type; /* Allows 4-byte read, but no write */
2630 __u32 protocol; /* Allows 4-byte read, but no write */
2631 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2632 * Stored in network byte order.
2633 */
2634 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2635 * Stored in network byte order.
2636 */
2637 };
2638
2639 /* User bpf_sock_ops struct to access socket values and specify request ops
2640 * and their replies.
2641 * Some of this fields are in network (bigendian) byte order and may need
2642 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2643 * New fields can only be added at the end of this structure
2644 */
2645 struct bpf_sock_ops {
2646 __u32 op;
2647 union {
2648 __u32 args[4]; /* Optionally passed to bpf program */
2649 __u32 reply; /* Returned by bpf program */
2650 __u32 replylong[4]; /* Optionally returned by bpf prog */
2651 };
2652 __u32 family;
2653 __u32 remote_ip4; /* Stored in network byte order */
2654 __u32 local_ip4; /* Stored in network byte order */
2655 __u32 remote_ip6[4]; /* Stored in network byte order */
2656 __u32 local_ip6[4]; /* Stored in network byte order */
2657 __u32 remote_port; /* Stored in network byte order */
2658 __u32 local_port; /* stored in host byte order */
2659 __u32 is_fullsock; /* Some TCP fields are only valid if
2660 * there is a full socket. If not, the
2661 * fields read as zero.
2662 */
2663 __u32 snd_cwnd;
2664 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2665 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2666 __u32 state;
2667 __u32 rtt_min;
2668 __u32 snd_ssthresh;
2669 __u32 rcv_nxt;
2670 __u32 snd_nxt;
2671 __u32 snd_una;
2672 __u32 mss_cache;
2673 __u32 ecn_flags;
2674 __u32 rate_delivered;
2675 __u32 rate_interval_us;
2676 __u32 packets_out;
2677 __u32 retrans_out;
2678 __u32 total_retrans;
2679 __u32 segs_in;
2680 __u32 data_segs_in;
2681 __u32 segs_out;
2682 __u32 data_segs_out;
2683 __u32 lost_out;
2684 __u32 sacked_out;
2685 __u32 sk_txhash;
2686 __u64 bytes_received;
2687 __u64 bytes_acked;
2688 };
2689
2690 /* Definitions for bpf_sock_ops_cb_flags */
2691 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2692 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2693 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2694 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2695 * supported cb flags
2696 */
2697
2698 /* List of known BPF sock_ops operators.
2699 * New entries can only be added at the end
2700 */
2701 enum {
2702 BPF_SOCK_OPS_VOID,
2703 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2704 * -1 if default value should be used
2705 */
2706 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2707 * window (in packets) or -1 if default
2708 * value should be used
2709 */
2710 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2711 * active connection is initialized
2712 */
2713 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2714 * active connection is
2715 * established
2716 */
2717 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2718 * passive connection is
2719 * established
2720 */
2721 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2722 * needs ECN
2723 */
2724 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2725 * based on the path and may be
2726 * dependent on the congestion control
2727 * algorithm. In general it indicates
2728 * a congestion threshold. RTTs above
2729 * this indicate congestion
2730 */
2731 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2732 * Arg1: value of icsk_retransmits
2733 * Arg2: value of icsk_rto
2734 * Arg3: whether RTO has expired
2735 */
2736 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2737 * Arg1: sequence number of 1st byte
2738 * Arg2: # segments
2739 * Arg3: return value of
2740 * tcp_transmit_skb (0 => success)
2741 */
2742 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2743 * Arg1: old_state
2744 * Arg2: new_state
2745 */
2746 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
2747 * socket transition to LISTEN state.
2748 */
2749 };
2750
2751 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2752 * changes between the TCP and BPF versions. Ideally this should never happen.
2753 * If it does, we need to add code to convert them before calling
2754 * the BPF sock_ops function.
2755 */
2756 enum {
2757 BPF_TCP_ESTABLISHED = 1,
2758 BPF_TCP_SYN_SENT,
2759 BPF_TCP_SYN_RECV,
2760 BPF_TCP_FIN_WAIT1,
2761 BPF_TCP_FIN_WAIT2,
2762 BPF_TCP_TIME_WAIT,
2763 BPF_TCP_CLOSE,
2764 BPF_TCP_CLOSE_WAIT,
2765 BPF_TCP_LAST_ACK,
2766 BPF_TCP_LISTEN,
2767 BPF_TCP_CLOSING, /* Now a valid state */
2768 BPF_TCP_NEW_SYN_RECV,
2769
2770 BPF_TCP_MAX_STATES /* Leave at the end! */
2771 };
2772
2773 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
2774 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
2775
2776 struct bpf_perf_event_value {
2777 __u64 counter;
2778 __u64 enabled;
2779 __u64 running;
2780 };
2781
2782 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2783 #define BPF_DEVCG_ACC_READ (1ULL << 1)
2784 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2785
2786 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2787 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2788
2789 struct bpf_cgroup_dev_ctx {
2790 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2791 __u32 access_type;
2792 __u32 major;
2793 __u32 minor;
2794 };
2795
2796 struct bpf_raw_tracepoint_args {
2797 __u64 args[0];
2798 };
2799
2800 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
2801 * OUTPUT: Do lookup from egress perspective; default is ingress
2802 */
2803 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
2804 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
2805
2806 enum {
2807 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
2808 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
2809 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
2810 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
2811 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
2812 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
2813 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
2814 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
2815 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
2816 };
2817
2818 struct bpf_fib_lookup {
2819 /* input: network family for lookup (AF_INET, AF_INET6)
2820 * output: network family of egress nexthop
2821 */
2822 __u8 family;
2823
2824 /* set if lookup is to consider L4 data - e.g., FIB rules */
2825 __u8 l4_protocol;
2826 __be16 sport;
2827 __be16 dport;
2828
2829 /* total length of packet from network header - used for MTU check */
2830 __u16 tot_len;
2831
2832 /* input: L3 device index for lookup
2833 * output: device index from FIB lookup
2834 */
2835 __u32 ifindex;
2836
2837 union {
2838 /* inputs to lookup */
2839 __u8 tos; /* AF_INET */
2840 __be32 flowinfo; /* AF_INET6, flow_label + priority */
2841
2842 /* output: metric of fib result (IPv4/IPv6 only) */
2843 __u32 rt_metric;
2844 };
2845
2846 union {
2847 __be32 ipv4_src;
2848 __u32 ipv6_src[4]; /* in6_addr; network order */
2849 };
2850
2851 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2852 * network header. output: bpf_fib_lookup sets to gateway address
2853 * if FIB lookup returns gateway route
2854 */
2855 union {
2856 __be32 ipv4_dst;
2857 __u32 ipv6_dst[4]; /* in6_addr; network order */
2858 };
2859
2860 /* output */
2861 __be16 h_vlan_proto;
2862 __be16 h_vlan_TCI;
2863 __u8 smac[6]; /* ETH_ALEN */
2864 __u8 dmac[6]; /* ETH_ALEN */
2865 };
2866
2867 enum bpf_task_fd_type {
2868 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
2869 BPF_FD_TYPE_TRACEPOINT, /* tp name */
2870 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
2871 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
2872 BPF_FD_TYPE_UPROBE, /* filename + offset */
2873 BPF_FD_TYPE_URETPROBE, /* filename + offset */
2874 };
2875
2876 struct bpf_flow_keys {
2877 __u16 nhoff;
2878 __u16 thoff;
2879 __u16 addr_proto; /* ETH_P_* of valid addrs */
2880 __u8 is_frag;
2881 __u8 is_first_frag;
2882 __u8 is_encap;
2883 __u8 ip_proto;
2884 __be16 n_proto;
2885 __be16 sport;
2886 __be16 dport;
2887 union {
2888 struct {
2889 __be32 ipv4_src;
2890 __be32 ipv4_dst;
2891 };
2892 struct {
2893 __u32 ipv6_src[4]; /* in6_addr; network order */
2894 __u32 ipv6_dst[4]; /* in6_addr; network order */
2895 };
2896 };
2897 };
2898
2899 #endif /* _UAPI__LINUX_BPF_H__ */