]> git.proxmox.com Git - mirror_iproute2.git/blob - include/uapi/linux/bpf.h
vdpa: add .gitignore
[mirror_iproute2.git] / include / uapi / linux / bpf.h
1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 */
8 #ifndef __LINUX_BPF_H__
9 #define __LINUX_BPF_H__
10
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
13
14 /* Extended instruction set based on top of classic BPF */
15
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
19
20 /* ld/ldx fields */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */
23 #define BPF_XADD 0xc0 /* exclusive add - legacy name */
24
25 /* alu/jmp fields */
26 #define BPF_MOV 0xb0 /* mov reg to reg */
27 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28
29 /* change endianness of a register */
30 #define BPF_END 0xd0 /* flags for endianness conversion: */
31 #define BPF_TO_LE 0x00 /* convert to little-endian */
32 #define BPF_TO_BE 0x08 /* convert to big-endian */
33 #define BPF_FROM_LE BPF_TO_LE
34 #define BPF_FROM_BE BPF_TO_BE
35
36 /* jmp encodings */
37 #define BPF_JNE 0x50 /* jump != */
38 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
39 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
40 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
41 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
42 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
43 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
44 #define BPF_CALL 0x80 /* function call */
45 #define BPF_EXIT 0x90 /* function return */
46
47 /* atomic op type fields (stored in immediate) */
48 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */
49 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */
50 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */
51
52 /* Register numbers */
53 enum {
54 BPF_REG_0 = 0,
55 BPF_REG_1,
56 BPF_REG_2,
57 BPF_REG_3,
58 BPF_REG_4,
59 BPF_REG_5,
60 BPF_REG_6,
61 BPF_REG_7,
62 BPF_REG_8,
63 BPF_REG_9,
64 BPF_REG_10,
65 __MAX_BPF_REG,
66 };
67
68 /* BPF has 10 general purpose 64-bit registers and stack frame. */
69 #define MAX_BPF_REG __MAX_BPF_REG
70
71 struct bpf_insn {
72 __u8 code; /* opcode */
73 __u8 dst_reg:4; /* dest register */
74 __u8 src_reg:4; /* source register */
75 __s16 off; /* signed offset */
76 __s32 imm; /* signed immediate constant */
77 };
78
79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
80 struct bpf_lpm_trie_key {
81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
82 __u8 data[0]; /* Arbitrary size */
83 };
84
85 struct bpf_cgroup_storage_key {
86 __u64 cgroup_inode_id; /* cgroup inode id */
87 __u32 attach_type; /* program attach type */
88 };
89
90 union bpf_iter_link_info {
91 struct {
92 __u32 map_fd;
93 } map;
94 };
95
96 /* BPF syscall commands, see bpf(2) man-page for details. */
97 enum bpf_cmd {
98 BPF_MAP_CREATE,
99 BPF_MAP_LOOKUP_ELEM,
100 BPF_MAP_UPDATE_ELEM,
101 BPF_MAP_DELETE_ELEM,
102 BPF_MAP_GET_NEXT_KEY,
103 BPF_PROG_LOAD,
104 BPF_OBJ_PIN,
105 BPF_OBJ_GET,
106 BPF_PROG_ATTACH,
107 BPF_PROG_DETACH,
108 BPF_PROG_TEST_RUN,
109 BPF_PROG_GET_NEXT_ID,
110 BPF_MAP_GET_NEXT_ID,
111 BPF_PROG_GET_FD_BY_ID,
112 BPF_MAP_GET_FD_BY_ID,
113 BPF_OBJ_GET_INFO_BY_FD,
114 BPF_PROG_QUERY,
115 BPF_RAW_TRACEPOINT_OPEN,
116 BPF_BTF_LOAD,
117 BPF_BTF_GET_FD_BY_ID,
118 BPF_TASK_FD_QUERY,
119 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
120 BPF_MAP_FREEZE,
121 BPF_BTF_GET_NEXT_ID,
122 BPF_MAP_LOOKUP_BATCH,
123 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
124 BPF_MAP_UPDATE_BATCH,
125 BPF_MAP_DELETE_BATCH,
126 BPF_LINK_CREATE,
127 BPF_LINK_UPDATE,
128 BPF_LINK_GET_FD_BY_ID,
129 BPF_LINK_GET_NEXT_ID,
130 BPF_ENABLE_STATS,
131 BPF_ITER_CREATE,
132 BPF_LINK_DETACH,
133 BPF_PROG_BIND_MAP,
134 };
135
136 enum bpf_map_type {
137 BPF_MAP_TYPE_UNSPEC,
138 BPF_MAP_TYPE_HASH,
139 BPF_MAP_TYPE_ARRAY,
140 BPF_MAP_TYPE_PROG_ARRAY,
141 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
142 BPF_MAP_TYPE_PERCPU_HASH,
143 BPF_MAP_TYPE_PERCPU_ARRAY,
144 BPF_MAP_TYPE_STACK_TRACE,
145 BPF_MAP_TYPE_CGROUP_ARRAY,
146 BPF_MAP_TYPE_LRU_HASH,
147 BPF_MAP_TYPE_LRU_PERCPU_HASH,
148 BPF_MAP_TYPE_LPM_TRIE,
149 BPF_MAP_TYPE_ARRAY_OF_MAPS,
150 BPF_MAP_TYPE_HASH_OF_MAPS,
151 BPF_MAP_TYPE_DEVMAP,
152 BPF_MAP_TYPE_SOCKMAP,
153 BPF_MAP_TYPE_CPUMAP,
154 BPF_MAP_TYPE_XSKMAP,
155 BPF_MAP_TYPE_SOCKHASH,
156 BPF_MAP_TYPE_CGROUP_STORAGE,
157 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
158 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
159 BPF_MAP_TYPE_QUEUE,
160 BPF_MAP_TYPE_STACK,
161 BPF_MAP_TYPE_SK_STORAGE,
162 BPF_MAP_TYPE_DEVMAP_HASH,
163 BPF_MAP_TYPE_STRUCT_OPS,
164 BPF_MAP_TYPE_RINGBUF,
165 BPF_MAP_TYPE_INODE_STORAGE,
166 BPF_MAP_TYPE_TASK_STORAGE,
167 };
168
169 /* Note that tracing related programs such as
170 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
171 * are not subject to a stable API since kernel internal data
172 * structures can change from release to release and may
173 * therefore break existing tracing BPF programs. Tracing BPF
174 * programs correspond to /a/ specific kernel which is to be
175 * analyzed, and not /a/ specific kernel /and/ all future ones.
176 */
177 enum bpf_prog_type {
178 BPF_PROG_TYPE_UNSPEC,
179 BPF_PROG_TYPE_SOCKET_FILTER,
180 BPF_PROG_TYPE_KPROBE,
181 BPF_PROG_TYPE_SCHED_CLS,
182 BPF_PROG_TYPE_SCHED_ACT,
183 BPF_PROG_TYPE_TRACEPOINT,
184 BPF_PROG_TYPE_XDP,
185 BPF_PROG_TYPE_PERF_EVENT,
186 BPF_PROG_TYPE_CGROUP_SKB,
187 BPF_PROG_TYPE_CGROUP_SOCK,
188 BPF_PROG_TYPE_LWT_IN,
189 BPF_PROG_TYPE_LWT_OUT,
190 BPF_PROG_TYPE_LWT_XMIT,
191 BPF_PROG_TYPE_SOCK_OPS,
192 BPF_PROG_TYPE_SK_SKB,
193 BPF_PROG_TYPE_CGROUP_DEVICE,
194 BPF_PROG_TYPE_SK_MSG,
195 BPF_PROG_TYPE_RAW_TRACEPOINT,
196 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
197 BPF_PROG_TYPE_LWT_SEG6LOCAL,
198 BPF_PROG_TYPE_LIRC_MODE2,
199 BPF_PROG_TYPE_SK_REUSEPORT,
200 BPF_PROG_TYPE_FLOW_DISSECTOR,
201 BPF_PROG_TYPE_CGROUP_SYSCTL,
202 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
203 BPF_PROG_TYPE_CGROUP_SOCKOPT,
204 BPF_PROG_TYPE_TRACING,
205 BPF_PROG_TYPE_STRUCT_OPS,
206 BPF_PROG_TYPE_EXT,
207 BPF_PROG_TYPE_LSM,
208 BPF_PROG_TYPE_SK_LOOKUP,
209 };
210
211 enum bpf_attach_type {
212 BPF_CGROUP_INET_INGRESS,
213 BPF_CGROUP_INET_EGRESS,
214 BPF_CGROUP_INET_SOCK_CREATE,
215 BPF_CGROUP_SOCK_OPS,
216 BPF_SK_SKB_STREAM_PARSER,
217 BPF_SK_SKB_STREAM_VERDICT,
218 BPF_CGROUP_DEVICE,
219 BPF_SK_MSG_VERDICT,
220 BPF_CGROUP_INET4_BIND,
221 BPF_CGROUP_INET6_BIND,
222 BPF_CGROUP_INET4_CONNECT,
223 BPF_CGROUP_INET6_CONNECT,
224 BPF_CGROUP_INET4_POST_BIND,
225 BPF_CGROUP_INET6_POST_BIND,
226 BPF_CGROUP_UDP4_SENDMSG,
227 BPF_CGROUP_UDP6_SENDMSG,
228 BPF_LIRC_MODE2,
229 BPF_FLOW_DISSECTOR,
230 BPF_CGROUP_SYSCTL,
231 BPF_CGROUP_UDP4_RECVMSG,
232 BPF_CGROUP_UDP6_RECVMSG,
233 BPF_CGROUP_GETSOCKOPT,
234 BPF_CGROUP_SETSOCKOPT,
235 BPF_TRACE_RAW_TP,
236 BPF_TRACE_FENTRY,
237 BPF_TRACE_FEXIT,
238 BPF_MODIFY_RETURN,
239 BPF_LSM_MAC,
240 BPF_TRACE_ITER,
241 BPF_CGROUP_INET4_GETPEERNAME,
242 BPF_CGROUP_INET6_GETPEERNAME,
243 BPF_CGROUP_INET4_GETSOCKNAME,
244 BPF_CGROUP_INET6_GETSOCKNAME,
245 BPF_XDP_DEVMAP,
246 BPF_CGROUP_INET_SOCK_RELEASE,
247 BPF_XDP_CPUMAP,
248 BPF_SK_LOOKUP,
249 BPF_XDP,
250 __MAX_BPF_ATTACH_TYPE
251 };
252
253 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
254
255 enum bpf_link_type {
256 BPF_LINK_TYPE_UNSPEC = 0,
257 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
258 BPF_LINK_TYPE_TRACING = 2,
259 BPF_LINK_TYPE_CGROUP = 3,
260 BPF_LINK_TYPE_ITER = 4,
261 BPF_LINK_TYPE_NETNS = 5,
262 BPF_LINK_TYPE_XDP = 6,
263
264 MAX_BPF_LINK_TYPE,
265 };
266
267 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
268 *
269 * NONE(default): No further bpf programs allowed in the subtree.
270 *
271 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
272 * the program in this cgroup yields to sub-cgroup program.
273 *
274 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
275 * that cgroup program gets run in addition to the program in this cgroup.
276 *
277 * Only one program is allowed to be attached to a cgroup with
278 * NONE or BPF_F_ALLOW_OVERRIDE flag.
279 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
280 * release old program and attach the new one. Attach flags has to match.
281 *
282 * Multiple programs are allowed to be attached to a cgroup with
283 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
284 * (those that were attached first, run first)
285 * The programs of sub-cgroup are executed first, then programs of
286 * this cgroup and then programs of parent cgroup.
287 * When children program makes decision (like picking TCP CA or sock bind)
288 * parent program has a chance to override it.
289 *
290 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
291 * programs for a cgroup. Though it's possible to replace an old program at
292 * any position by also specifying BPF_F_REPLACE flag and position itself in
293 * replace_bpf_fd attribute. Old program at this position will be released.
294 *
295 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
296 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
297 * Ex1:
298 * cgrp1 (MULTI progs A, B) ->
299 * cgrp2 (OVERRIDE prog C) ->
300 * cgrp3 (MULTI prog D) ->
301 * cgrp4 (OVERRIDE prog E) ->
302 * cgrp5 (NONE prog F)
303 * the event in cgrp5 triggers execution of F,D,A,B in that order.
304 * if prog F is detached, the execution is E,D,A,B
305 * if prog F and D are detached, the execution is E,A,B
306 * if prog F, E and D are detached, the execution is C,A,B
307 *
308 * All eligible programs are executed regardless of return code from
309 * earlier programs.
310 */
311 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
312 #define BPF_F_ALLOW_MULTI (1U << 1)
313 #define BPF_F_REPLACE (1U << 2)
314
315 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
316 * verifier will perform strict alignment checking as if the kernel
317 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
318 * and NET_IP_ALIGN defined to 2.
319 */
320 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
321
322 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
323 * verifier will allow any alignment whatsoever. On platforms
324 * with strict alignment requirements for loads ands stores (such
325 * as sparc and mips) the verifier validates that all loads and
326 * stores provably follow this requirement. This flag turns that
327 * checking and enforcement off.
328 *
329 * It is mostly used for testing when we want to validate the
330 * context and memory access aspects of the verifier, but because
331 * of an unaligned access the alignment check would trigger before
332 * the one we are interested in.
333 */
334 #define BPF_F_ANY_ALIGNMENT (1U << 1)
335
336 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
337 * Verifier does sub-register def/use analysis and identifies instructions whose
338 * def only matters for low 32-bit, high 32-bit is never referenced later
339 * through implicit zero extension. Therefore verifier notifies JIT back-ends
340 * that it is safe to ignore clearing high 32-bit for these instructions. This
341 * saves some back-ends a lot of code-gen. However such optimization is not
342 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
343 * hence hasn't used verifier's analysis result. But, we really want to have a
344 * way to be able to verify the correctness of the described optimization on
345 * x86_64 on which testsuites are frequently exercised.
346 *
347 * So, this flag is introduced. Once it is set, verifier will randomize high
348 * 32-bit for those instructions who has been identified as safe to ignore them.
349 * Then, if verifier is not doing correct analysis, such randomization will
350 * regress tests to expose bugs.
351 */
352 #define BPF_F_TEST_RND_HI32 (1U << 2)
353
354 /* The verifier internal test flag. Behavior is undefined */
355 #define BPF_F_TEST_STATE_FREQ (1U << 3)
356
357 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
358 * restrict map and helper usage for such programs. Sleepable BPF programs can
359 * only be attached to hooks where kernel execution context allows sleeping.
360 * Such programs are allowed to use helpers that may sleep like
361 * bpf_copy_from_user().
362 */
363 #define BPF_F_SLEEPABLE (1U << 4)
364
365 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
366 * the following extensions:
367 *
368 * insn[0].src_reg: BPF_PSEUDO_MAP_FD
369 * insn[0].imm: map fd
370 * insn[1].imm: 0
371 * insn[0].off: 0
372 * insn[1].off: 0
373 * ldimm64 rewrite: address of map
374 * verifier type: CONST_PTR_TO_MAP
375 */
376 #define BPF_PSEUDO_MAP_FD 1
377 /* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE
378 * insn[0].imm: map fd
379 * insn[1].imm: offset into value
380 * insn[0].off: 0
381 * insn[1].off: 0
382 * ldimm64 rewrite: address of map[0]+offset
383 * verifier type: PTR_TO_MAP_VALUE
384 */
385 #define BPF_PSEUDO_MAP_VALUE 2
386 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID
387 * insn[0].imm: kernel btd id of VAR
388 * insn[1].imm: 0
389 * insn[0].off: 0
390 * insn[1].off: 0
391 * ldimm64 rewrite: address of the kernel variable
392 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
393 * is struct/union.
394 */
395 #define BPF_PSEUDO_BTF_ID 3
396
397 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
398 * offset to another bpf function
399 */
400 #define BPF_PSEUDO_CALL 1
401
402 /* flags for BPF_MAP_UPDATE_ELEM command */
403 enum {
404 BPF_ANY = 0, /* create new element or update existing */
405 BPF_NOEXIST = 1, /* create new element if it didn't exist */
406 BPF_EXIST = 2, /* update existing element */
407 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
408 };
409
410 /* flags for BPF_MAP_CREATE command */
411 enum {
412 BPF_F_NO_PREALLOC = (1U << 0),
413 /* Instead of having one common LRU list in the
414 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
415 * which can scale and perform better.
416 * Note, the LRU nodes (including free nodes) cannot be moved
417 * across different LRU lists.
418 */
419 BPF_F_NO_COMMON_LRU = (1U << 1),
420 /* Specify numa node during map creation */
421 BPF_F_NUMA_NODE = (1U << 2),
422
423 /* Flags for accessing BPF object from syscall side. */
424 BPF_F_RDONLY = (1U << 3),
425 BPF_F_WRONLY = (1U << 4),
426
427 /* Flag for stack_map, store build_id+offset instead of pointer */
428 BPF_F_STACK_BUILD_ID = (1U << 5),
429
430 /* Zero-initialize hash function seed. This should only be used for testing. */
431 BPF_F_ZERO_SEED = (1U << 6),
432
433 /* Flags for accessing BPF object from program side. */
434 BPF_F_RDONLY_PROG = (1U << 7),
435 BPF_F_WRONLY_PROG = (1U << 8),
436
437 /* Clone map from listener for newly accepted socket */
438 BPF_F_CLONE = (1U << 9),
439
440 /* Enable memory-mapping BPF map */
441 BPF_F_MMAPABLE = (1U << 10),
442
443 /* Share perf_event among processes */
444 BPF_F_PRESERVE_ELEMS = (1U << 11),
445
446 /* Create a map that is suitable to be an inner map with dynamic max entries */
447 BPF_F_INNER_MAP = (1U << 12),
448 };
449
450 /* Flags for BPF_PROG_QUERY. */
451
452 /* Query effective (directly attached + inherited from ancestor cgroups)
453 * programs that will be executed for events within a cgroup.
454 * attach_flags with this flag are returned only for directly attached programs.
455 */
456 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
457
458 /* Flags for BPF_PROG_TEST_RUN */
459
460 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
461 #define BPF_F_TEST_RUN_ON_CPU (1U << 0)
462
463 /* type for BPF_ENABLE_STATS */
464 enum bpf_stats_type {
465 /* enabled run_time_ns and run_cnt */
466 BPF_STATS_RUN_TIME = 0,
467 };
468
469 enum bpf_stack_build_id_status {
470 /* user space need an empty entry to identify end of a trace */
471 BPF_STACK_BUILD_ID_EMPTY = 0,
472 /* with valid build_id and offset */
473 BPF_STACK_BUILD_ID_VALID = 1,
474 /* couldn't get build_id, fallback to ip */
475 BPF_STACK_BUILD_ID_IP = 2,
476 };
477
478 #define BPF_BUILD_ID_SIZE 20
479 struct bpf_stack_build_id {
480 __s32 status;
481 unsigned char build_id[BPF_BUILD_ID_SIZE];
482 union {
483 __u64 offset;
484 __u64 ip;
485 };
486 };
487
488 #define BPF_OBJ_NAME_LEN 16U
489
490 union bpf_attr {
491 struct { /* anonymous struct used by BPF_MAP_CREATE command */
492 __u32 map_type; /* one of enum bpf_map_type */
493 __u32 key_size; /* size of key in bytes */
494 __u32 value_size; /* size of value in bytes */
495 __u32 max_entries; /* max number of entries in a map */
496 __u32 map_flags; /* BPF_MAP_CREATE related
497 * flags defined above.
498 */
499 __u32 inner_map_fd; /* fd pointing to the inner map */
500 __u32 numa_node; /* numa node (effective only if
501 * BPF_F_NUMA_NODE is set).
502 */
503 char map_name[BPF_OBJ_NAME_LEN];
504 __u32 map_ifindex; /* ifindex of netdev to create on */
505 __u32 btf_fd; /* fd pointing to a BTF type data */
506 __u32 btf_key_type_id; /* BTF type_id of the key */
507 __u32 btf_value_type_id; /* BTF type_id of the value */
508 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
509 * struct stored as the
510 * map value
511 */
512 };
513
514 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
515 __u32 map_fd;
516 __aligned_u64 key;
517 union {
518 __aligned_u64 value;
519 __aligned_u64 next_key;
520 };
521 __u64 flags;
522 };
523
524 struct { /* struct used by BPF_MAP_*_BATCH commands */
525 __aligned_u64 in_batch; /* start batch,
526 * NULL to start from beginning
527 */
528 __aligned_u64 out_batch; /* output: next start batch */
529 __aligned_u64 keys;
530 __aligned_u64 values;
531 __u32 count; /* input/output:
532 * input: # of key/value
533 * elements
534 * output: # of filled elements
535 */
536 __u32 map_fd;
537 __u64 elem_flags;
538 __u64 flags;
539 } batch;
540
541 struct { /* anonymous struct used by BPF_PROG_LOAD command */
542 __u32 prog_type; /* one of enum bpf_prog_type */
543 __u32 insn_cnt;
544 __aligned_u64 insns;
545 __aligned_u64 license;
546 __u32 log_level; /* verbosity level of verifier */
547 __u32 log_size; /* size of user buffer */
548 __aligned_u64 log_buf; /* user supplied buffer */
549 __u32 kern_version; /* not used */
550 __u32 prog_flags;
551 char prog_name[BPF_OBJ_NAME_LEN];
552 __u32 prog_ifindex; /* ifindex of netdev to prep for */
553 /* For some prog types expected attach type must be known at
554 * load time to verify attach type specific parts of prog
555 * (context accesses, allowed helpers, etc).
556 */
557 __u32 expected_attach_type;
558 __u32 prog_btf_fd; /* fd pointing to BTF type data */
559 __u32 func_info_rec_size; /* userspace bpf_func_info size */
560 __aligned_u64 func_info; /* func info */
561 __u32 func_info_cnt; /* number of bpf_func_info records */
562 __u32 line_info_rec_size; /* userspace bpf_line_info size */
563 __aligned_u64 line_info; /* line info */
564 __u32 line_info_cnt; /* number of bpf_line_info records */
565 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
566 union {
567 /* valid prog_fd to attach to bpf prog */
568 __u32 attach_prog_fd;
569 /* or valid module BTF object fd or 0 to attach to vmlinux */
570 __u32 attach_btf_obj_fd;
571 };
572 };
573
574 struct { /* anonymous struct used by BPF_OBJ_* commands */
575 __aligned_u64 pathname;
576 __u32 bpf_fd;
577 __u32 file_flags;
578 };
579
580 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
581 __u32 target_fd; /* container object to attach to */
582 __u32 attach_bpf_fd; /* eBPF program to attach */
583 __u32 attach_type;
584 __u32 attach_flags;
585 __u32 replace_bpf_fd; /* previously attached eBPF
586 * program to replace if
587 * BPF_F_REPLACE is used
588 */
589 };
590
591 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
592 __u32 prog_fd;
593 __u32 retval;
594 __u32 data_size_in; /* input: len of data_in */
595 __u32 data_size_out; /* input/output: len of data_out
596 * returns ENOSPC if data_out
597 * is too small.
598 */
599 __aligned_u64 data_in;
600 __aligned_u64 data_out;
601 __u32 repeat;
602 __u32 duration;
603 __u32 ctx_size_in; /* input: len of ctx_in */
604 __u32 ctx_size_out; /* input/output: len of ctx_out
605 * returns ENOSPC if ctx_out
606 * is too small.
607 */
608 __aligned_u64 ctx_in;
609 __aligned_u64 ctx_out;
610 __u32 flags;
611 __u32 cpu;
612 } test;
613
614 struct { /* anonymous struct used by BPF_*_GET_*_ID */
615 union {
616 __u32 start_id;
617 __u32 prog_id;
618 __u32 map_id;
619 __u32 btf_id;
620 __u32 link_id;
621 };
622 __u32 next_id;
623 __u32 open_flags;
624 };
625
626 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
627 __u32 bpf_fd;
628 __u32 info_len;
629 __aligned_u64 info;
630 } info;
631
632 struct { /* anonymous struct used by BPF_PROG_QUERY command */
633 __u32 target_fd; /* container object to query */
634 __u32 attach_type;
635 __u32 query_flags;
636 __u32 attach_flags;
637 __aligned_u64 prog_ids;
638 __u32 prog_cnt;
639 } query;
640
641 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
642 __u64 name;
643 __u32 prog_fd;
644 } raw_tracepoint;
645
646 struct { /* anonymous struct for BPF_BTF_LOAD */
647 __aligned_u64 btf;
648 __aligned_u64 btf_log_buf;
649 __u32 btf_size;
650 __u32 btf_log_size;
651 __u32 btf_log_level;
652 };
653
654 struct {
655 __u32 pid; /* input: pid */
656 __u32 fd; /* input: fd */
657 __u32 flags; /* input: flags */
658 __u32 buf_len; /* input/output: buf len */
659 __aligned_u64 buf; /* input/output:
660 * tp_name for tracepoint
661 * symbol for kprobe
662 * filename for uprobe
663 */
664 __u32 prog_id; /* output: prod_id */
665 __u32 fd_type; /* output: BPF_FD_TYPE_* */
666 __u64 probe_offset; /* output: probe_offset */
667 __u64 probe_addr; /* output: probe_addr */
668 } task_fd_query;
669
670 struct { /* struct used by BPF_LINK_CREATE command */
671 __u32 prog_fd; /* eBPF program to attach */
672 union {
673 __u32 target_fd; /* object to attach to */
674 __u32 target_ifindex; /* target ifindex */
675 };
676 __u32 attach_type; /* attach type */
677 __u32 flags; /* extra flags */
678 union {
679 __u32 target_btf_id; /* btf_id of target to attach to */
680 struct {
681 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
682 __u32 iter_info_len; /* iter_info length */
683 };
684 };
685 } link_create;
686
687 struct { /* struct used by BPF_LINK_UPDATE command */
688 __u32 link_fd; /* link fd */
689 /* new program fd to update link with */
690 __u32 new_prog_fd;
691 __u32 flags; /* extra flags */
692 /* expected link's program fd; is specified only if
693 * BPF_F_REPLACE flag is set in flags */
694 __u32 old_prog_fd;
695 } link_update;
696
697 struct {
698 __u32 link_fd;
699 } link_detach;
700
701 struct { /* struct used by BPF_ENABLE_STATS command */
702 __u32 type;
703 } enable_stats;
704
705 struct { /* struct used by BPF_ITER_CREATE command */
706 __u32 link_fd;
707 __u32 flags;
708 } iter_create;
709
710 struct { /* struct used by BPF_PROG_BIND_MAP command */
711 __u32 prog_fd;
712 __u32 map_fd;
713 __u32 flags; /* extra flags */
714 } prog_bind_map;
715
716 } __attribute__((aligned(8)));
717
718 /* The description below is an attempt at providing documentation to eBPF
719 * developers about the multiple available eBPF helper functions. It can be
720 * parsed and used to produce a manual page. The workflow is the following,
721 * and requires the rst2man utility:
722 *
723 * $ ./scripts/bpf_helpers_doc.py \
724 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
725 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
726 * $ man /tmp/bpf-helpers.7
727 *
728 * Note that in order to produce this external documentation, some RST
729 * formatting is used in the descriptions to get "bold" and "italics" in
730 * manual pages. Also note that the few trailing white spaces are
731 * intentional, removing them would break paragraphs for rst2man.
732 *
733 * Start of BPF helper function descriptions:
734 *
735 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
736 * Description
737 * Perform a lookup in *map* for an entry associated to *key*.
738 * Return
739 * Map value associated to *key*, or **NULL** if no entry was
740 * found.
741 *
742 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
743 * Description
744 * Add or update the value of the entry associated to *key* in
745 * *map* with *value*. *flags* is one of:
746 *
747 * **BPF_NOEXIST**
748 * The entry for *key* must not exist in the map.
749 * **BPF_EXIST**
750 * The entry for *key* must already exist in the map.
751 * **BPF_ANY**
752 * No condition on the existence of the entry for *key*.
753 *
754 * Flag value **BPF_NOEXIST** cannot be used for maps of types
755 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
756 * elements always exist), the helper would return an error.
757 * Return
758 * 0 on success, or a negative error in case of failure.
759 *
760 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
761 * Description
762 * Delete entry with *key* from *map*.
763 * Return
764 * 0 on success, or a negative error in case of failure.
765 *
766 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
767 * Description
768 * For tracing programs, safely attempt to read *size* bytes from
769 * kernel space address *unsafe_ptr* and store the data in *dst*.
770 *
771 * Generally, use **bpf_probe_read_user**\ () or
772 * **bpf_probe_read_kernel**\ () instead.
773 * Return
774 * 0 on success, or a negative error in case of failure.
775 *
776 * u64 bpf_ktime_get_ns(void)
777 * Description
778 * Return the time elapsed since system boot, in nanoseconds.
779 * Does not include time the system was suspended.
780 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
781 * Return
782 * Current *ktime*.
783 *
784 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
785 * Description
786 * This helper is a "printk()-like" facility for debugging. It
787 * prints a message defined by format *fmt* (of size *fmt_size*)
788 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
789 * available. It can take up to three additional **u64**
790 * arguments (as an eBPF helpers, the total number of arguments is
791 * limited to five).
792 *
793 * Each time the helper is called, it appends a line to the trace.
794 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
795 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
796 * The format of the trace is customizable, and the exact output
797 * one will get depends on the options set in
798 * *\/sys/kernel/debug/tracing/trace_options* (see also the
799 * *README* file under the same directory). However, it usually
800 * defaults to something like:
801 *
802 * ::
803 *
804 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
805 *
806 * In the above:
807 *
808 * * ``telnet`` is the name of the current task.
809 * * ``470`` is the PID of the current task.
810 * * ``001`` is the CPU number on which the task is
811 * running.
812 * * In ``.N..``, each character refers to a set of
813 * options (whether irqs are enabled, scheduling
814 * options, whether hard/softirqs are running, level of
815 * preempt_disabled respectively). **N** means that
816 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
817 * are set.
818 * * ``419421.045894`` is a timestamp.
819 * * ``0x00000001`` is a fake value used by BPF for the
820 * instruction pointer register.
821 * * ``<formatted msg>`` is the message formatted with
822 * *fmt*.
823 *
824 * The conversion specifiers supported by *fmt* are similar, but
825 * more limited than for printk(). They are **%d**, **%i**,
826 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
827 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
828 * of field, padding with zeroes, etc.) is available, and the
829 * helper will return **-EINVAL** (but print nothing) if it
830 * encounters an unknown specifier.
831 *
832 * Also, note that **bpf_trace_printk**\ () is slow, and should
833 * only be used for debugging purposes. For this reason, a notice
834 * block (spanning several lines) is printed to kernel logs and
835 * states that the helper should not be used "for production use"
836 * the first time this helper is used (or more precisely, when
837 * **trace_printk**\ () buffers are allocated). For passing values
838 * to user space, perf events should be preferred.
839 * Return
840 * The number of bytes written to the buffer, or a negative error
841 * in case of failure.
842 *
843 * u32 bpf_get_prandom_u32(void)
844 * Description
845 * Get a pseudo-random number.
846 *
847 * From a security point of view, this helper uses its own
848 * pseudo-random internal state, and cannot be used to infer the
849 * seed of other random functions in the kernel. However, it is
850 * essential to note that the generator used by the helper is not
851 * cryptographically secure.
852 * Return
853 * A random 32-bit unsigned value.
854 *
855 * u32 bpf_get_smp_processor_id(void)
856 * Description
857 * Get the SMP (symmetric multiprocessing) processor id. Note that
858 * all programs run with preemption disabled, which means that the
859 * SMP processor id is stable during all the execution of the
860 * program.
861 * Return
862 * The SMP id of the processor running the program.
863 *
864 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
865 * Description
866 * Store *len* bytes from address *from* into the packet
867 * associated to *skb*, at *offset*. *flags* are a combination of
868 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
869 * checksum for the packet after storing the bytes) and
870 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
871 * **->swhash** and *skb*\ **->l4hash** to 0).
872 *
873 * A call to this helper is susceptible to change the underlying
874 * packet buffer. Therefore, at load time, all checks on pointers
875 * previously done by the verifier are invalidated and must be
876 * performed again, if the helper is used in combination with
877 * direct packet access.
878 * Return
879 * 0 on success, or a negative error in case of failure.
880 *
881 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
882 * Description
883 * Recompute the layer 3 (e.g. IP) checksum for the packet
884 * associated to *skb*. Computation is incremental, so the helper
885 * must know the former value of the header field that was
886 * modified (*from*), the new value of this field (*to*), and the
887 * number of bytes (2 or 4) for this field, stored in *size*.
888 * Alternatively, it is possible to store the difference between
889 * the previous and the new values of the header field in *to*, by
890 * setting *from* and *size* to 0. For both methods, *offset*
891 * indicates the location of the IP checksum within the packet.
892 *
893 * This helper works in combination with **bpf_csum_diff**\ (),
894 * which does not update the checksum in-place, but offers more
895 * flexibility and can handle sizes larger than 2 or 4 for the
896 * checksum to update.
897 *
898 * A call to this helper is susceptible to change the underlying
899 * packet buffer. Therefore, at load time, all checks on pointers
900 * previously done by the verifier are invalidated and must be
901 * performed again, if the helper is used in combination with
902 * direct packet access.
903 * Return
904 * 0 on success, or a negative error in case of failure.
905 *
906 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
907 * Description
908 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
909 * packet associated to *skb*. Computation is incremental, so the
910 * helper must know the former value of the header field that was
911 * modified (*from*), the new value of this field (*to*), and the
912 * number of bytes (2 or 4) for this field, stored on the lowest
913 * four bits of *flags*. Alternatively, it is possible to store
914 * the difference between the previous and the new values of the
915 * header field in *to*, by setting *from* and the four lowest
916 * bits of *flags* to 0. For both methods, *offset* indicates the
917 * location of the IP checksum within the packet. In addition to
918 * the size of the field, *flags* can be added (bitwise OR) actual
919 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
920 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
921 * for updates resulting in a null checksum the value is set to
922 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
923 * the checksum is to be computed against a pseudo-header.
924 *
925 * This helper works in combination with **bpf_csum_diff**\ (),
926 * which does not update the checksum in-place, but offers more
927 * flexibility and can handle sizes larger than 2 or 4 for the
928 * checksum to update.
929 *
930 * A call to this helper is susceptible to change the underlying
931 * packet buffer. Therefore, at load time, all checks on pointers
932 * previously done by the verifier are invalidated and must be
933 * performed again, if the helper is used in combination with
934 * direct packet access.
935 * Return
936 * 0 on success, or a negative error in case of failure.
937 *
938 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
939 * Description
940 * This special helper is used to trigger a "tail call", or in
941 * other words, to jump into another eBPF program. The same stack
942 * frame is used (but values on stack and in registers for the
943 * caller are not accessible to the callee). This mechanism allows
944 * for program chaining, either for raising the maximum number of
945 * available eBPF instructions, or to execute given programs in
946 * conditional blocks. For security reasons, there is an upper
947 * limit to the number of successive tail calls that can be
948 * performed.
949 *
950 * Upon call of this helper, the program attempts to jump into a
951 * program referenced at index *index* in *prog_array_map*, a
952 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
953 * *ctx*, a pointer to the context.
954 *
955 * If the call succeeds, the kernel immediately runs the first
956 * instruction of the new program. This is not a function call,
957 * and it never returns to the previous program. If the call
958 * fails, then the helper has no effect, and the caller continues
959 * to run its subsequent instructions. A call can fail if the
960 * destination program for the jump does not exist (i.e. *index*
961 * is superior to the number of entries in *prog_array_map*), or
962 * if the maximum number of tail calls has been reached for this
963 * chain of programs. This limit is defined in the kernel by the
964 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
965 * which is currently set to 32.
966 * Return
967 * 0 on success, or a negative error in case of failure.
968 *
969 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
970 * Description
971 * Clone and redirect the packet associated to *skb* to another
972 * net device of index *ifindex*. Both ingress and egress
973 * interfaces can be used for redirection. The **BPF_F_INGRESS**
974 * value in *flags* is used to make the distinction (ingress path
975 * is selected if the flag is present, egress path otherwise).
976 * This is the only flag supported for now.
977 *
978 * In comparison with **bpf_redirect**\ () helper,
979 * **bpf_clone_redirect**\ () has the associated cost of
980 * duplicating the packet buffer, but this can be executed out of
981 * the eBPF program. Conversely, **bpf_redirect**\ () is more
982 * efficient, but it is handled through an action code where the
983 * redirection happens only after the eBPF program has returned.
984 *
985 * A call to this helper is susceptible to change the underlying
986 * packet buffer. Therefore, at load time, all checks on pointers
987 * previously done by the verifier are invalidated and must be
988 * performed again, if the helper is used in combination with
989 * direct packet access.
990 * Return
991 * 0 on success, or a negative error in case of failure.
992 *
993 * u64 bpf_get_current_pid_tgid(void)
994 * Return
995 * A 64-bit integer containing the current tgid and pid, and
996 * created as such:
997 * *current_task*\ **->tgid << 32 \|**
998 * *current_task*\ **->pid**.
999 *
1000 * u64 bpf_get_current_uid_gid(void)
1001 * Return
1002 * A 64-bit integer containing the current GID and UID, and
1003 * created as such: *current_gid* **<< 32 \|** *current_uid*.
1004 *
1005 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
1006 * Description
1007 * Copy the **comm** attribute of the current task into *buf* of
1008 * *size_of_buf*. The **comm** attribute contains the name of
1009 * the executable (excluding the path) for the current task. The
1010 * *size_of_buf* must be strictly positive. On success, the
1011 * helper makes sure that the *buf* is NUL-terminated. On failure,
1012 * it is filled with zeroes.
1013 * Return
1014 * 0 on success, or a negative error in case of failure.
1015 *
1016 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1017 * Description
1018 * Retrieve the classid for the current task, i.e. for the net_cls
1019 * cgroup to which *skb* belongs.
1020 *
1021 * This helper can be used on TC egress path, but not on ingress.
1022 *
1023 * The net_cls cgroup provides an interface to tag network packets
1024 * based on a user-provided identifier for all traffic coming from
1025 * the tasks belonging to the related cgroup. See also the related
1026 * kernel documentation, available from the Linux sources in file
1027 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1028 *
1029 * The Linux kernel has two versions for cgroups: there are
1030 * cgroups v1 and cgroups v2. Both are available to users, who can
1031 * use a mixture of them, but note that the net_cls cgroup is for
1032 * cgroup v1 only. This makes it incompatible with BPF programs
1033 * run on cgroups, which is a cgroup-v2-only feature (a socket can
1034 * only hold data for one version of cgroups at a time).
1035 *
1036 * This helper is only available is the kernel was compiled with
1037 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1038 * "**y**" or to "**m**".
1039 * Return
1040 * The classid, or 0 for the default unconfigured classid.
1041 *
1042 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1043 * Description
1044 * Push a *vlan_tci* (VLAN tag control information) of protocol
1045 * *vlan_proto* to the packet associated to *skb*, then update
1046 * the checksum. Note that if *vlan_proto* is different from
1047 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1048 * be **ETH_P_8021Q**.
1049 *
1050 * A call to this helper is susceptible to change the underlying
1051 * packet buffer. Therefore, at load time, all checks on pointers
1052 * previously done by the verifier are invalidated and must be
1053 * performed again, if the helper is used in combination with
1054 * direct packet access.
1055 * Return
1056 * 0 on success, or a negative error in case of failure.
1057 *
1058 * long bpf_skb_vlan_pop(struct sk_buff *skb)
1059 * Description
1060 * Pop a VLAN header from the packet associated to *skb*.
1061 *
1062 * A call to this helper is susceptible to change the underlying
1063 * packet buffer. Therefore, at load time, all checks on pointers
1064 * previously done by the verifier are invalidated and must be
1065 * performed again, if the helper is used in combination with
1066 * direct packet access.
1067 * Return
1068 * 0 on success, or a negative error in case of failure.
1069 *
1070 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1071 * Description
1072 * Get tunnel metadata. This helper takes a pointer *key* to an
1073 * empty **struct bpf_tunnel_key** of **size**, that will be
1074 * filled with tunnel metadata for the packet associated to *skb*.
1075 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1076 * indicates that the tunnel is based on IPv6 protocol instead of
1077 * IPv4.
1078 *
1079 * The **struct bpf_tunnel_key** is an object that generalizes the
1080 * principal parameters used by various tunneling protocols into a
1081 * single struct. This way, it can be used to easily make a
1082 * decision based on the contents of the encapsulation header,
1083 * "summarized" in this struct. In particular, it holds the IP
1084 * address of the remote end (IPv4 or IPv6, depending on the case)
1085 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1086 * this struct exposes the *key*\ **->tunnel_id**, which is
1087 * generally mapped to a VNI (Virtual Network Identifier), making
1088 * it programmable together with the **bpf_skb_set_tunnel_key**\
1089 * () helper.
1090 *
1091 * Let's imagine that the following code is part of a program
1092 * attached to the TC ingress interface, on one end of a GRE
1093 * tunnel, and is supposed to filter out all messages coming from
1094 * remote ends with IPv4 address other than 10.0.0.1:
1095 *
1096 * ::
1097 *
1098 * int ret;
1099 * struct bpf_tunnel_key key = {};
1100 *
1101 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1102 * if (ret < 0)
1103 * return TC_ACT_SHOT; // drop packet
1104 *
1105 * if (key.remote_ipv4 != 0x0a000001)
1106 * return TC_ACT_SHOT; // drop packet
1107 *
1108 * return TC_ACT_OK; // accept packet
1109 *
1110 * This interface can also be used with all encapsulation devices
1111 * that can operate in "collect metadata" mode: instead of having
1112 * one network device per specific configuration, the "collect
1113 * metadata" mode only requires a single device where the
1114 * configuration can be extracted from this helper.
1115 *
1116 * This can be used together with various tunnels such as VXLan,
1117 * Geneve, GRE or IP in IP (IPIP).
1118 * Return
1119 * 0 on success, or a negative error in case of failure.
1120 *
1121 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1122 * Description
1123 * Populate tunnel metadata for packet associated to *skb.* The
1124 * tunnel metadata is set to the contents of *key*, of *size*. The
1125 * *flags* can be set to a combination of the following values:
1126 *
1127 * **BPF_F_TUNINFO_IPV6**
1128 * Indicate that the tunnel is based on IPv6 protocol
1129 * instead of IPv4.
1130 * **BPF_F_ZERO_CSUM_TX**
1131 * For IPv4 packets, add a flag to tunnel metadata
1132 * indicating that checksum computation should be skipped
1133 * and checksum set to zeroes.
1134 * **BPF_F_DONT_FRAGMENT**
1135 * Add a flag to tunnel metadata indicating that the
1136 * packet should not be fragmented.
1137 * **BPF_F_SEQ_NUMBER**
1138 * Add a flag to tunnel metadata indicating that a
1139 * sequence number should be added to tunnel header before
1140 * sending the packet. This flag was added for GRE
1141 * encapsulation, but might be used with other protocols
1142 * as well in the future.
1143 *
1144 * Here is a typical usage on the transmit path:
1145 *
1146 * ::
1147 *
1148 * struct bpf_tunnel_key key;
1149 * populate key ...
1150 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1151 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1152 *
1153 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1154 * helper for additional information.
1155 * Return
1156 * 0 on success, or a negative error in case of failure.
1157 *
1158 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1159 * Description
1160 * Read the value of a perf event counter. This helper relies on a
1161 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1162 * the perf event counter is selected when *map* is updated with
1163 * perf event file descriptors. The *map* is an array whose size
1164 * is the number of available CPUs, and each cell contains a value
1165 * relative to one CPU. The value to retrieve is indicated by
1166 * *flags*, that contains the index of the CPU to look up, masked
1167 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1168 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1169 * current CPU should be retrieved.
1170 *
1171 * Note that before Linux 4.13, only hardware perf event can be
1172 * retrieved.
1173 *
1174 * Also, be aware that the newer helper
1175 * **bpf_perf_event_read_value**\ () is recommended over
1176 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1177 * quirks where error and counter value are used as a return code
1178 * (which is wrong to do since ranges may overlap). This issue is
1179 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1180 * time provides more features over the **bpf_perf_event_read**\
1181 * () interface. Please refer to the description of
1182 * **bpf_perf_event_read_value**\ () for details.
1183 * Return
1184 * The value of the perf event counter read from the map, or a
1185 * negative error code in case of failure.
1186 *
1187 * long bpf_redirect(u32 ifindex, u64 flags)
1188 * Description
1189 * Redirect the packet to another net device of index *ifindex*.
1190 * This helper is somewhat similar to **bpf_clone_redirect**\
1191 * (), except that the packet is not cloned, which provides
1192 * increased performance.
1193 *
1194 * Except for XDP, both ingress and egress interfaces can be used
1195 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1196 * to make the distinction (ingress path is selected if the flag
1197 * is present, egress path otherwise). Currently, XDP only
1198 * supports redirection to the egress interface, and accepts no
1199 * flag at all.
1200 *
1201 * The same effect can also be attained with the more generic
1202 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1203 * redirect target instead of providing it directly to the helper.
1204 * Return
1205 * For XDP, the helper returns **XDP_REDIRECT** on success or
1206 * **XDP_ABORTED** on error. For other program types, the values
1207 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1208 * error.
1209 *
1210 * u32 bpf_get_route_realm(struct sk_buff *skb)
1211 * Description
1212 * Retrieve the realm or the route, that is to say the
1213 * **tclassid** field of the destination for the *skb*. The
1214 * identifier retrieved is a user-provided tag, similar to the
1215 * one used with the net_cls cgroup (see description for
1216 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1217 * held by a route (a destination entry), not by a task.
1218 *
1219 * Retrieving this identifier works with the clsact TC egress hook
1220 * (see also **tc-bpf(8)**), or alternatively on conventional
1221 * classful egress qdiscs, but not on TC ingress path. In case of
1222 * clsact TC egress hook, this has the advantage that, internally,
1223 * the destination entry has not been dropped yet in the transmit
1224 * path. Therefore, the destination entry does not need to be
1225 * artificially held via **netif_keep_dst**\ () for a classful
1226 * qdisc until the *skb* is freed.
1227 *
1228 * This helper is available only if the kernel was compiled with
1229 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1230 * Return
1231 * The realm of the route for the packet associated to *skb*, or 0
1232 * if none was found.
1233 *
1234 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1235 * Description
1236 * Write raw *data* blob into a special BPF perf event held by
1237 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1238 * event must have the following attributes: **PERF_SAMPLE_RAW**
1239 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1240 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1241 *
1242 * The *flags* are used to indicate the index in *map* for which
1243 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1244 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1245 * to indicate that the index of the current CPU core should be
1246 * used.
1247 *
1248 * The value to write, of *size*, is passed through eBPF stack and
1249 * pointed by *data*.
1250 *
1251 * The context of the program *ctx* needs also be passed to the
1252 * helper.
1253 *
1254 * On user space, a program willing to read the values needs to
1255 * call **perf_event_open**\ () on the perf event (either for
1256 * one or for all CPUs) and to store the file descriptor into the
1257 * *map*. This must be done before the eBPF program can send data
1258 * into it. An example is available in file
1259 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1260 * tree (the eBPF program counterpart is in
1261 * *samples/bpf/trace_output_kern.c*).
1262 *
1263 * **bpf_perf_event_output**\ () achieves better performance
1264 * than **bpf_trace_printk**\ () for sharing data with user
1265 * space, and is much better suitable for streaming data from eBPF
1266 * programs.
1267 *
1268 * Note that this helper is not restricted to tracing use cases
1269 * and can be used with programs attached to TC or XDP as well,
1270 * where it allows for passing data to user space listeners. Data
1271 * can be:
1272 *
1273 * * Only custom structs,
1274 * * Only the packet payload, or
1275 * * A combination of both.
1276 * Return
1277 * 0 on success, or a negative error in case of failure.
1278 *
1279 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1280 * Description
1281 * This helper was provided as an easy way to load data from a
1282 * packet. It can be used to load *len* bytes from *offset* from
1283 * the packet associated to *skb*, into the buffer pointed by
1284 * *to*.
1285 *
1286 * Since Linux 4.7, usage of this helper has mostly been replaced
1287 * by "direct packet access", enabling packet data to be
1288 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1289 * pointing respectively to the first byte of packet data and to
1290 * the byte after the last byte of packet data. However, it
1291 * remains useful if one wishes to read large quantities of data
1292 * at once from a packet into the eBPF stack.
1293 * Return
1294 * 0 on success, or a negative error in case of failure.
1295 *
1296 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
1297 * Description
1298 * Walk a user or a kernel stack and return its id. To achieve
1299 * this, the helper needs *ctx*, which is a pointer to the context
1300 * on which the tracing program is executed, and a pointer to a
1301 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1302 *
1303 * The last argument, *flags*, holds the number of stack frames to
1304 * skip (from 0 to 255), masked with
1305 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1306 * a combination of the following flags:
1307 *
1308 * **BPF_F_USER_STACK**
1309 * Collect a user space stack instead of a kernel stack.
1310 * **BPF_F_FAST_STACK_CMP**
1311 * Compare stacks by hash only.
1312 * **BPF_F_REUSE_STACKID**
1313 * If two different stacks hash into the same *stackid*,
1314 * discard the old one.
1315 *
1316 * The stack id retrieved is a 32 bit long integer handle which
1317 * can be further combined with other data (including other stack
1318 * ids) and used as a key into maps. This can be useful for
1319 * generating a variety of graphs (such as flame graphs or off-cpu
1320 * graphs).
1321 *
1322 * For walking a stack, this helper is an improvement over
1323 * **bpf_probe_read**\ (), which can be used with unrolled loops
1324 * but is not efficient and consumes a lot of eBPF instructions.
1325 * Instead, **bpf_get_stackid**\ () can collect up to
1326 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1327 * this limit can be controlled with the **sysctl** program, and
1328 * that it should be manually increased in order to profile long
1329 * user stacks (such as stacks for Java programs). To do so, use:
1330 *
1331 * ::
1332 *
1333 * # sysctl kernel.perf_event_max_stack=<new value>
1334 * Return
1335 * The positive or null stack id on success, or a negative error
1336 * in case of failure.
1337 *
1338 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1339 * Description
1340 * Compute a checksum difference, from the raw buffer pointed by
1341 * *from*, of length *from_size* (that must be a multiple of 4),
1342 * towards the raw buffer pointed by *to*, of size *to_size*
1343 * (same remark). An optional *seed* can be added to the value
1344 * (this can be cascaded, the seed may come from a previous call
1345 * to the helper).
1346 *
1347 * This is flexible enough to be used in several ways:
1348 *
1349 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1350 * checksum, it can be used when pushing new data.
1351 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1352 * checksum, it can be used when removing data from a packet.
1353 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1354 * can be used to compute a diff. Note that *from_size* and
1355 * *to_size* do not need to be equal.
1356 *
1357 * This helper can be used in combination with
1358 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1359 * which one can feed in the difference computed with
1360 * **bpf_csum_diff**\ ().
1361 * Return
1362 * The checksum result, or a negative error code in case of
1363 * failure.
1364 *
1365 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1366 * Description
1367 * Retrieve tunnel options metadata for the packet associated to
1368 * *skb*, and store the raw tunnel option data to the buffer *opt*
1369 * of *size*.
1370 *
1371 * This helper can be used with encapsulation devices that can
1372 * operate in "collect metadata" mode (please refer to the related
1373 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1374 * more details). A particular example where this can be used is
1375 * in combination with the Geneve encapsulation protocol, where it
1376 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1377 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1378 * the eBPF program. This allows for full customization of these
1379 * headers.
1380 * Return
1381 * The size of the option data retrieved.
1382 *
1383 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1384 * Description
1385 * Set tunnel options metadata for the packet associated to *skb*
1386 * to the option data contained in the raw buffer *opt* of *size*.
1387 *
1388 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1389 * helper for additional information.
1390 * Return
1391 * 0 on success, or a negative error in case of failure.
1392 *
1393 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1394 * Description
1395 * Change the protocol of the *skb* to *proto*. Currently
1396 * supported are transition from IPv4 to IPv6, and from IPv6 to
1397 * IPv4. The helper takes care of the groundwork for the
1398 * transition, including resizing the socket buffer. The eBPF
1399 * program is expected to fill the new headers, if any, via
1400 * **skb_store_bytes**\ () and to recompute the checksums with
1401 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1402 * (). The main case for this helper is to perform NAT64
1403 * operations out of an eBPF program.
1404 *
1405 * Internally, the GSO type is marked as dodgy so that headers are
1406 * checked and segments are recalculated by the GSO/GRO engine.
1407 * The size for GSO target is adapted as well.
1408 *
1409 * All values for *flags* are reserved for future usage, and must
1410 * be left at zero.
1411 *
1412 * A call to this helper is susceptible to change the underlying
1413 * packet buffer. Therefore, at load time, all checks on pointers
1414 * previously done by the verifier are invalidated and must be
1415 * performed again, if the helper is used in combination with
1416 * direct packet access.
1417 * Return
1418 * 0 on success, or a negative error in case of failure.
1419 *
1420 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
1421 * Description
1422 * Change the packet type for the packet associated to *skb*. This
1423 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1424 * the eBPF program does not have a write access to *skb*\
1425 * **->pkt_type** beside this helper. Using a helper here allows
1426 * for graceful handling of errors.
1427 *
1428 * The major use case is to change incoming *skb*s to
1429 * **PACKET_HOST** in a programmatic way instead of having to
1430 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1431 * example.
1432 *
1433 * Note that *type* only allows certain values. At this time, they
1434 * are:
1435 *
1436 * **PACKET_HOST**
1437 * Packet is for us.
1438 * **PACKET_BROADCAST**
1439 * Send packet to all.
1440 * **PACKET_MULTICAST**
1441 * Send packet to group.
1442 * **PACKET_OTHERHOST**
1443 * Send packet to someone else.
1444 * Return
1445 * 0 on success, or a negative error in case of failure.
1446 *
1447 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1448 * Description
1449 * Check whether *skb* is a descendant of the cgroup2 held by
1450 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1451 * Return
1452 * The return value depends on the result of the test, and can be:
1453 *
1454 * * 0, if the *skb* failed the cgroup2 descendant test.
1455 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1456 * * A negative error code, if an error occurred.
1457 *
1458 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1459 * Description
1460 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1461 * not set, in particular if the hash was cleared due to mangling,
1462 * recompute this hash. Later accesses to the hash can be done
1463 * directly with *skb*\ **->hash**.
1464 *
1465 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1466 * prototype with **bpf_skb_change_proto**\ (), or calling
1467 * **bpf_skb_store_bytes**\ () with the
1468 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1469 * the hash and to trigger a new computation for the next call to
1470 * **bpf_get_hash_recalc**\ ().
1471 * Return
1472 * The 32-bit hash.
1473 *
1474 * u64 bpf_get_current_task(void)
1475 * Return
1476 * A pointer to the current task struct.
1477 *
1478 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
1479 * Description
1480 * Attempt in a safe way to write *len* bytes from the buffer
1481 * *src* to *dst* in memory. It only works for threads that are in
1482 * user context, and *dst* must be a valid user space address.
1483 *
1484 * This helper should not be used to implement any kind of
1485 * security mechanism because of TOC-TOU attacks, but rather to
1486 * debug, divert, and manipulate execution of semi-cooperative
1487 * processes.
1488 *
1489 * Keep in mind that this feature is meant for experiments, and it
1490 * has a risk of crashing the system and running programs.
1491 * Therefore, when an eBPF program using this helper is attached,
1492 * a warning including PID and process name is printed to kernel
1493 * logs.
1494 * Return
1495 * 0 on success, or a negative error in case of failure.
1496 *
1497 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1498 * Description
1499 * Check whether the probe is being run is the context of a given
1500 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1501 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1502 * Return
1503 * The return value depends on the result of the test, and can be:
1504 *
1505 * * 0, if current task belongs to the cgroup2.
1506 * * 1, if current task does not belong to the cgroup2.
1507 * * A negative error code, if an error occurred.
1508 *
1509 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1510 * Description
1511 * Resize (trim or grow) the packet associated to *skb* to the
1512 * new *len*. The *flags* are reserved for future usage, and must
1513 * be left at zero.
1514 *
1515 * The basic idea is that the helper performs the needed work to
1516 * change the size of the packet, then the eBPF program rewrites
1517 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1518 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1519 * and others. This helper is a slow path utility intended for
1520 * replies with control messages. And because it is targeted for
1521 * slow path, the helper itself can afford to be slow: it
1522 * implicitly linearizes, unclones and drops offloads from the
1523 * *skb*.
1524 *
1525 * A call to this helper is susceptible to change the underlying
1526 * packet buffer. Therefore, at load time, all checks on pointers
1527 * previously done by the verifier are invalidated and must be
1528 * performed again, if the helper is used in combination with
1529 * direct packet access.
1530 * Return
1531 * 0 on success, or a negative error in case of failure.
1532 *
1533 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1534 * Description
1535 * Pull in non-linear data in case the *skb* is non-linear and not
1536 * all of *len* are part of the linear section. Make *len* bytes
1537 * from *skb* readable and writable. If a zero value is passed for
1538 * *len*, then the whole length of the *skb* is pulled.
1539 *
1540 * This helper is only needed for reading and writing with direct
1541 * packet access.
1542 *
1543 * For direct packet access, testing that offsets to access
1544 * are within packet boundaries (test on *skb*\ **->data_end**) is
1545 * susceptible to fail if offsets are invalid, or if the requested
1546 * data is in non-linear parts of the *skb*. On failure the
1547 * program can just bail out, or in the case of a non-linear
1548 * buffer, use a helper to make the data available. The
1549 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1550 * the data. Another one consists in using **bpf_skb_pull_data**
1551 * to pull in once the non-linear parts, then retesting and
1552 * eventually access the data.
1553 *
1554 * At the same time, this also makes sure the *skb* is uncloned,
1555 * which is a necessary condition for direct write. As this needs
1556 * to be an invariant for the write part only, the verifier
1557 * detects writes and adds a prologue that is calling
1558 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1559 * the very beginning in case it is indeed cloned.
1560 *
1561 * A call to this helper is susceptible to change the underlying
1562 * packet buffer. Therefore, at load time, all checks on pointers
1563 * previously done by the verifier are invalidated and must be
1564 * performed again, if the helper is used in combination with
1565 * direct packet access.
1566 * Return
1567 * 0 on success, or a negative error in case of failure.
1568 *
1569 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1570 * Description
1571 * Add the checksum *csum* into *skb*\ **->csum** in case the
1572 * driver has supplied a checksum for the entire packet into that
1573 * field. Return an error otherwise. This helper is intended to be
1574 * used in combination with **bpf_csum_diff**\ (), in particular
1575 * when the checksum needs to be updated after data has been
1576 * written into the packet through direct packet access.
1577 * Return
1578 * The checksum on success, or a negative error code in case of
1579 * failure.
1580 *
1581 * void bpf_set_hash_invalid(struct sk_buff *skb)
1582 * Description
1583 * Invalidate the current *skb*\ **->hash**. It can be used after
1584 * mangling on headers through direct packet access, in order to
1585 * indicate that the hash is outdated and to trigger a
1586 * recalculation the next time the kernel tries to access this
1587 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1588 *
1589 * long bpf_get_numa_node_id(void)
1590 * Description
1591 * Return the id of the current NUMA node. The primary use case
1592 * for this helper is the selection of sockets for the local NUMA
1593 * node, when the program is attached to sockets using the
1594 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1595 * but the helper is also available to other eBPF program types,
1596 * similarly to **bpf_get_smp_processor_id**\ ().
1597 * Return
1598 * The id of current NUMA node.
1599 *
1600 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1601 * Description
1602 * Grows headroom of packet associated to *skb* and adjusts the
1603 * offset of the MAC header accordingly, adding *len* bytes of
1604 * space. It automatically extends and reallocates memory as
1605 * required.
1606 *
1607 * This helper can be used on a layer 3 *skb* to push a MAC header
1608 * for redirection into a layer 2 device.
1609 *
1610 * All values for *flags* are reserved for future usage, and must
1611 * be left at zero.
1612 *
1613 * A call to this helper is susceptible to change the underlying
1614 * packet buffer. Therefore, at load time, all checks on pointers
1615 * previously done by the verifier are invalidated and must be
1616 * performed again, if the helper is used in combination with
1617 * direct packet access.
1618 * Return
1619 * 0 on success, or a negative error in case of failure.
1620 *
1621 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1622 * Description
1623 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1624 * it is possible to use a negative value for *delta*. This helper
1625 * can be used to prepare the packet for pushing or popping
1626 * headers.
1627 *
1628 * A call to this helper is susceptible to change the underlying
1629 * packet buffer. Therefore, at load time, all checks on pointers
1630 * previously done by the verifier are invalidated and must be
1631 * performed again, if the helper is used in combination with
1632 * direct packet access.
1633 * Return
1634 * 0 on success, or a negative error in case of failure.
1635 *
1636 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
1637 * Description
1638 * Copy a NUL terminated string from an unsafe kernel address
1639 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
1640 * more details.
1641 *
1642 * Generally, use **bpf_probe_read_user_str**\ () or
1643 * **bpf_probe_read_kernel_str**\ () instead.
1644 * Return
1645 * On success, the strictly positive length of the string,
1646 * including the trailing NUL character. On error, a negative
1647 * value.
1648 *
1649 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1650 * Description
1651 * If the **struct sk_buff** pointed by *skb* has a known socket,
1652 * retrieve the cookie (generated by the kernel) of this socket.
1653 * If no cookie has been set yet, generate a new cookie. Once
1654 * generated, the socket cookie remains stable for the life of the
1655 * socket. This helper can be useful for monitoring per socket
1656 * networking traffic statistics as it provides a global socket
1657 * identifier that can be assumed unique.
1658 * Return
1659 * A 8-byte long unique number on success, or 0 if the socket
1660 * field is missing inside *skb*.
1661 *
1662 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1663 * Description
1664 * Equivalent to bpf_get_socket_cookie() helper that accepts
1665 * *skb*, but gets socket from **struct bpf_sock_addr** context.
1666 * Return
1667 * A 8-byte long unique number.
1668 *
1669 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1670 * Description
1671 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
1672 * *skb*, but gets socket from **struct bpf_sock_ops** context.
1673 * Return
1674 * A 8-byte long unique number.
1675 *
1676 * u64 bpf_get_socket_cookie(struct sock *sk)
1677 * Description
1678 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
1679 * *sk*, but gets socket from a BTF **struct sock**. This helper
1680 * also works for sleepable programs.
1681 * Return
1682 * A 8-byte long unique number or 0 if *sk* is NULL.
1683 *
1684 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1685 * Return
1686 * The owner UID of the socket associated to *skb*. If the socket
1687 * is **NULL**, or if it is not a full socket (i.e. if it is a
1688 * time-wait or a request socket instead), **overflowuid** value
1689 * is returned (note that **overflowuid** might also be the actual
1690 * UID value for the socket).
1691 *
1692 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
1693 * Description
1694 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1695 * to value *hash*.
1696 * Return
1697 * 0
1698 *
1699 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1700 * Description
1701 * Emulate a call to **setsockopt()** on the socket associated to
1702 * *bpf_socket*, which must be a full socket. The *level* at
1703 * which the option resides and the name *optname* of the option
1704 * must be specified, see **setsockopt(2)** for more information.
1705 * The option value of length *optlen* is pointed by *optval*.
1706 *
1707 * *bpf_socket* should be one of the following:
1708 *
1709 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1710 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1711 * and **BPF_CGROUP_INET6_CONNECT**.
1712 *
1713 * This helper actually implements a subset of **setsockopt()**.
1714 * It supports the following *level*\ s:
1715 *
1716 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1717 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1718 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
1719 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
1720 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1721 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1722 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
1723 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
1724 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**.
1725 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1726 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1727 * Return
1728 * 0 on success, or a negative error in case of failure.
1729 *
1730 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1731 * Description
1732 * Grow or shrink the room for data in the packet associated to
1733 * *skb* by *len_diff*, and according to the selected *mode*.
1734 *
1735 * By default, the helper will reset any offloaded checksum
1736 * indicator of the skb to CHECKSUM_NONE. This can be avoided
1737 * by the following flag:
1738 *
1739 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
1740 * checksum data of the skb to CHECKSUM_NONE.
1741 *
1742 * There are two supported modes at this time:
1743 *
1744 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1745 * (room space is added or removed below the layer 2 header).
1746 *
1747 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1748 * (room space is added or removed below the layer 3 header).
1749 *
1750 * The following flags are supported at this time:
1751 *
1752 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1753 * Adjusting mss in this way is not allowed for datagrams.
1754 *
1755 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1756 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
1757 * Any new space is reserved to hold a tunnel header.
1758 * Configure skb offsets and other fields accordingly.
1759 *
1760 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1761 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
1762 * Use with ENCAP_L3 flags to further specify the tunnel type.
1763 *
1764 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1765 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1766 * type; *len* is the length of the inner MAC header.
1767 *
1768 * A call to this helper is susceptible to change the underlying
1769 * packet buffer. Therefore, at load time, all checks on pointers
1770 * previously done by the verifier are invalidated and must be
1771 * performed again, if the helper is used in combination with
1772 * direct packet access.
1773 * Return
1774 * 0 on success, or a negative error in case of failure.
1775 *
1776 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1777 * Description
1778 * Redirect the packet to the endpoint referenced by *map* at
1779 * index *key*. Depending on its type, this *map* can contain
1780 * references to net devices (for forwarding packets through other
1781 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1782 * but this is only implemented for native XDP (with driver
1783 * support) as of this writing).
1784 *
1785 * The lower two bits of *flags* are used as the return code if
1786 * the map lookup fails. This is so that the return value can be
1787 * one of the XDP program return codes up to **XDP_TX**, as chosen
1788 * by the caller. Any higher bits in the *flags* argument must be
1789 * unset.
1790 *
1791 * See also **bpf_redirect**\ (), which only supports redirecting
1792 * to an ifindex, but doesn't require a map to do so.
1793 * Return
1794 * **XDP_REDIRECT** on success, or the value of the two lower bits
1795 * of the *flags* argument on error.
1796 *
1797 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
1798 * Description
1799 * Redirect the packet to the socket referenced by *map* (of type
1800 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1801 * egress interfaces can be used for redirection. The
1802 * **BPF_F_INGRESS** value in *flags* is used to make the
1803 * distinction (ingress path is selected if the flag is present,
1804 * egress path otherwise). This is the only flag supported for now.
1805 * Return
1806 * **SK_PASS** on success, or **SK_DROP** on error.
1807 *
1808 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1809 * Description
1810 * Add an entry to, or update a *map* referencing sockets. The
1811 * *skops* is used as a new value for the entry associated to
1812 * *key*. *flags* is one of:
1813 *
1814 * **BPF_NOEXIST**
1815 * The entry for *key* must not exist in the map.
1816 * **BPF_EXIST**
1817 * The entry for *key* must already exist in the map.
1818 * **BPF_ANY**
1819 * No condition on the existence of the entry for *key*.
1820 *
1821 * If the *map* has eBPF programs (parser and verdict), those will
1822 * be inherited by the socket being added. If the socket is
1823 * already attached to eBPF programs, this results in an error.
1824 * Return
1825 * 0 on success, or a negative error in case of failure.
1826 *
1827 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1828 * Description
1829 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1830 * *delta* (which can be positive or negative). Note that this
1831 * operation modifies the address stored in *xdp_md*\ **->data**,
1832 * so the latter must be loaded only after the helper has been
1833 * called.
1834 *
1835 * The use of *xdp_md*\ **->data_meta** is optional and programs
1836 * are not required to use it. The rationale is that when the
1837 * packet is processed with XDP (e.g. as DoS filter), it is
1838 * possible to push further meta data along with it before passing
1839 * to the stack, and to give the guarantee that an ingress eBPF
1840 * program attached as a TC classifier on the same device can pick
1841 * this up for further post-processing. Since TC works with socket
1842 * buffers, it remains possible to set from XDP the **mark** or
1843 * **priority** pointers, or other pointers for the socket buffer.
1844 * Having this scratch space generic and programmable allows for
1845 * more flexibility as the user is free to store whatever meta
1846 * data they need.
1847 *
1848 * A call to this helper is susceptible to change the underlying
1849 * packet buffer. Therefore, at load time, all checks on pointers
1850 * previously done by the verifier are invalidated and must be
1851 * performed again, if the helper is used in combination with
1852 * direct packet access.
1853 * Return
1854 * 0 on success, or a negative error in case of failure.
1855 *
1856 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1857 * Description
1858 * Read the value of a perf event counter, and store it into *buf*
1859 * of size *buf_size*. This helper relies on a *map* of type
1860 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1861 * counter is selected when *map* is updated with perf event file
1862 * descriptors. The *map* is an array whose size is the number of
1863 * available CPUs, and each cell contains a value relative to one
1864 * CPU. The value to retrieve is indicated by *flags*, that
1865 * contains the index of the CPU to look up, masked with
1866 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1867 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1868 * current CPU should be retrieved.
1869 *
1870 * This helper behaves in a way close to
1871 * **bpf_perf_event_read**\ () helper, save that instead of
1872 * just returning the value observed, it fills the *buf*
1873 * structure. This allows for additional data to be retrieved: in
1874 * particular, the enabled and running times (in *buf*\
1875 * **->enabled** and *buf*\ **->running**, respectively) are
1876 * copied. In general, **bpf_perf_event_read_value**\ () is
1877 * recommended over **bpf_perf_event_read**\ (), which has some
1878 * ABI issues and provides fewer functionalities.
1879 *
1880 * These values are interesting, because hardware PMU (Performance
1881 * Monitoring Unit) counters are limited resources. When there are
1882 * more PMU based perf events opened than available counters,
1883 * kernel will multiplex these events so each event gets certain
1884 * percentage (but not all) of the PMU time. In case that
1885 * multiplexing happens, the number of samples or counter value
1886 * will not reflect the case compared to when no multiplexing
1887 * occurs. This makes comparison between different runs difficult.
1888 * Typically, the counter value should be normalized before
1889 * comparing to other experiments. The usual normalization is done
1890 * as follows.
1891 *
1892 * ::
1893 *
1894 * normalized_counter = counter * t_enabled / t_running
1895 *
1896 * Where t_enabled is the time enabled for event and t_running is
1897 * the time running for event since last normalization. The
1898 * enabled and running times are accumulated since the perf event
1899 * open. To achieve scaling factor between two invocations of an
1900 * eBPF program, users can use CPU id as the key (which is
1901 * typical for perf array usage model) to remember the previous
1902 * value and do the calculation inside the eBPF program.
1903 * Return
1904 * 0 on success, or a negative error in case of failure.
1905 *
1906 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1907 * Description
1908 * For en eBPF program attached to a perf event, retrieve the
1909 * value of the event counter associated to *ctx* and store it in
1910 * the structure pointed by *buf* and of size *buf_size*. Enabled
1911 * and running times are also stored in the structure (see
1912 * description of helper **bpf_perf_event_read_value**\ () for
1913 * more details).
1914 * Return
1915 * 0 on success, or a negative error in case of failure.
1916 *
1917 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
1918 * Description
1919 * Emulate a call to **getsockopt()** on the socket associated to
1920 * *bpf_socket*, which must be a full socket. The *level* at
1921 * which the option resides and the name *optname* of the option
1922 * must be specified, see **getsockopt(2)** for more information.
1923 * The retrieved value is stored in the structure pointed by
1924 * *opval* and of length *optlen*.
1925 *
1926 * *bpf_socket* should be one of the following:
1927 *
1928 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1929 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1930 * and **BPF_CGROUP_INET6_CONNECT**.
1931 *
1932 * This helper actually implements a subset of **getsockopt()**.
1933 * It supports the following *level*\ s:
1934 *
1935 * * **IPPROTO_TCP**, which supports *optname*
1936 * **TCP_CONGESTION**.
1937 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1938 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1939 * Return
1940 * 0 on success, or a negative error in case of failure.
1941 *
1942 * long bpf_override_return(struct pt_regs *regs, u64 rc)
1943 * Description
1944 * Used for error injection, this helper uses kprobes to override
1945 * the return value of the probed function, and to set it to *rc*.
1946 * The first argument is the context *regs* on which the kprobe
1947 * works.
1948 *
1949 * This helper works by setting the PC (program counter)
1950 * to an override function which is run in place of the original
1951 * probed function. This means the probed function is not run at
1952 * all. The replacement function just returns with the required
1953 * value.
1954 *
1955 * This helper has security implications, and thus is subject to
1956 * restrictions. It is only available if the kernel was compiled
1957 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1958 * option, and in this case it only works on functions tagged with
1959 * **ALLOW_ERROR_INJECTION** in the kernel code.
1960 *
1961 * Also, the helper is only available for the architectures having
1962 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1963 * x86 architecture is the only one to support this feature.
1964 * Return
1965 * 0
1966 *
1967 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1968 * Description
1969 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1970 * for the full TCP socket associated to *bpf_sock_ops* to
1971 * *argval*.
1972 *
1973 * The primary use of this field is to determine if there should
1974 * be calls to eBPF programs of type
1975 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1976 * code. A program of the same type can change its value, per
1977 * connection and as necessary, when the connection is
1978 * established. This field is directly accessible for reading, but
1979 * this helper must be used for updates in order to return an
1980 * error if an eBPF program tries to set a callback that is not
1981 * supported in the current kernel.
1982 *
1983 * *argval* is a flag array which can combine these flags:
1984 *
1985 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1986 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1987 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1988 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
1989 *
1990 * Therefore, this function can be used to clear a callback flag by
1991 * setting the appropriate bit to zero. e.g. to disable the RTO
1992 * callback:
1993 *
1994 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1995 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1996 *
1997 * Here are some examples of where one could call such eBPF
1998 * program:
1999 *
2000 * * When RTO fires.
2001 * * When a packet is retransmitted.
2002 * * When the connection terminates.
2003 * * When a packet is sent.
2004 * * When a packet is received.
2005 * Return
2006 * Code **-EINVAL** if the socket is not a full TCP socket;
2007 * otherwise, a positive number containing the bits that could not
2008 * be set is returned (which comes down to 0 if all bits were set
2009 * as required).
2010 *
2011 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
2012 * Description
2013 * This helper is used in programs implementing policies at the
2014 * socket level. If the message *msg* is allowed to pass (i.e. if
2015 * the verdict eBPF program returns **SK_PASS**), redirect it to
2016 * the socket referenced by *map* (of type
2017 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2018 * egress interfaces can be used for redirection. The
2019 * **BPF_F_INGRESS** value in *flags* is used to make the
2020 * distinction (ingress path is selected if the flag is present,
2021 * egress path otherwise). This is the only flag supported for now.
2022 * Return
2023 * **SK_PASS** on success, or **SK_DROP** on error.
2024 *
2025 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2026 * Description
2027 * For socket policies, apply the verdict of the eBPF program to
2028 * the next *bytes* (number of bytes) of message *msg*.
2029 *
2030 * For example, this helper can be used in the following cases:
2031 *
2032 * * A single **sendmsg**\ () or **sendfile**\ () system call
2033 * contains multiple logical messages that the eBPF program is
2034 * supposed to read and for which it should apply a verdict.
2035 * * An eBPF program only cares to read the first *bytes* of a
2036 * *msg*. If the message has a large payload, then setting up
2037 * and calling the eBPF program repeatedly for all bytes, even
2038 * though the verdict is already known, would create unnecessary
2039 * overhead.
2040 *
2041 * When called from within an eBPF program, the helper sets a
2042 * counter internal to the BPF infrastructure, that is used to
2043 * apply the last verdict to the next *bytes*. If *bytes* is
2044 * smaller than the current data being processed from a
2045 * **sendmsg**\ () or **sendfile**\ () system call, the first
2046 * *bytes* will be sent and the eBPF program will be re-run with
2047 * the pointer for start of data pointing to byte number *bytes*
2048 * **+ 1**. If *bytes* is larger than the current data being
2049 * processed, then the eBPF verdict will be applied to multiple
2050 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2051 * consumed.
2052 *
2053 * Note that if a socket closes with the internal counter holding
2054 * a non-zero value, this is not a problem because data is not
2055 * being buffered for *bytes* and is sent as it is received.
2056 * Return
2057 * 0
2058 *
2059 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2060 * Description
2061 * For socket policies, prevent the execution of the verdict eBPF
2062 * program for message *msg* until *bytes* (byte number) have been
2063 * accumulated.
2064 *
2065 * This can be used when one needs a specific number of bytes
2066 * before a verdict can be assigned, even if the data spans
2067 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2068 * case would be a user calling **sendmsg**\ () repeatedly with
2069 * 1-byte long message segments. Obviously, this is bad for
2070 * performance, but it is still valid. If the eBPF program needs
2071 * *bytes* bytes to validate a header, this helper can be used to
2072 * prevent the eBPF program to be called again until *bytes* have
2073 * been accumulated.
2074 * Return
2075 * 0
2076 *
2077 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2078 * Description
2079 * For socket policies, pull in non-linear data from user space
2080 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2081 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2082 * respectively.
2083 *
2084 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2085 * *msg* it can only parse data that the (**data**, **data_end**)
2086 * pointers have already consumed. For **sendmsg**\ () hooks this
2087 * is likely the first scatterlist element. But for calls relying
2088 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2089 * be the range (**0**, **0**) because the data is shared with
2090 * user space and by default the objective is to avoid allowing
2091 * user space to modify data while (or after) eBPF verdict is
2092 * being decided. This helper can be used to pull in data and to
2093 * set the start and end pointer to given values. Data will be
2094 * copied if necessary (i.e. if data was not linear and if start
2095 * and end pointers do not point to the same chunk).
2096 *
2097 * A call to this helper is susceptible to change the underlying
2098 * packet buffer. Therefore, at load time, all checks on pointers
2099 * previously done by the verifier are invalidated and must be
2100 * performed again, if the helper is used in combination with
2101 * direct packet access.
2102 *
2103 * All values for *flags* are reserved for future usage, and must
2104 * be left at zero.
2105 * Return
2106 * 0 on success, or a negative error in case of failure.
2107 *
2108 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2109 * Description
2110 * Bind the socket associated to *ctx* to the address pointed by
2111 * *addr*, of length *addr_len*. This allows for making outgoing
2112 * connection from the desired IP address, which can be useful for
2113 * example when all processes inside a cgroup should use one
2114 * single IP address on a host that has multiple IP configured.
2115 *
2116 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2117 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2118 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2119 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2120 * behavior and lets the kernel efficiently pick up an unused
2121 * port as long as 4-tuple is unique. Passing non-zero port might
2122 * lead to degraded performance.
2123 * Return
2124 * 0 on success, or a negative error in case of failure.
2125 *
2126 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2127 * Description
2128 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2129 * possible to both shrink and grow the packet tail.
2130 * Shrink done via *delta* being a negative integer.
2131 *
2132 * A call to this helper is susceptible to change the underlying
2133 * packet buffer. Therefore, at load time, all checks on pointers
2134 * previously done by the verifier are invalidated and must be
2135 * performed again, if the helper is used in combination with
2136 * direct packet access.
2137 * Return
2138 * 0 on success, or a negative error in case of failure.
2139 *
2140 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2141 * Description
2142 * Retrieve the XFRM state (IP transform framework, see also
2143 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2144 *
2145 * The retrieved value is stored in the **struct bpf_xfrm_state**
2146 * pointed by *xfrm_state* and of length *size*.
2147 *
2148 * All values for *flags* are reserved for future usage, and must
2149 * be left at zero.
2150 *
2151 * This helper is available only if the kernel was compiled with
2152 * **CONFIG_XFRM** configuration option.
2153 * Return
2154 * 0 on success, or a negative error in case of failure.
2155 *
2156 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2157 * Description
2158 * Return a user or a kernel stack in bpf program provided buffer.
2159 * To achieve this, the helper needs *ctx*, which is a pointer
2160 * to the context on which the tracing program is executed.
2161 * To store the stacktrace, the bpf program provides *buf* with
2162 * a nonnegative *size*.
2163 *
2164 * The last argument, *flags*, holds the number of stack frames to
2165 * skip (from 0 to 255), masked with
2166 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2167 * the following flags:
2168 *
2169 * **BPF_F_USER_STACK**
2170 * Collect a user space stack instead of a kernel stack.
2171 * **BPF_F_USER_BUILD_ID**
2172 * Collect buildid+offset instead of ips for user stack,
2173 * only valid if **BPF_F_USER_STACK** is also specified.
2174 *
2175 * **bpf_get_stack**\ () can collect up to
2176 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2177 * to sufficient large buffer size. Note that
2178 * this limit can be controlled with the **sysctl** program, and
2179 * that it should be manually increased in order to profile long
2180 * user stacks (such as stacks for Java programs). To do so, use:
2181 *
2182 * ::
2183 *
2184 * # sysctl kernel.perf_event_max_stack=<new value>
2185 * Return
2186 * A non-negative value equal to or less than *size* on success,
2187 * or a negative error in case of failure.
2188 *
2189 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2190 * Description
2191 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2192 * it provides an easy way to load *len* bytes from *offset*
2193 * from the packet associated to *skb*, into the buffer pointed
2194 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2195 * a fifth argument *start_header* exists in order to select a
2196 * base offset to start from. *start_header* can be one of:
2197 *
2198 * **BPF_HDR_START_MAC**
2199 * Base offset to load data from is *skb*'s mac header.
2200 * **BPF_HDR_START_NET**
2201 * Base offset to load data from is *skb*'s network header.
2202 *
2203 * In general, "direct packet access" is the preferred method to
2204 * access packet data, however, this helper is in particular useful
2205 * in socket filters where *skb*\ **->data** does not always point
2206 * to the start of the mac header and where "direct packet access"
2207 * is not available.
2208 * Return
2209 * 0 on success, or a negative error in case of failure.
2210 *
2211 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2212 * Description
2213 * Do FIB lookup in kernel tables using parameters in *params*.
2214 * If lookup is successful and result shows packet is to be
2215 * forwarded, the neighbor tables are searched for the nexthop.
2216 * If successful (ie., FIB lookup shows forwarding and nexthop
2217 * is resolved), the nexthop address is returned in ipv4_dst
2218 * or ipv6_dst based on family, smac is set to mac address of
2219 * egress device, dmac is set to nexthop mac address, rt_metric
2220 * is set to metric from route (IPv4/IPv6 only), and ifindex
2221 * is set to the device index of the nexthop from the FIB lookup.
2222 *
2223 * *plen* argument is the size of the passed in struct.
2224 * *flags* argument can be a combination of one or more of the
2225 * following values:
2226 *
2227 * **BPF_FIB_LOOKUP_DIRECT**
2228 * Do a direct table lookup vs full lookup using FIB
2229 * rules.
2230 * **BPF_FIB_LOOKUP_OUTPUT**
2231 * Perform lookup from an egress perspective (default is
2232 * ingress).
2233 *
2234 * *ctx* is either **struct xdp_md** for XDP programs or
2235 * **struct sk_buff** tc cls_act programs.
2236 * Return
2237 * * < 0 if any input argument is invalid
2238 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2239 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2240 * packet is not forwarded or needs assist from full stack
2241 *
2242 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
2243 * was exceeded and output params->mtu_result contains the MTU.
2244 *
2245 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2246 * Description
2247 * Add an entry to, or update a sockhash *map* referencing sockets.
2248 * The *skops* is used as a new value for the entry associated to
2249 * *key*. *flags* is one of:
2250 *
2251 * **BPF_NOEXIST**
2252 * The entry for *key* must not exist in the map.
2253 * **BPF_EXIST**
2254 * The entry for *key* must already exist in the map.
2255 * **BPF_ANY**
2256 * No condition on the existence of the entry for *key*.
2257 *
2258 * If the *map* has eBPF programs (parser and verdict), those will
2259 * be inherited by the socket being added. If the socket is
2260 * already attached to eBPF programs, this results in an error.
2261 * Return
2262 * 0 on success, or a negative error in case of failure.
2263 *
2264 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2265 * Description
2266 * This helper is used in programs implementing policies at the
2267 * socket level. If the message *msg* is allowed to pass (i.e. if
2268 * the verdict eBPF program returns **SK_PASS**), redirect it to
2269 * the socket referenced by *map* (of type
2270 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2271 * egress interfaces can be used for redirection. The
2272 * **BPF_F_INGRESS** value in *flags* is used to make the
2273 * distinction (ingress path is selected if the flag is present,
2274 * egress path otherwise). This is the only flag supported for now.
2275 * Return
2276 * **SK_PASS** on success, or **SK_DROP** on error.
2277 *
2278 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2279 * Description
2280 * This helper is used in programs implementing policies at the
2281 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2282 * if the verdict eBPF program returns **SK_PASS**), redirect it
2283 * to the socket referenced by *map* (of type
2284 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2285 * egress interfaces can be used for redirection. The
2286 * **BPF_F_INGRESS** value in *flags* is used to make the
2287 * distinction (ingress path is selected if the flag is present,
2288 * egress otherwise). This is the only flag supported for now.
2289 * Return
2290 * **SK_PASS** on success, or **SK_DROP** on error.
2291 *
2292 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2293 * Description
2294 * Encapsulate the packet associated to *skb* within a Layer 3
2295 * protocol header. This header is provided in the buffer at
2296 * address *hdr*, with *len* its size in bytes. *type* indicates
2297 * the protocol of the header and can be one of:
2298 *
2299 * **BPF_LWT_ENCAP_SEG6**
2300 * IPv6 encapsulation with Segment Routing Header
2301 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2302 * the IPv6 header is computed by the kernel.
2303 * **BPF_LWT_ENCAP_SEG6_INLINE**
2304 * Only works if *skb* contains an IPv6 packet. Insert a
2305 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2306 * the IPv6 header.
2307 * **BPF_LWT_ENCAP_IP**
2308 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2309 * must be IPv4 or IPv6, followed by zero or more
2310 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2311 * total bytes in all prepended headers. Please note that
2312 * if **skb_is_gso**\ (*skb*) is true, no more than two
2313 * headers can be prepended, and the inner header, if
2314 * present, should be either GRE or UDP/GUE.
2315 *
2316 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2317 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2318 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2319 * **BPF_PROG_TYPE_LWT_XMIT**.
2320 *
2321 * A call to this helper is susceptible to change the underlying
2322 * packet buffer. Therefore, at load time, all checks on pointers
2323 * previously done by the verifier are invalidated and must be
2324 * performed again, if the helper is used in combination with
2325 * direct packet access.
2326 * Return
2327 * 0 on success, or a negative error in case of failure.
2328 *
2329 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2330 * Description
2331 * Store *len* bytes from address *from* into the packet
2332 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2333 * inside the outermost IPv6 Segment Routing Header can be
2334 * modified through this helper.
2335 *
2336 * A call to this helper is susceptible to change the underlying
2337 * packet buffer. Therefore, at load time, all checks on pointers
2338 * previously done by the verifier are invalidated and must be
2339 * performed again, if the helper is used in combination with
2340 * direct packet access.
2341 * Return
2342 * 0 on success, or a negative error in case of failure.
2343 *
2344 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2345 * Description
2346 * Adjust the size allocated to TLVs in the outermost IPv6
2347 * Segment Routing Header contained in the packet associated to
2348 * *skb*, at position *offset* by *delta* bytes. Only offsets
2349 * after the segments are accepted. *delta* can be as well
2350 * positive (growing) as negative (shrinking).
2351 *
2352 * A call to this helper is susceptible to change the underlying
2353 * packet buffer. Therefore, at load time, all checks on pointers
2354 * previously done by the verifier are invalidated and must be
2355 * performed again, if the helper is used in combination with
2356 * direct packet access.
2357 * Return
2358 * 0 on success, or a negative error in case of failure.
2359 *
2360 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2361 * Description
2362 * Apply an IPv6 Segment Routing action of type *action* to the
2363 * packet associated to *skb*. Each action takes a parameter
2364 * contained at address *param*, and of length *param_len* bytes.
2365 * *action* can be one of:
2366 *
2367 * **SEG6_LOCAL_ACTION_END_X**
2368 * End.X action: Endpoint with Layer-3 cross-connect.
2369 * Type of *param*: **struct in6_addr**.
2370 * **SEG6_LOCAL_ACTION_END_T**
2371 * End.T action: Endpoint with specific IPv6 table lookup.
2372 * Type of *param*: **int**.
2373 * **SEG6_LOCAL_ACTION_END_B6**
2374 * End.B6 action: Endpoint bound to an SRv6 policy.
2375 * Type of *param*: **struct ipv6_sr_hdr**.
2376 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2377 * End.B6.Encap action: Endpoint bound to an SRv6
2378 * encapsulation policy.
2379 * Type of *param*: **struct ipv6_sr_hdr**.
2380 *
2381 * A call to this helper is susceptible to change the underlying
2382 * packet buffer. Therefore, at load time, all checks on pointers
2383 * previously done by the verifier are invalidated and must be
2384 * performed again, if the helper is used in combination with
2385 * direct packet access.
2386 * Return
2387 * 0 on success, or a negative error in case of failure.
2388 *
2389 * long bpf_rc_repeat(void *ctx)
2390 * Description
2391 * This helper is used in programs implementing IR decoding, to
2392 * report a successfully decoded repeat key message. This delays
2393 * the generation of a key up event for previously generated
2394 * key down event.
2395 *
2396 * Some IR protocols like NEC have a special IR message for
2397 * repeating last button, for when a button is held down.
2398 *
2399 * The *ctx* should point to the lirc sample as passed into
2400 * the program.
2401 *
2402 * This helper is only available is the kernel was compiled with
2403 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2404 * "**y**".
2405 * Return
2406 * 0
2407 *
2408 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2409 * Description
2410 * This helper is used in programs implementing IR decoding, to
2411 * report a successfully decoded key press with *scancode*,
2412 * *toggle* value in the given *protocol*. The scancode will be
2413 * translated to a keycode using the rc keymap, and reported as
2414 * an input key down event. After a period a key up event is
2415 * generated. This period can be extended by calling either
2416 * **bpf_rc_keydown**\ () again with the same values, or calling
2417 * **bpf_rc_repeat**\ ().
2418 *
2419 * Some protocols include a toggle bit, in case the button was
2420 * released and pressed again between consecutive scancodes.
2421 *
2422 * The *ctx* should point to the lirc sample as passed into
2423 * the program.
2424 *
2425 * The *protocol* is the decoded protocol number (see
2426 * **enum rc_proto** for some predefined values).
2427 *
2428 * This helper is only available is the kernel was compiled with
2429 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2430 * "**y**".
2431 * Return
2432 * 0
2433 *
2434 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
2435 * Description
2436 * Return the cgroup v2 id of the socket associated with the *skb*.
2437 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2438 * helper for cgroup v1 by providing a tag resp. identifier that
2439 * can be matched on or used for map lookups e.g. to implement
2440 * policy. The cgroup v2 id of a given path in the hierarchy is
2441 * exposed in user space through the f_handle API in order to get
2442 * to the same 64-bit id.
2443 *
2444 * This helper can be used on TC egress path, but not on ingress,
2445 * and is available only if the kernel was compiled with the
2446 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2447 * Return
2448 * The id is returned or 0 in case the id could not be retrieved.
2449 *
2450 * u64 bpf_get_current_cgroup_id(void)
2451 * Return
2452 * A 64-bit integer containing the current cgroup id based
2453 * on the cgroup within which the current task is running.
2454 *
2455 * void *bpf_get_local_storage(void *map, u64 flags)
2456 * Description
2457 * Get the pointer to the local storage area.
2458 * The type and the size of the local storage is defined
2459 * by the *map* argument.
2460 * The *flags* meaning is specific for each map type,
2461 * and has to be 0 for cgroup local storage.
2462 *
2463 * Depending on the BPF program type, a local storage area
2464 * can be shared between multiple instances of the BPF program,
2465 * running simultaneously.
2466 *
2467 * A user should care about the synchronization by himself.
2468 * For example, by using the **BPF_ATOMIC** instructions to alter
2469 * the shared data.
2470 * Return
2471 * A pointer to the local storage area.
2472 *
2473 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2474 * Description
2475 * Select a **SO_REUSEPORT** socket from a
2476 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2477 * It checks the selected socket is matching the incoming
2478 * request in the socket buffer.
2479 * Return
2480 * 0 on success, or a negative error in case of failure.
2481 *
2482 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2483 * Description
2484 * Return id of cgroup v2 that is ancestor of cgroup associated
2485 * with the *skb* at the *ancestor_level*. The root cgroup is at
2486 * *ancestor_level* zero and each step down the hierarchy
2487 * increments the level. If *ancestor_level* == level of cgroup
2488 * associated with *skb*, then return value will be same as that
2489 * of **bpf_skb_cgroup_id**\ ().
2490 *
2491 * The helper is useful to implement policies based on cgroups
2492 * that are upper in hierarchy than immediate cgroup associated
2493 * with *skb*.
2494 *
2495 * The format of returned id and helper limitations are same as in
2496 * **bpf_skb_cgroup_id**\ ().
2497 * Return
2498 * The id is returned or 0 in case the id could not be retrieved.
2499 *
2500 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2501 * Description
2502 * Look for TCP socket matching *tuple*, optionally in a child
2503 * network namespace *netns*. The return value must be checked,
2504 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2505 *
2506 * The *ctx* should point to the context of the program, such as
2507 * the skb or socket (depending on the hook in use). This is used
2508 * to determine the base network namespace for the lookup.
2509 *
2510 * *tuple_size* must be one of:
2511 *
2512 * **sizeof**\ (*tuple*\ **->ipv4**)
2513 * Look for an IPv4 socket.
2514 * **sizeof**\ (*tuple*\ **->ipv6**)
2515 * Look for an IPv6 socket.
2516 *
2517 * If the *netns* is a negative signed 32-bit integer, then the
2518 * socket lookup table in the netns associated with the *ctx*
2519 * will be used. For the TC hooks, this is the netns of the device
2520 * in the skb. For socket hooks, this is the netns of the socket.
2521 * If *netns* is any other signed 32-bit value greater than or
2522 * equal to zero then it specifies the ID of the netns relative to
2523 * the netns associated with the *ctx*. *netns* values beyond the
2524 * range of 32-bit integers are reserved for future use.
2525 *
2526 * All values for *flags* are reserved for future usage, and must
2527 * be left at zero.
2528 *
2529 * This helper is available only if the kernel was compiled with
2530 * **CONFIG_NET** configuration option.
2531 * Return
2532 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2533 * For sockets with reuseport option, the **struct bpf_sock**
2534 * result is from *reuse*\ **->socks**\ [] using the hash of the
2535 * tuple.
2536 *
2537 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2538 * Description
2539 * Look for UDP socket matching *tuple*, optionally in a child
2540 * network namespace *netns*. The return value must be checked,
2541 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2542 *
2543 * The *ctx* should point to the context of the program, such as
2544 * the skb or socket (depending on the hook in use). This is used
2545 * to determine the base network namespace for the lookup.
2546 *
2547 * *tuple_size* must be one of:
2548 *
2549 * **sizeof**\ (*tuple*\ **->ipv4**)
2550 * Look for an IPv4 socket.
2551 * **sizeof**\ (*tuple*\ **->ipv6**)
2552 * Look for an IPv6 socket.
2553 *
2554 * If the *netns* is a negative signed 32-bit integer, then the
2555 * socket lookup table in the netns associated with the *ctx*
2556 * will be used. For the TC hooks, this is the netns of the device
2557 * in the skb. For socket hooks, this is the netns of the socket.
2558 * If *netns* is any other signed 32-bit value greater than or
2559 * equal to zero then it specifies the ID of the netns relative to
2560 * the netns associated with the *ctx*. *netns* values beyond the
2561 * range of 32-bit integers are reserved for future use.
2562 *
2563 * All values for *flags* are reserved for future usage, and must
2564 * be left at zero.
2565 *
2566 * This helper is available only if the kernel was compiled with
2567 * **CONFIG_NET** configuration option.
2568 * Return
2569 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2570 * For sockets with reuseport option, the **struct bpf_sock**
2571 * result is from *reuse*\ **->socks**\ [] using the hash of the
2572 * tuple.
2573 *
2574 * long bpf_sk_release(void *sock)
2575 * Description
2576 * Release the reference held by *sock*. *sock* must be a
2577 * non-**NULL** pointer that was returned from
2578 * **bpf_sk_lookup_xxx**\ ().
2579 * Return
2580 * 0 on success, or a negative error in case of failure.
2581 *
2582 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2583 * Description
2584 * Push an element *value* in *map*. *flags* is one of:
2585 *
2586 * **BPF_EXIST**
2587 * If the queue/stack is full, the oldest element is
2588 * removed to make room for this.
2589 * Return
2590 * 0 on success, or a negative error in case of failure.
2591 *
2592 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
2593 * Description
2594 * Pop an element from *map*.
2595 * Return
2596 * 0 on success, or a negative error in case of failure.
2597 *
2598 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
2599 * Description
2600 * Get an element from *map* without removing it.
2601 * Return
2602 * 0 on success, or a negative error in case of failure.
2603 *
2604 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2605 * Description
2606 * For socket policies, insert *len* bytes into *msg* at offset
2607 * *start*.
2608 *
2609 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2610 * *msg* it may want to insert metadata or options into the *msg*.
2611 * This can later be read and used by any of the lower layer BPF
2612 * hooks.
2613 *
2614 * This helper may fail if under memory pressure (a malloc
2615 * fails) in these cases BPF programs will get an appropriate
2616 * error and BPF programs will need to handle them.
2617 * Return
2618 * 0 on success, or a negative error in case of failure.
2619 *
2620 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
2621 * Description
2622 * Will remove *len* bytes from a *msg* starting at byte *start*.
2623 * This may result in **ENOMEM** errors under certain situations if
2624 * an allocation and copy are required due to a full ring buffer.
2625 * However, the helper will try to avoid doing the allocation
2626 * if possible. Other errors can occur if input parameters are
2627 * invalid either due to *start* byte not being valid part of *msg*
2628 * payload and/or *pop* value being to large.
2629 * Return
2630 * 0 on success, or a negative error in case of failure.
2631 *
2632 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2633 * Description
2634 * This helper is used in programs implementing IR decoding, to
2635 * report a successfully decoded pointer movement.
2636 *
2637 * The *ctx* should point to the lirc sample as passed into
2638 * the program.
2639 *
2640 * This helper is only available is the kernel was compiled with
2641 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2642 * "**y**".
2643 * Return
2644 * 0
2645 *
2646 * long bpf_spin_lock(struct bpf_spin_lock *lock)
2647 * Description
2648 * Acquire a spinlock represented by the pointer *lock*, which is
2649 * stored as part of a value of a map. Taking the lock allows to
2650 * safely update the rest of the fields in that value. The
2651 * spinlock can (and must) later be released with a call to
2652 * **bpf_spin_unlock**\ (\ *lock*\ ).
2653 *
2654 * Spinlocks in BPF programs come with a number of restrictions
2655 * and constraints:
2656 *
2657 * * **bpf_spin_lock** objects are only allowed inside maps of
2658 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2659 * list could be extended in the future).
2660 * * BTF description of the map is mandatory.
2661 * * The BPF program can take ONE lock at a time, since taking two
2662 * or more could cause dead locks.
2663 * * Only one **struct bpf_spin_lock** is allowed per map element.
2664 * * When the lock is taken, calls (either BPF to BPF or helpers)
2665 * are not allowed.
2666 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2667 * allowed inside a spinlock-ed region.
2668 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2669 * the lock, on all execution paths, before it returns.
2670 * * The BPF program can access **struct bpf_spin_lock** only via
2671 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2672 * helpers. Loading or storing data into the **struct
2673 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2674 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2675 * of the map value must be a struct and have **struct
2676 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2677 * Nested lock inside another struct is not allowed.
2678 * * The **struct bpf_spin_lock** *lock* field in a map value must
2679 * be aligned on a multiple of 4 bytes in that value.
2680 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2681 * the **bpf_spin_lock** field to user space.
2682 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2683 * a BPF program, do not update the **bpf_spin_lock** field.
2684 * * **bpf_spin_lock** cannot be on the stack or inside a
2685 * networking packet (it can only be inside of a map values).
2686 * * **bpf_spin_lock** is available to root only.
2687 * * Tracing programs and socket filter programs cannot use
2688 * **bpf_spin_lock**\ () due to insufficient preemption checks
2689 * (but this may change in the future).
2690 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2691 * Return
2692 * 0
2693 *
2694 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
2695 * Description
2696 * Release the *lock* previously locked by a call to
2697 * **bpf_spin_lock**\ (\ *lock*\ ).
2698 * Return
2699 * 0
2700 *
2701 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2702 * Description
2703 * This helper gets a **struct bpf_sock** pointer such
2704 * that all the fields in this **bpf_sock** can be accessed.
2705 * Return
2706 * A **struct bpf_sock** pointer on success, or **NULL** in
2707 * case of failure.
2708 *
2709 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2710 * Description
2711 * This helper gets a **struct bpf_tcp_sock** pointer from a
2712 * **struct bpf_sock** pointer.
2713 * Return
2714 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
2715 * case of failure.
2716 *
2717 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
2718 * Description
2719 * Set ECN (Explicit Congestion Notification) field of IP header
2720 * to **CE** (Congestion Encountered) if current value is **ECT**
2721 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2722 * and IPv4.
2723 * Return
2724 * 1 if the **CE** flag is set (either by the current helper call
2725 * or because it was already present), 0 if it is not set.
2726 *
2727 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2728 * Description
2729 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2730 * **bpf_sk_release**\ () is unnecessary and not allowed.
2731 * Return
2732 * A **struct bpf_sock** pointer on success, or **NULL** in
2733 * case of failure.
2734 *
2735 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2736 * Description
2737 * Look for TCP socket matching *tuple*, optionally in a child
2738 * network namespace *netns*. The return value must be checked,
2739 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2740 *
2741 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2742 * that it also returns timewait or request sockets. Use
2743 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2744 * full structure.
2745 *
2746 * This helper is available only if the kernel was compiled with
2747 * **CONFIG_NET** configuration option.
2748 * Return
2749 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2750 * For sockets with reuseport option, the **struct bpf_sock**
2751 * result is from *reuse*\ **->socks**\ [] using the hash of the
2752 * tuple.
2753 *
2754 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2755 * Description
2756 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2757 * the listening socket in *sk*.
2758 *
2759 * *iph* points to the start of the IPv4 or IPv6 header, while
2760 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2761 * **sizeof**\ (**struct ip6hdr**).
2762 *
2763 * *th* points to the start of the TCP header, while *th_len*
2764 * contains **sizeof**\ (**struct tcphdr**).
2765 * Return
2766 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2767 * error otherwise.
2768 *
2769 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2770 * Description
2771 * Get name of sysctl in /proc/sys/ and copy it into provided by
2772 * program buffer *buf* of size *buf_len*.
2773 *
2774 * The buffer is always NUL terminated, unless it's zero-sized.
2775 *
2776 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2777 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2778 * only (e.g. "tcp_mem").
2779 * Return
2780 * Number of character copied (not including the trailing NUL).
2781 *
2782 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2783 * truncated name in this case).
2784 *
2785 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2786 * Description
2787 * Get current value of sysctl as it is presented in /proc/sys
2788 * (incl. newline, etc), and copy it as a string into provided
2789 * by program buffer *buf* of size *buf_len*.
2790 *
2791 * The whole value is copied, no matter what file position user
2792 * space issued e.g. sys_read at.
2793 *
2794 * The buffer is always NUL terminated, unless it's zero-sized.
2795 * Return
2796 * Number of character copied (not including the trailing NUL).
2797 *
2798 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2799 * truncated name in this case).
2800 *
2801 * **-EINVAL** if current value was unavailable, e.g. because
2802 * sysctl is uninitialized and read returns -EIO for it.
2803 *
2804 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2805 * Description
2806 * Get new value being written by user space to sysctl (before
2807 * the actual write happens) and copy it as a string into
2808 * provided by program buffer *buf* of size *buf_len*.
2809 *
2810 * User space may write new value at file position > 0.
2811 *
2812 * The buffer is always NUL terminated, unless it's zero-sized.
2813 * Return
2814 * Number of character copied (not including the trailing NUL).
2815 *
2816 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2817 * truncated name in this case).
2818 *
2819 * **-EINVAL** if sysctl is being read.
2820 *
2821 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2822 * Description
2823 * Override new value being written by user space to sysctl with
2824 * value provided by program in buffer *buf* of size *buf_len*.
2825 *
2826 * *buf* should contain a string in same form as provided by user
2827 * space on sysctl write.
2828 *
2829 * User space may write new value at file position > 0. To override
2830 * the whole sysctl value file position should be set to zero.
2831 * Return
2832 * 0 on success.
2833 *
2834 * **-E2BIG** if the *buf_len* is too big.
2835 *
2836 * **-EINVAL** if sysctl is being read.
2837 *
2838 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2839 * Description
2840 * Convert the initial part of the string from buffer *buf* of
2841 * size *buf_len* to a long integer according to the given base
2842 * and save the result in *res*.
2843 *
2844 * The string may begin with an arbitrary amount of white space
2845 * (as determined by **isspace**\ (3)) followed by a single
2846 * optional '**-**' sign.
2847 *
2848 * Five least significant bits of *flags* encode base, other bits
2849 * are currently unused.
2850 *
2851 * Base must be either 8, 10, 16 or 0 to detect it automatically
2852 * similar to user space **strtol**\ (3).
2853 * Return
2854 * Number of characters consumed on success. Must be positive but
2855 * no more than *buf_len*.
2856 *
2857 * **-EINVAL** if no valid digits were found or unsupported base
2858 * was provided.
2859 *
2860 * **-ERANGE** if resulting value was out of range.
2861 *
2862 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2863 * Description
2864 * Convert the initial part of the string from buffer *buf* of
2865 * size *buf_len* to an unsigned long integer according to the
2866 * given base and save the result in *res*.
2867 *
2868 * The string may begin with an arbitrary amount of white space
2869 * (as determined by **isspace**\ (3)).
2870 *
2871 * Five least significant bits of *flags* encode base, other bits
2872 * are currently unused.
2873 *
2874 * Base must be either 8, 10, 16 or 0 to detect it automatically
2875 * similar to user space **strtoul**\ (3).
2876 * Return
2877 * Number of characters consumed on success. Must be positive but
2878 * no more than *buf_len*.
2879 *
2880 * **-EINVAL** if no valid digits were found or unsupported base
2881 * was provided.
2882 *
2883 * **-ERANGE** if resulting value was out of range.
2884 *
2885 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
2886 * Description
2887 * Get a bpf-local-storage from a *sk*.
2888 *
2889 * Logically, it could be thought of getting the value from
2890 * a *map* with *sk* as the **key**. From this
2891 * perspective, the usage is not much different from
2892 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2893 * helper enforces the key must be a full socket and the map must
2894 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
2895 *
2896 * Underneath, the value is stored locally at *sk* instead of
2897 * the *map*. The *map* is used as the bpf-local-storage
2898 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2899 * searched against all bpf-local-storages residing at *sk*.
2900 *
2901 * *sk* is a kernel **struct sock** pointer for LSM program.
2902 * *sk* is a **struct bpf_sock** pointer for other program types.
2903 *
2904 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
2905 * used such that a new bpf-local-storage will be
2906 * created if one does not exist. *value* can be used
2907 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
2908 * the initial value of a bpf-local-storage. If *value* is
2909 * **NULL**, the new bpf-local-storage will be zero initialized.
2910 * Return
2911 * A bpf-local-storage pointer is returned on success.
2912 *
2913 * **NULL** if not found or there was an error in adding
2914 * a new bpf-local-storage.
2915 *
2916 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
2917 * Description
2918 * Delete a bpf-local-storage from a *sk*.
2919 * Return
2920 * 0 on success.
2921 *
2922 * **-ENOENT** if the bpf-local-storage cannot be found.
2923 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
2924 *
2925 * long bpf_send_signal(u32 sig)
2926 * Description
2927 * Send signal *sig* to the process of the current task.
2928 * The signal may be delivered to any of this process's threads.
2929 * Return
2930 * 0 on success or successfully queued.
2931 *
2932 * **-EBUSY** if work queue under nmi is full.
2933 *
2934 * **-EINVAL** if *sig* is invalid.
2935 *
2936 * **-EPERM** if no permission to send the *sig*.
2937 *
2938 * **-EAGAIN** if bpf program can try again.
2939 *
2940 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2941 * Description
2942 * Try to issue a SYN cookie for the packet with corresponding
2943 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2944 *
2945 * *iph* points to the start of the IPv4 or IPv6 header, while
2946 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2947 * **sizeof**\ (**struct ip6hdr**).
2948 *
2949 * *th* points to the start of the TCP header, while *th_len*
2950 * contains the length of the TCP header.
2951 * Return
2952 * On success, lower 32 bits hold the generated SYN cookie in
2953 * followed by 16 bits which hold the MSS value for that cookie,
2954 * and the top 16 bits are unused.
2955 *
2956 * On failure, the returned value is one of the following:
2957 *
2958 * **-EINVAL** SYN cookie cannot be issued due to error
2959 *
2960 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2961 *
2962 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2963 *
2964 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
2965 *
2966 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2967 * Description
2968 * Write raw *data* blob into a special BPF perf event held by
2969 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2970 * event must have the following attributes: **PERF_SAMPLE_RAW**
2971 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2972 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2973 *
2974 * The *flags* are used to indicate the index in *map* for which
2975 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2976 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2977 * to indicate that the index of the current CPU core should be
2978 * used.
2979 *
2980 * The value to write, of *size*, is passed through eBPF stack and
2981 * pointed by *data*.
2982 *
2983 * *ctx* is a pointer to in-kernel struct sk_buff.
2984 *
2985 * This helper is similar to **bpf_perf_event_output**\ () but
2986 * restricted to raw_tracepoint bpf programs.
2987 * Return
2988 * 0 on success, or a negative error in case of failure.
2989 *
2990 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2991 * Description
2992 * Safely attempt to read *size* bytes from user space address
2993 * *unsafe_ptr* and store the data in *dst*.
2994 * Return
2995 * 0 on success, or a negative error in case of failure.
2996 *
2997 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2998 * Description
2999 * Safely attempt to read *size* bytes from kernel space address
3000 * *unsafe_ptr* and store the data in *dst*.
3001 * Return
3002 * 0 on success, or a negative error in case of failure.
3003 *
3004 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
3005 * Description
3006 * Copy a NUL terminated string from an unsafe user address
3007 * *unsafe_ptr* to *dst*. The *size* should include the
3008 * terminating NUL byte. In case the string length is smaller than
3009 * *size*, the target is not padded with further NUL bytes. If the
3010 * string length is larger than *size*, just *size*-1 bytes are
3011 * copied and the last byte is set to NUL.
3012 *
3013 * On success, returns the number of bytes that were written,
3014 * including the terminal NUL. This makes this helper useful in
3015 * tracing programs for reading strings, and more importantly to
3016 * get its length at runtime. See the following snippet:
3017 *
3018 * ::
3019 *
3020 * SEC("kprobe/sys_open")
3021 * void bpf_sys_open(struct pt_regs *ctx)
3022 * {
3023 * char buf[PATHLEN]; // PATHLEN is defined to 256
3024 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
3025 * ctx->di);
3026 *
3027 * // Consume buf, for example push it to
3028 * // userspace via bpf_perf_event_output(); we
3029 * // can use res (the string length) as event
3030 * // size, after checking its boundaries.
3031 * }
3032 *
3033 * In comparison, using **bpf_probe_read_user**\ () helper here
3034 * instead to read the string would require to estimate the length
3035 * at compile time, and would often result in copying more memory
3036 * than necessary.
3037 *
3038 * Another useful use case is when parsing individual process
3039 * arguments or individual environment variables navigating
3040 * *current*\ **->mm->arg_start** and *current*\
3041 * **->mm->env_start**: using this helper and the return value,
3042 * one can quickly iterate at the right offset of the memory area.
3043 * Return
3044 * On success, the strictly positive length of the output string,
3045 * including the trailing NUL character. On error, a negative
3046 * value.
3047 *
3048 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3049 * Description
3050 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3051 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3052 * Return
3053 * On success, the strictly positive length of the string, including
3054 * the trailing NUL character. On error, a negative value.
3055 *
3056 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3057 * Description
3058 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3059 * *rcv_nxt* is the ack_seq to be sent out.
3060 * Return
3061 * 0 on success, or a negative error in case of failure.
3062 *
3063 * long bpf_send_signal_thread(u32 sig)
3064 * Description
3065 * Send signal *sig* to the thread corresponding to the current task.
3066 * Return
3067 * 0 on success or successfully queued.
3068 *
3069 * **-EBUSY** if work queue under nmi is full.
3070 *
3071 * **-EINVAL** if *sig* is invalid.
3072 *
3073 * **-EPERM** if no permission to send the *sig*.
3074 *
3075 * **-EAGAIN** if bpf program can try again.
3076 *
3077 * u64 bpf_jiffies64(void)
3078 * Description
3079 * Obtain the 64bit jiffies
3080 * Return
3081 * The 64 bit jiffies
3082 *
3083 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3084 * Description
3085 * For an eBPF program attached to a perf event, retrieve the
3086 * branch records (**struct perf_branch_entry**) associated to *ctx*
3087 * and store it in the buffer pointed by *buf* up to size
3088 * *size* bytes.
3089 * Return
3090 * On success, number of bytes written to *buf*. On error, a
3091 * negative value.
3092 *
3093 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3094 * instead return the number of bytes required to store all the
3095 * branch entries. If this flag is set, *buf* may be NULL.
3096 *
3097 * **-EINVAL** if arguments invalid or **size** not a multiple
3098 * of **sizeof**\ (**struct perf_branch_entry**\ ).
3099 *
3100 * **-ENOENT** if architecture does not support branch records.
3101 *
3102 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3103 * Description
3104 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3105 * *namespace* will be returned in *nsdata*.
3106 * Return
3107 * 0 on success, or one of the following in case of failure:
3108 *
3109 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3110 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3111 *
3112 * **-ENOENT** if pidns does not exists for the current task.
3113 *
3114 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3115 * Description
3116 * Write raw *data* blob into a special BPF perf event held by
3117 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3118 * event must have the following attributes: **PERF_SAMPLE_RAW**
3119 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3120 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3121 *
3122 * The *flags* are used to indicate the index in *map* for which
3123 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3124 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3125 * to indicate that the index of the current CPU core should be
3126 * used.
3127 *
3128 * The value to write, of *size*, is passed through eBPF stack and
3129 * pointed by *data*.
3130 *
3131 * *ctx* is a pointer to in-kernel struct xdp_buff.
3132 *
3133 * This helper is similar to **bpf_perf_eventoutput**\ () but
3134 * restricted to raw_tracepoint bpf programs.
3135 * Return
3136 * 0 on success, or a negative error in case of failure.
3137 *
3138 * u64 bpf_get_netns_cookie(void *ctx)
3139 * Description
3140 * Retrieve the cookie (generated by the kernel) of the network
3141 * namespace the input *ctx* is associated with. The network
3142 * namespace cookie remains stable for its lifetime and provides
3143 * a global identifier that can be assumed unique. If *ctx* is
3144 * NULL, then the helper returns the cookie for the initial
3145 * network namespace. The cookie itself is very similar to that
3146 * of **bpf_get_socket_cookie**\ () helper, but for network
3147 * namespaces instead of sockets.
3148 * Return
3149 * A 8-byte long opaque number.
3150 *
3151 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3152 * Description
3153 * Return id of cgroup v2 that is ancestor of the cgroup associated
3154 * with the current task at the *ancestor_level*. The root cgroup
3155 * is at *ancestor_level* zero and each step down the hierarchy
3156 * increments the level. If *ancestor_level* == level of cgroup
3157 * associated with the current task, then return value will be the
3158 * same as that of **bpf_get_current_cgroup_id**\ ().
3159 *
3160 * The helper is useful to implement policies based on cgroups
3161 * that are upper in hierarchy than immediate cgroup associated
3162 * with the current task.
3163 *
3164 * The format of returned id and helper limitations are same as in
3165 * **bpf_get_current_cgroup_id**\ ().
3166 * Return
3167 * The id is returned or 0 in case the id could not be retrieved.
3168 *
3169 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
3170 * Description
3171 * Helper is overloaded depending on BPF program type. This
3172 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3173 * **BPF_PROG_TYPE_SCHED_ACT** programs.
3174 *
3175 * Assign the *sk* to the *skb*. When combined with appropriate
3176 * routing configuration to receive the packet towards the socket,
3177 * will cause *skb* to be delivered to the specified socket.
3178 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3179 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3180 * interfere with successful delivery to the socket.
3181 *
3182 * This operation is only valid from TC ingress path.
3183 *
3184 * The *flags* argument must be zero.
3185 * Return
3186 * 0 on success, or a negative error in case of failure:
3187 *
3188 * **-EINVAL** if specified *flags* are not supported.
3189 *
3190 * **-ENOENT** if the socket is unavailable for assignment.
3191 *
3192 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3193 *
3194 * **-EOPNOTSUPP** if the operation is not supported, for example
3195 * a call from outside of TC ingress.
3196 *
3197 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3198 * (reuseport).
3199 *
3200 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3201 * Description
3202 * Helper is overloaded depending on BPF program type. This
3203 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
3204 *
3205 * Select the *sk* as a result of a socket lookup.
3206 *
3207 * For the operation to succeed passed socket must be compatible
3208 * with the packet description provided by the *ctx* object.
3209 *
3210 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
3211 * be an exact match. While IP family (**AF_INET** or
3212 * **AF_INET6**) must be compatible, that is IPv6 sockets
3213 * that are not v6-only can be selected for IPv4 packets.
3214 *
3215 * Only TCP listeners and UDP unconnected sockets can be
3216 * selected. *sk* can also be NULL to reset any previous
3217 * selection.
3218 *
3219 * *flags* argument can combination of following values:
3220 *
3221 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
3222 * socket selection, potentially done by a BPF program
3223 * that ran before us.
3224 *
3225 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
3226 * load-balancing within reuseport group for the socket
3227 * being selected.
3228 *
3229 * On success *ctx->sk* will point to the selected socket.
3230 *
3231 * Return
3232 * 0 on success, or a negative errno in case of failure.
3233 *
3234 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
3235 * not compatible with packet family (*ctx->family*).
3236 *
3237 * * **-EEXIST** if socket has been already selected,
3238 * potentially by another program, and
3239 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
3240 *
3241 * * **-EINVAL** if unsupported flags were specified.
3242 *
3243 * * **-EPROTOTYPE** if socket L4 protocol
3244 * (*sk->protocol*) doesn't match packet protocol
3245 * (*ctx->protocol*).
3246 *
3247 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
3248 * state (TCP listening or UDP unconnected).
3249 *
3250 * u64 bpf_ktime_get_boot_ns(void)
3251 * Description
3252 * Return the time elapsed since system boot, in nanoseconds.
3253 * Does include the time the system was suspended.
3254 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
3255 * Return
3256 * Current *ktime*.
3257 *
3258 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3259 * Description
3260 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3261 * out the format string.
3262 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3263 * the format string itself. The *data* and *data_len* are format string
3264 * arguments. The *data* are a **u64** array and corresponding format string
3265 * values are stored in the array. For strings and pointers where pointees
3266 * are accessed, only the pointer values are stored in the *data* array.
3267 * The *data_len* is the size of *data* in bytes.
3268 *
3269 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3270 * Reading kernel memory may fail due to either invalid address or
3271 * valid address but requiring a major memory fault. If reading kernel memory
3272 * fails, the string for **%s** will be an empty string, and the ip
3273 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
3274 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
3275 * Return
3276 * 0 on success, or a negative error in case of failure:
3277 *
3278 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3279 * by returning 1 from bpf program.
3280 *
3281 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3282 *
3283 * **-E2BIG** if *fmt* contains too many format specifiers.
3284 *
3285 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3286 *
3287 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
3288 * Description
3289 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
3290 * The *m* represents the seq_file. The *data* and *len* represent the
3291 * data to write in bytes.
3292 * Return
3293 * 0 on success, or a negative error in case of failure:
3294 *
3295 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
3296 *
3297 * u64 bpf_sk_cgroup_id(void *sk)
3298 * Description
3299 * Return the cgroup v2 id of the socket *sk*.
3300 *
3301 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
3302 * returned from **bpf_sk_lookup_xxx**\ (),
3303 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3304 * same as in **bpf_skb_cgroup_id**\ ().
3305 *
3306 * This helper is available only if the kernel was compiled with
3307 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3308 * Return
3309 * The id is returned or 0 in case the id could not be retrieved.
3310 *
3311 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
3312 * Description
3313 * Return id of cgroup v2 that is ancestor of cgroup associated
3314 * with the *sk* at the *ancestor_level*. The root cgroup is at
3315 * *ancestor_level* zero and each step down the hierarchy
3316 * increments the level. If *ancestor_level* == level of cgroup
3317 * associated with *sk*, then return value will be same as that
3318 * of **bpf_sk_cgroup_id**\ ().
3319 *
3320 * The helper is useful to implement policies based on cgroups
3321 * that are upper in hierarchy than immediate cgroup associated
3322 * with *sk*.
3323 *
3324 * The format of returned id and helper limitations are same as in
3325 * **bpf_sk_cgroup_id**\ ().
3326 * Return
3327 * The id is returned or 0 in case the id could not be retrieved.
3328 *
3329 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
3330 * Description
3331 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
3332 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3333 * of new data availability is sent.
3334 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3335 * of new data availability is sent unconditionally.
3336 * Return
3337 * 0 on success, or a negative error in case of failure.
3338 *
3339 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3340 * Description
3341 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3342 * Return
3343 * Valid pointer with *size* bytes of memory available; NULL,
3344 * otherwise.
3345 *
3346 * void bpf_ringbuf_submit(void *data, u64 flags)
3347 * Description
3348 * Submit reserved ring buffer sample, pointed to by *data*.
3349 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3350 * of new data availability is sent.
3351 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3352 * of new data availability is sent unconditionally.
3353 * Return
3354 * Nothing. Always succeeds.
3355 *
3356 * void bpf_ringbuf_discard(void *data, u64 flags)
3357 * Description
3358 * Discard reserved ring buffer sample, pointed to by *data*.
3359 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3360 * of new data availability is sent.
3361 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3362 * of new data availability is sent unconditionally.
3363 * Return
3364 * Nothing. Always succeeds.
3365 *
3366 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3367 * Description
3368 * Query various characteristics of provided ring buffer. What
3369 * exactly is queries is determined by *flags*:
3370 *
3371 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
3372 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
3373 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
3374 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
3375 *
3376 * Data returned is just a momentary snapshot of actual values
3377 * and could be inaccurate, so this facility should be used to
3378 * power heuristics and for reporting, not to make 100% correct
3379 * calculation.
3380 * Return
3381 * Requested value, or 0, if *flags* are not recognized.
3382 *
3383 * long bpf_csum_level(struct sk_buff *skb, u64 level)
3384 * Description
3385 * Change the skbs checksum level by one layer up or down, or
3386 * reset it entirely to none in order to have the stack perform
3387 * checksum validation. The level is applicable to the following
3388 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
3389 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
3390 * through **bpf_skb_adjust_room**\ () helper with passing in
3391 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
3392 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
3393 * the UDP header is removed. Similarly, an encap of the latter
3394 * into the former could be accompanied by a helper call to
3395 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
3396 * skb is still intended to be processed in higher layers of the
3397 * stack instead of just egressing at tc.
3398 *
3399 * There are three supported level settings at this time:
3400 *
3401 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
3402 * with CHECKSUM_UNNECESSARY.
3403 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
3404 * with CHECKSUM_UNNECESSARY.
3405 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
3406 * sets CHECKSUM_NONE to force checksum validation by the stack.
3407 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
3408 * skb->csum_level.
3409 * Return
3410 * 0 on success, or a negative error in case of failure. In the
3411 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
3412 * is returned or the error code -EACCES in case the skb is not
3413 * subject to CHECKSUM_UNNECESSARY.
3414 *
3415 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
3416 * Description
3417 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
3418 * Return
3419 * *sk* if casting is valid, or **NULL** otherwise.
3420 *
3421 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
3422 * Description
3423 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
3424 * Return
3425 * *sk* if casting is valid, or **NULL** otherwise.
3426 *
3427 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
3428 * Description
3429 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
3430 * Return
3431 * *sk* if casting is valid, or **NULL** otherwise.
3432 *
3433 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
3434 * Description
3435 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
3436 * Return
3437 * *sk* if casting is valid, or **NULL** otherwise.
3438 *
3439 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
3440 * Description
3441 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
3442 * Return
3443 * *sk* if casting is valid, or **NULL** otherwise.
3444 *
3445 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
3446 * Description
3447 * Return a user or a kernel stack in bpf program provided buffer.
3448 * To achieve this, the helper needs *task*, which is a valid
3449 * pointer to **struct task_struct**. To store the stacktrace, the
3450 * bpf program provides *buf* with a nonnegative *size*.
3451 *
3452 * The last argument, *flags*, holds the number of stack frames to
3453 * skip (from 0 to 255), masked with
3454 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3455 * the following flags:
3456 *
3457 * **BPF_F_USER_STACK**
3458 * Collect a user space stack instead of a kernel stack.
3459 * **BPF_F_USER_BUILD_ID**
3460 * Collect buildid+offset instead of ips for user stack,
3461 * only valid if **BPF_F_USER_STACK** is also specified.
3462 *
3463 * **bpf_get_task_stack**\ () can collect up to
3464 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3465 * to sufficient large buffer size. Note that
3466 * this limit can be controlled with the **sysctl** program, and
3467 * that it should be manually increased in order to profile long
3468 * user stacks (such as stacks for Java programs). To do so, use:
3469 *
3470 * ::
3471 *
3472 * # sysctl kernel.perf_event_max_stack=<new value>
3473 * Return
3474 * A non-negative value equal to or less than *size* on success,
3475 * or a negative error in case of failure.
3476 *
3477 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
3478 * Description
3479 * Load header option. Support reading a particular TCP header
3480 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
3481 *
3482 * If *flags* is 0, it will search the option from the
3483 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
3484 * has details on what skb_data contains under different
3485 * *skops*\ **->op**.
3486 *
3487 * The first byte of the *searchby_res* specifies the
3488 * kind that it wants to search.
3489 *
3490 * If the searching kind is an experimental kind
3491 * (i.e. 253 or 254 according to RFC6994). It also
3492 * needs to specify the "magic" which is either
3493 * 2 bytes or 4 bytes. It then also needs to
3494 * specify the size of the magic by using
3495 * the 2nd byte which is "kind-length" of a TCP
3496 * header option and the "kind-length" also
3497 * includes the first 2 bytes "kind" and "kind-length"
3498 * itself as a normal TCP header option also does.
3499 *
3500 * For example, to search experimental kind 254 with
3501 * 2 byte magic 0xeB9F, the searchby_res should be
3502 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
3503 *
3504 * To search for the standard window scale option (3),
3505 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
3506 * Note, kind-length must be 0 for regular option.
3507 *
3508 * Searching for No-Op (0) and End-of-Option-List (1) are
3509 * not supported.
3510 *
3511 * *len* must be at least 2 bytes which is the minimal size
3512 * of a header option.
3513 *
3514 * Supported flags:
3515 *
3516 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
3517 * saved_syn packet or the just-received syn packet.
3518 *
3519 * Return
3520 * > 0 when found, the header option is copied to *searchby_res*.
3521 * The return value is the total length copied. On failure, a
3522 * negative error code is returned:
3523 *
3524 * **-EINVAL** if a parameter is invalid.
3525 *
3526 * **-ENOMSG** if the option is not found.
3527 *
3528 * **-ENOENT** if no syn packet is available when
3529 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
3530 *
3531 * **-ENOSPC** if there is not enough space. Only *len* number of
3532 * bytes are copied.
3533 *
3534 * **-EFAULT** on failure to parse the header options in the
3535 * packet.
3536 *
3537 * **-EPERM** if the helper cannot be used under the current
3538 * *skops*\ **->op**.
3539 *
3540 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
3541 * Description
3542 * Store header option. The data will be copied
3543 * from buffer *from* with length *len* to the TCP header.
3544 *
3545 * The buffer *from* should have the whole option that
3546 * includes the kind, kind-length, and the actual
3547 * option data. The *len* must be at least kind-length
3548 * long. The kind-length does not have to be 4 byte
3549 * aligned. The kernel will take care of the padding
3550 * and setting the 4 bytes aligned value to th->doff.
3551 *
3552 * This helper will check for duplicated option
3553 * by searching the same option in the outgoing skb.
3554 *
3555 * This helper can only be called during
3556 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
3557 *
3558 * Return
3559 * 0 on success, or negative error in case of failure:
3560 *
3561 * **-EINVAL** If param is invalid.
3562 *
3563 * **-ENOSPC** if there is not enough space in the header.
3564 * Nothing has been written
3565 *
3566 * **-EEXIST** if the option already exists.
3567 *
3568 * **-EFAULT** on failrue to parse the existing header options.
3569 *
3570 * **-EPERM** if the helper cannot be used under the current
3571 * *skops*\ **->op**.
3572 *
3573 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
3574 * Description
3575 * Reserve *len* bytes for the bpf header option. The
3576 * space will be used by **bpf_store_hdr_opt**\ () later in
3577 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
3578 *
3579 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
3580 * the total number of bytes will be reserved.
3581 *
3582 * This helper can only be called during
3583 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
3584 *
3585 * Return
3586 * 0 on success, or negative error in case of failure:
3587 *
3588 * **-EINVAL** if a parameter is invalid.
3589 *
3590 * **-ENOSPC** if there is not enough space in the header.
3591 *
3592 * **-EPERM** if the helper cannot be used under the current
3593 * *skops*\ **->op**.
3594 *
3595 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
3596 * Description
3597 * Get a bpf_local_storage from an *inode*.
3598 *
3599 * Logically, it could be thought of as getting the value from
3600 * a *map* with *inode* as the **key**. From this
3601 * perspective, the usage is not much different from
3602 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
3603 * helper enforces the key must be an inode and the map must also
3604 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
3605 *
3606 * Underneath, the value is stored locally at *inode* instead of
3607 * the *map*. The *map* is used as the bpf-local-storage
3608 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3609 * searched against all bpf_local_storage residing at *inode*.
3610 *
3611 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
3612 * used such that a new bpf_local_storage will be
3613 * created if one does not exist. *value* can be used
3614 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
3615 * the initial value of a bpf_local_storage. If *value* is
3616 * **NULL**, the new bpf_local_storage will be zero initialized.
3617 * Return
3618 * A bpf_local_storage pointer is returned on success.
3619 *
3620 * **NULL** if not found or there was an error in adding
3621 * a new bpf_local_storage.
3622 *
3623 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
3624 * Description
3625 * Delete a bpf_local_storage from an *inode*.
3626 * Return
3627 * 0 on success.
3628 *
3629 * **-ENOENT** if the bpf_local_storage cannot be found.
3630 *
3631 * long bpf_d_path(struct path *path, char *buf, u32 sz)
3632 * Description
3633 * Return full path for given **struct path** object, which
3634 * needs to be the kernel BTF *path* object. The path is
3635 * returned in the provided buffer *buf* of size *sz* and
3636 * is zero terminated.
3637 *
3638 * Return
3639 * On success, the strictly positive length of the string,
3640 * including the trailing NUL character. On error, a negative
3641 * value.
3642 *
3643 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
3644 * Description
3645 * Read *size* bytes from user space address *user_ptr* and store
3646 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
3647 * Return
3648 * 0 on success, or a negative error in case of failure.
3649 *
3650 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
3651 * Description
3652 * Use BTF to store a string representation of *ptr*->ptr in *str*,
3653 * using *ptr*->type_id. This value should specify the type
3654 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
3655 * can be used to look up vmlinux BTF type ids. Traversing the
3656 * data structure using BTF, the type information and values are
3657 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
3658 * the pointer data is carried out to avoid kernel crashes during
3659 * operation. Smaller types can use string space on the stack;
3660 * larger programs can use map data to store the string
3661 * representation.
3662 *
3663 * The string can be subsequently shared with userspace via
3664 * bpf_perf_event_output() or ring buffer interfaces.
3665 * bpf_trace_printk() is to be avoided as it places too small
3666 * a limit on string size to be useful.
3667 *
3668 * *flags* is a combination of
3669 *
3670 * **BTF_F_COMPACT**
3671 * no formatting around type information
3672 * **BTF_F_NONAME**
3673 * no struct/union member names/types
3674 * **BTF_F_PTR_RAW**
3675 * show raw (unobfuscated) pointer values;
3676 * equivalent to printk specifier %px.
3677 * **BTF_F_ZERO**
3678 * show zero-valued struct/union members; they
3679 * are not displayed by default
3680 *
3681 * Return
3682 * The number of bytes that were written (or would have been
3683 * written if output had to be truncated due to string size),
3684 * or a negative error in cases of failure.
3685 *
3686 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
3687 * Description
3688 * Use BTF to write to seq_write a string representation of
3689 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
3690 * *flags* are identical to those used for bpf_snprintf_btf.
3691 * Return
3692 * 0 on success or a negative error in case of failure.
3693 *
3694 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
3695 * Description
3696 * See **bpf_get_cgroup_classid**\ () for the main description.
3697 * This helper differs from **bpf_get_cgroup_classid**\ () in that
3698 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
3699 * associated socket instead of the current process.
3700 * Return
3701 * The id is returned or 0 in case the id could not be retrieved.
3702 *
3703 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
3704 * Description
3705 * Redirect the packet to another net device of index *ifindex*
3706 * and fill in L2 addresses from neighboring subsystem. This helper
3707 * is somewhat similar to **bpf_redirect**\ (), except that it
3708 * populates L2 addresses as well, meaning, internally, the helper
3709 * relies on the neighbor lookup for the L2 address of the nexthop.
3710 *
3711 * The helper will perform a FIB lookup based on the skb's
3712 * networking header to get the address of the next hop, unless
3713 * this is supplied by the caller in the *params* argument. The
3714 * *plen* argument indicates the len of *params* and should be set
3715 * to 0 if *params* is NULL.
3716 *
3717 * The *flags* argument is reserved and must be 0. The helper is
3718 * currently only supported for tc BPF program types, and enabled
3719 * for IPv4 and IPv6 protocols.
3720 * Return
3721 * The helper returns **TC_ACT_REDIRECT** on success or
3722 * **TC_ACT_SHOT** on error.
3723 *
3724 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
3725 * Description
3726 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
3727 * pointer to the percpu kernel variable on *cpu*. A ksym is an
3728 * extern variable decorated with '__ksym'. For ksym, there is a
3729 * global var (either static or global) defined of the same name
3730 * in the kernel. The ksym is percpu if the global var is percpu.
3731 * The returned pointer points to the global percpu var on *cpu*.
3732 *
3733 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
3734 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
3735 * happens if *cpu* is larger than nr_cpu_ids. The caller of
3736 * bpf_per_cpu_ptr() must check the returned value.
3737 * Return
3738 * A pointer pointing to the kernel percpu variable on *cpu*, or
3739 * NULL, if *cpu* is invalid.
3740 *
3741 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
3742 * Description
3743 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
3744 * pointer to the percpu kernel variable on this cpu. See the
3745 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
3746 *
3747 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
3748 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
3749 * never return NULL.
3750 * Return
3751 * A pointer pointing to the kernel percpu variable on this cpu.
3752 *
3753 * long bpf_redirect_peer(u32 ifindex, u64 flags)
3754 * Description
3755 * Redirect the packet to another net device of index *ifindex*.
3756 * This helper is somewhat similar to **bpf_redirect**\ (), except
3757 * that the redirection happens to the *ifindex*' peer device and
3758 * the netns switch takes place from ingress to ingress without
3759 * going through the CPU's backlog queue.
3760 *
3761 * The *flags* argument is reserved and must be 0. The helper is
3762 * currently only supported for tc BPF program types at the ingress
3763 * hook and for veth device types. The peer device must reside in a
3764 * different network namespace.
3765 * Return
3766 * The helper returns **TC_ACT_REDIRECT** on success or
3767 * **TC_ACT_SHOT** on error.
3768 *
3769 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
3770 * Description
3771 * Get a bpf_local_storage from the *task*.
3772 *
3773 * Logically, it could be thought of as getting the value from
3774 * a *map* with *task* as the **key**. From this
3775 * perspective, the usage is not much different from
3776 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
3777 * helper enforces the key must be an task_struct and the map must also
3778 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
3779 *
3780 * Underneath, the value is stored locally at *task* instead of
3781 * the *map*. The *map* is used as the bpf-local-storage
3782 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3783 * searched against all bpf_local_storage residing at *task*.
3784 *
3785 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
3786 * used such that a new bpf_local_storage will be
3787 * created if one does not exist. *value* can be used
3788 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
3789 * the initial value of a bpf_local_storage. If *value* is
3790 * **NULL**, the new bpf_local_storage will be zero initialized.
3791 * Return
3792 * A bpf_local_storage pointer is returned on success.
3793 *
3794 * **NULL** if not found or there was an error in adding
3795 * a new bpf_local_storage.
3796 *
3797 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
3798 * Description
3799 * Delete a bpf_local_storage from a *task*.
3800 * Return
3801 * 0 on success.
3802 *
3803 * **-ENOENT** if the bpf_local_storage cannot be found.
3804 *
3805 * struct task_struct *bpf_get_current_task_btf(void)
3806 * Description
3807 * Return a BTF pointer to the "current" task.
3808 * This pointer can also be used in helpers that accept an
3809 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
3810 * Return
3811 * Pointer to the current task.
3812 *
3813 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
3814 * Description
3815 * Set or clear certain options on *bprm*:
3816 *
3817 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
3818 * which sets the **AT_SECURE** auxv for glibc. The bit
3819 * is cleared if the flag is not specified.
3820 * Return
3821 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
3822 *
3823 * u64 bpf_ktime_get_coarse_ns(void)
3824 * Description
3825 * Return a coarse-grained version of the time elapsed since
3826 * system boot, in nanoseconds. Does not include time the system
3827 * was suspended.
3828 *
3829 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
3830 * Return
3831 * Current *ktime*.
3832 *
3833 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
3834 * Description
3835 * Returns the stored IMA hash of the *inode* (if it's avaialable).
3836 * If the hash is larger than *size*, then only *size*
3837 * bytes will be copied to *dst*
3838 * Return
3839 * The **hash_algo** is returned on success,
3840 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
3841 * invalid arguments are passed.
3842 *
3843 * struct socket *bpf_sock_from_file(struct file *file)
3844 * Description
3845 * If the given file represents a socket, returns the associated
3846 * socket.
3847 * Return
3848 * A pointer to a struct socket on success or NULL if the file is
3849 * not a socket.
3850 *
3851 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
3852 * Description
3853
3854 * Check ctx packet size against exceeding MTU of net device (based
3855 * on *ifindex*). This helper will likely be used in combination
3856 * with helpers that adjust/change the packet size.
3857 *
3858 * The argument *len_diff* can be used for querying with a planned
3859 * size change. This allows to check MTU prior to changing packet
3860 * ctx. Providing an *len_diff* adjustment that is larger than the
3861 * actual packet size (resulting in negative packet size) will in
3862 * principle not exceed the MTU, why it is not considered a
3863 * failure. Other BPF-helpers are needed for performing the
3864 * planned size change, why the responsability for catch a negative
3865 * packet size belong in those helpers.
3866 *
3867 * Specifying *ifindex* zero means the MTU check is performed
3868 * against the current net device. This is practical if this isn't
3869 * used prior to redirect.
3870 *
3871 * The Linux kernel route table can configure MTUs on a more
3872 * specific per route level, which is not provided by this helper.
3873 * For route level MTU checks use the **bpf_fib_lookup**\ ()
3874 * helper.
3875 *
3876 * *ctx* is either **struct xdp_md** for XDP programs or
3877 * **struct sk_buff** for tc cls_act programs.
3878 *
3879 * The *flags* argument can be a combination of one or more of the
3880 * following values:
3881 *
3882 * **BPF_MTU_CHK_SEGS**
3883 * This flag will only works for *ctx* **struct sk_buff**.
3884 * If packet context contains extra packet segment buffers
3885 * (often knows as GSO skb), then MTU check is harder to
3886 * check at this point, because in transmit path it is
3887 * possible for the skb packet to get re-segmented
3888 * (depending on net device features). This could still be
3889 * a MTU violation, so this flag enables performing MTU
3890 * check against segments, with a different violation
3891 * return code to tell it apart. Check cannot use len_diff.
3892 *
3893 * On return *mtu_len* pointer contains the MTU value of the net
3894 * device. Remember the net device configured MTU is the L3 size,
3895 * which is returned here and XDP and TX length operate at L2.
3896 * Helper take this into account for you, but remember when using
3897 * MTU value in your BPF-code. On input *mtu_len* must be a valid
3898 * pointer and be initialized (to zero), else verifier will reject
3899 * BPF program.
3900 *
3901 * Return
3902 * * 0 on success, and populate MTU value in *mtu_len* pointer.
3903 *
3904 * * < 0 if any input argument is invalid (*mtu_len* not updated)
3905 *
3906 * MTU violations return positive values, but also populate MTU
3907 * value in *mtu_len* pointer, as this can be needed for
3908 * implementing PMTU handing:
3909 *
3910 * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
3911 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
3912 *
3913 */
3914 #define __BPF_FUNC_MAPPER(FN) \
3915 FN(unspec), \
3916 FN(map_lookup_elem), \
3917 FN(map_update_elem), \
3918 FN(map_delete_elem), \
3919 FN(probe_read), \
3920 FN(ktime_get_ns), \
3921 FN(trace_printk), \
3922 FN(get_prandom_u32), \
3923 FN(get_smp_processor_id), \
3924 FN(skb_store_bytes), \
3925 FN(l3_csum_replace), \
3926 FN(l4_csum_replace), \
3927 FN(tail_call), \
3928 FN(clone_redirect), \
3929 FN(get_current_pid_tgid), \
3930 FN(get_current_uid_gid), \
3931 FN(get_current_comm), \
3932 FN(get_cgroup_classid), \
3933 FN(skb_vlan_push), \
3934 FN(skb_vlan_pop), \
3935 FN(skb_get_tunnel_key), \
3936 FN(skb_set_tunnel_key), \
3937 FN(perf_event_read), \
3938 FN(redirect), \
3939 FN(get_route_realm), \
3940 FN(perf_event_output), \
3941 FN(skb_load_bytes), \
3942 FN(get_stackid), \
3943 FN(csum_diff), \
3944 FN(skb_get_tunnel_opt), \
3945 FN(skb_set_tunnel_opt), \
3946 FN(skb_change_proto), \
3947 FN(skb_change_type), \
3948 FN(skb_under_cgroup), \
3949 FN(get_hash_recalc), \
3950 FN(get_current_task), \
3951 FN(probe_write_user), \
3952 FN(current_task_under_cgroup), \
3953 FN(skb_change_tail), \
3954 FN(skb_pull_data), \
3955 FN(csum_update), \
3956 FN(set_hash_invalid), \
3957 FN(get_numa_node_id), \
3958 FN(skb_change_head), \
3959 FN(xdp_adjust_head), \
3960 FN(probe_read_str), \
3961 FN(get_socket_cookie), \
3962 FN(get_socket_uid), \
3963 FN(set_hash), \
3964 FN(setsockopt), \
3965 FN(skb_adjust_room), \
3966 FN(redirect_map), \
3967 FN(sk_redirect_map), \
3968 FN(sock_map_update), \
3969 FN(xdp_adjust_meta), \
3970 FN(perf_event_read_value), \
3971 FN(perf_prog_read_value), \
3972 FN(getsockopt), \
3973 FN(override_return), \
3974 FN(sock_ops_cb_flags_set), \
3975 FN(msg_redirect_map), \
3976 FN(msg_apply_bytes), \
3977 FN(msg_cork_bytes), \
3978 FN(msg_pull_data), \
3979 FN(bind), \
3980 FN(xdp_adjust_tail), \
3981 FN(skb_get_xfrm_state), \
3982 FN(get_stack), \
3983 FN(skb_load_bytes_relative), \
3984 FN(fib_lookup), \
3985 FN(sock_hash_update), \
3986 FN(msg_redirect_hash), \
3987 FN(sk_redirect_hash), \
3988 FN(lwt_push_encap), \
3989 FN(lwt_seg6_store_bytes), \
3990 FN(lwt_seg6_adjust_srh), \
3991 FN(lwt_seg6_action), \
3992 FN(rc_repeat), \
3993 FN(rc_keydown), \
3994 FN(skb_cgroup_id), \
3995 FN(get_current_cgroup_id), \
3996 FN(get_local_storage), \
3997 FN(sk_select_reuseport), \
3998 FN(skb_ancestor_cgroup_id), \
3999 FN(sk_lookup_tcp), \
4000 FN(sk_lookup_udp), \
4001 FN(sk_release), \
4002 FN(map_push_elem), \
4003 FN(map_pop_elem), \
4004 FN(map_peek_elem), \
4005 FN(msg_push_data), \
4006 FN(msg_pop_data), \
4007 FN(rc_pointer_rel), \
4008 FN(spin_lock), \
4009 FN(spin_unlock), \
4010 FN(sk_fullsock), \
4011 FN(tcp_sock), \
4012 FN(skb_ecn_set_ce), \
4013 FN(get_listener_sock), \
4014 FN(skc_lookup_tcp), \
4015 FN(tcp_check_syncookie), \
4016 FN(sysctl_get_name), \
4017 FN(sysctl_get_current_value), \
4018 FN(sysctl_get_new_value), \
4019 FN(sysctl_set_new_value), \
4020 FN(strtol), \
4021 FN(strtoul), \
4022 FN(sk_storage_get), \
4023 FN(sk_storage_delete), \
4024 FN(send_signal), \
4025 FN(tcp_gen_syncookie), \
4026 FN(skb_output), \
4027 FN(probe_read_user), \
4028 FN(probe_read_kernel), \
4029 FN(probe_read_user_str), \
4030 FN(probe_read_kernel_str), \
4031 FN(tcp_send_ack), \
4032 FN(send_signal_thread), \
4033 FN(jiffies64), \
4034 FN(read_branch_records), \
4035 FN(get_ns_current_pid_tgid), \
4036 FN(xdp_output), \
4037 FN(get_netns_cookie), \
4038 FN(get_current_ancestor_cgroup_id), \
4039 FN(sk_assign), \
4040 FN(ktime_get_boot_ns), \
4041 FN(seq_printf), \
4042 FN(seq_write), \
4043 FN(sk_cgroup_id), \
4044 FN(sk_ancestor_cgroup_id), \
4045 FN(ringbuf_output), \
4046 FN(ringbuf_reserve), \
4047 FN(ringbuf_submit), \
4048 FN(ringbuf_discard), \
4049 FN(ringbuf_query), \
4050 FN(csum_level), \
4051 FN(skc_to_tcp6_sock), \
4052 FN(skc_to_tcp_sock), \
4053 FN(skc_to_tcp_timewait_sock), \
4054 FN(skc_to_tcp_request_sock), \
4055 FN(skc_to_udp6_sock), \
4056 FN(get_task_stack), \
4057 FN(load_hdr_opt), \
4058 FN(store_hdr_opt), \
4059 FN(reserve_hdr_opt), \
4060 FN(inode_storage_get), \
4061 FN(inode_storage_delete), \
4062 FN(d_path), \
4063 FN(copy_from_user), \
4064 FN(snprintf_btf), \
4065 FN(seq_printf_btf), \
4066 FN(skb_cgroup_classid), \
4067 FN(redirect_neigh), \
4068 FN(per_cpu_ptr), \
4069 FN(this_cpu_ptr), \
4070 FN(redirect_peer), \
4071 FN(task_storage_get), \
4072 FN(task_storage_delete), \
4073 FN(get_current_task_btf), \
4074 FN(bprm_opts_set), \
4075 FN(ktime_get_coarse_ns), \
4076 FN(ima_inode_hash), \
4077 FN(sock_from_file), \
4078 FN(check_mtu), \
4079 /* */
4080
4081 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
4082 * function eBPF program intends to call
4083 */
4084 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
4085 enum bpf_func_id {
4086 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
4087 __BPF_FUNC_MAX_ID,
4088 };
4089 #undef __BPF_ENUM_FN
4090
4091 /* All flags used by eBPF helper functions, placed here. */
4092
4093 /* BPF_FUNC_skb_store_bytes flags. */
4094 enum {
4095 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
4096 BPF_F_INVALIDATE_HASH = (1ULL << 1),
4097 };
4098
4099 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
4100 * First 4 bits are for passing the header field size.
4101 */
4102 enum {
4103 BPF_F_HDR_FIELD_MASK = 0xfULL,
4104 };
4105
4106 /* BPF_FUNC_l4_csum_replace flags. */
4107 enum {
4108 BPF_F_PSEUDO_HDR = (1ULL << 4),
4109 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
4110 BPF_F_MARK_ENFORCE = (1ULL << 6),
4111 };
4112
4113 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
4114 enum {
4115 BPF_F_INGRESS = (1ULL << 0),
4116 };
4117
4118 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
4119 enum {
4120 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
4121 };
4122
4123 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
4124 enum {
4125 BPF_F_SKIP_FIELD_MASK = 0xffULL,
4126 BPF_F_USER_STACK = (1ULL << 8),
4127 /* flags used by BPF_FUNC_get_stackid only. */
4128 BPF_F_FAST_STACK_CMP = (1ULL << 9),
4129 BPF_F_REUSE_STACKID = (1ULL << 10),
4130 /* flags used by BPF_FUNC_get_stack only. */
4131 BPF_F_USER_BUILD_ID = (1ULL << 11),
4132 };
4133
4134 /* BPF_FUNC_skb_set_tunnel_key flags. */
4135 enum {
4136 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
4137 BPF_F_DONT_FRAGMENT = (1ULL << 2),
4138 BPF_F_SEQ_NUMBER = (1ULL << 3),
4139 };
4140
4141 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
4142 * BPF_FUNC_perf_event_read_value flags.
4143 */
4144 enum {
4145 BPF_F_INDEX_MASK = 0xffffffffULL,
4146 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
4147 /* BPF_FUNC_perf_event_output for sk_buff input context. */
4148 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
4149 };
4150
4151 /* Current network namespace */
4152 enum {
4153 BPF_F_CURRENT_NETNS = (-1L),
4154 };
4155
4156 /* BPF_FUNC_csum_level level values. */
4157 enum {
4158 BPF_CSUM_LEVEL_QUERY,
4159 BPF_CSUM_LEVEL_INC,
4160 BPF_CSUM_LEVEL_DEC,
4161 BPF_CSUM_LEVEL_RESET,
4162 };
4163
4164 /* BPF_FUNC_skb_adjust_room flags. */
4165 enum {
4166 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
4167 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
4168 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
4169 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
4170 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
4171 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
4172 };
4173
4174 enum {
4175 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
4176 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
4177 };
4178
4179 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
4180 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
4181 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
4182
4183 /* BPF_FUNC_sysctl_get_name flags. */
4184 enum {
4185 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
4186 };
4187
4188 /* BPF_FUNC_<kernel_obj>_storage_get flags */
4189 enum {
4190 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
4191 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
4192 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
4193 */
4194 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
4195 };
4196
4197 /* BPF_FUNC_read_branch_records flags. */
4198 enum {
4199 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
4200 };
4201
4202 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
4203 * BPF_FUNC_bpf_ringbuf_output flags.
4204 */
4205 enum {
4206 BPF_RB_NO_WAKEUP = (1ULL << 0),
4207 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
4208 };
4209
4210 /* BPF_FUNC_bpf_ringbuf_query flags */
4211 enum {
4212 BPF_RB_AVAIL_DATA = 0,
4213 BPF_RB_RING_SIZE = 1,
4214 BPF_RB_CONS_POS = 2,
4215 BPF_RB_PROD_POS = 3,
4216 };
4217
4218 /* BPF ring buffer constants */
4219 enum {
4220 BPF_RINGBUF_BUSY_BIT = (1U << 31),
4221 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
4222 BPF_RINGBUF_HDR_SZ = 8,
4223 };
4224
4225 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
4226 enum {
4227 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
4228 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
4229 };
4230
4231 /* Mode for BPF_FUNC_skb_adjust_room helper. */
4232 enum bpf_adj_room_mode {
4233 BPF_ADJ_ROOM_NET,
4234 BPF_ADJ_ROOM_MAC,
4235 };
4236
4237 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
4238 enum bpf_hdr_start_off {
4239 BPF_HDR_START_MAC,
4240 BPF_HDR_START_NET,
4241 };
4242
4243 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
4244 enum bpf_lwt_encap_mode {
4245 BPF_LWT_ENCAP_SEG6,
4246 BPF_LWT_ENCAP_SEG6_INLINE,
4247 BPF_LWT_ENCAP_IP,
4248 };
4249
4250 /* Flags for bpf_bprm_opts_set helper */
4251 enum {
4252 BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
4253 };
4254
4255 #define __bpf_md_ptr(type, name) \
4256 union { \
4257 type name; \
4258 __u64 :64; \
4259 } __attribute__((aligned(8)))
4260
4261 /* user accessible mirror of in-kernel sk_buff.
4262 * new fields can only be added to the end of this structure
4263 */
4264 struct __sk_buff {
4265 __u32 len;
4266 __u32 pkt_type;
4267 __u32 mark;
4268 __u32 queue_mapping;
4269 __u32 protocol;
4270 __u32 vlan_present;
4271 __u32 vlan_tci;
4272 __u32 vlan_proto;
4273 __u32 priority;
4274 __u32 ingress_ifindex;
4275 __u32 ifindex;
4276 __u32 tc_index;
4277 __u32 cb[5];
4278 __u32 hash;
4279 __u32 tc_classid;
4280 __u32 data;
4281 __u32 data_end;
4282 __u32 napi_id;
4283
4284 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
4285 __u32 family;
4286 __u32 remote_ip4; /* Stored in network byte order */
4287 __u32 local_ip4; /* Stored in network byte order */
4288 __u32 remote_ip6[4]; /* Stored in network byte order */
4289 __u32 local_ip6[4]; /* Stored in network byte order */
4290 __u32 remote_port; /* Stored in network byte order */
4291 __u32 local_port; /* stored in host byte order */
4292 /* ... here. */
4293
4294 __u32 data_meta;
4295 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
4296 __u64 tstamp;
4297 __u32 wire_len;
4298 __u32 gso_segs;
4299 __bpf_md_ptr(struct bpf_sock *, sk);
4300 __u32 gso_size;
4301 };
4302
4303 struct bpf_tunnel_key {
4304 __u32 tunnel_id;
4305 union {
4306 __u32 remote_ipv4;
4307 __u32 remote_ipv6[4];
4308 };
4309 __u8 tunnel_tos;
4310 __u8 tunnel_ttl;
4311 __u16 tunnel_ext; /* Padding, future use. */
4312 __u32 tunnel_label;
4313 };
4314
4315 /* user accessible mirror of in-kernel xfrm_state.
4316 * new fields can only be added to the end of this structure
4317 */
4318 struct bpf_xfrm_state {
4319 __u32 reqid;
4320 __u32 spi; /* Stored in network byte order */
4321 __u16 family;
4322 __u16 ext; /* Padding, future use. */
4323 union {
4324 __u32 remote_ipv4; /* Stored in network byte order */
4325 __u32 remote_ipv6[4]; /* Stored in network byte order */
4326 };
4327 };
4328
4329 /* Generic BPF return codes which all BPF program types may support.
4330 * The values are binary compatible with their TC_ACT_* counter-part to
4331 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
4332 * programs.
4333 *
4334 * XDP is handled seprately, see XDP_*.
4335 */
4336 enum bpf_ret_code {
4337 BPF_OK = 0,
4338 /* 1 reserved */
4339 BPF_DROP = 2,
4340 /* 3-6 reserved */
4341 BPF_REDIRECT = 7,
4342 /* >127 are reserved for prog type specific return codes.
4343 *
4344 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
4345 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
4346 * changed and should be routed based on its new L3 header.
4347 * (This is an L3 redirect, as opposed to L2 redirect
4348 * represented by BPF_REDIRECT above).
4349 */
4350 BPF_LWT_REROUTE = 128,
4351 };
4352
4353 struct bpf_sock {
4354 __u32 bound_dev_if;
4355 __u32 family;
4356 __u32 type;
4357 __u32 protocol;
4358 __u32 mark;
4359 __u32 priority;
4360 /* IP address also allows 1 and 2 bytes access */
4361 __u32 src_ip4;
4362 __u32 src_ip6[4];
4363 __u32 src_port; /* host byte order */
4364 __u32 dst_port; /* network byte order */
4365 __u32 dst_ip4;
4366 __u32 dst_ip6[4];
4367 __u32 state;
4368 __s32 rx_queue_mapping;
4369 };
4370
4371 struct bpf_tcp_sock {
4372 __u32 snd_cwnd; /* Sending congestion window */
4373 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
4374 __u32 rtt_min;
4375 __u32 snd_ssthresh; /* Slow start size threshold */
4376 __u32 rcv_nxt; /* What we want to receive next */
4377 __u32 snd_nxt; /* Next sequence we send */
4378 __u32 snd_una; /* First byte we want an ack for */
4379 __u32 mss_cache; /* Cached effective mss, not including SACKS */
4380 __u32 ecn_flags; /* ECN status bits. */
4381 __u32 rate_delivered; /* saved rate sample: packets delivered */
4382 __u32 rate_interval_us; /* saved rate sample: time elapsed */
4383 __u32 packets_out; /* Packets which are "in flight" */
4384 __u32 retrans_out; /* Retransmitted packets out */
4385 __u32 total_retrans; /* Total retransmits for entire connection */
4386 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
4387 * total number of segments in.
4388 */
4389 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
4390 * total number of data segments in.
4391 */
4392 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
4393 * The total number of segments sent.
4394 */
4395 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
4396 * total number of data segments sent.
4397 */
4398 __u32 lost_out; /* Lost packets */
4399 __u32 sacked_out; /* SACK'd packets */
4400 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
4401 * sum(delta(rcv_nxt)), or how many bytes
4402 * were acked.
4403 */
4404 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
4405 * sum(delta(snd_una)), or how many bytes
4406 * were acked.
4407 */
4408 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
4409 * total number of DSACK blocks received
4410 */
4411 __u32 delivered; /* Total data packets delivered incl. rexmits */
4412 __u32 delivered_ce; /* Like the above but only ECE marked packets */
4413 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
4414 };
4415
4416 struct bpf_sock_tuple {
4417 union {
4418 struct {
4419 __be32 saddr;
4420 __be32 daddr;
4421 __be16 sport;
4422 __be16 dport;
4423 } ipv4;
4424 struct {
4425 __be32 saddr[4];
4426 __be32 daddr[4];
4427 __be16 sport;
4428 __be16 dport;
4429 } ipv6;
4430 };
4431 };
4432
4433 struct bpf_xdp_sock {
4434 __u32 queue_id;
4435 };
4436
4437 #define XDP_PACKET_HEADROOM 256
4438
4439 /* User return codes for XDP prog type.
4440 * A valid XDP program must return one of these defined values. All other
4441 * return codes are reserved for future use. Unknown return codes will
4442 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
4443 */
4444 enum xdp_action {
4445 XDP_ABORTED = 0,
4446 XDP_DROP,
4447 XDP_PASS,
4448 XDP_TX,
4449 XDP_REDIRECT,
4450 };
4451
4452 /* user accessible metadata for XDP packet hook
4453 * new fields must be added to the end of this structure
4454 */
4455 struct xdp_md {
4456 __u32 data;
4457 __u32 data_end;
4458 __u32 data_meta;
4459 /* Below access go through struct xdp_rxq_info */
4460 __u32 ingress_ifindex; /* rxq->dev->ifindex */
4461 __u32 rx_queue_index; /* rxq->queue_index */
4462
4463 __u32 egress_ifindex; /* txq->dev->ifindex */
4464 };
4465
4466 /* DEVMAP map-value layout
4467 *
4468 * The struct data-layout of map-value is a configuration interface.
4469 * New members can only be added to the end of this structure.
4470 */
4471 struct bpf_devmap_val {
4472 __u32 ifindex; /* device index */
4473 union {
4474 int fd; /* prog fd on map write */
4475 __u32 id; /* prog id on map read */
4476 } bpf_prog;
4477 };
4478
4479 /* CPUMAP map-value layout
4480 *
4481 * The struct data-layout of map-value is a configuration interface.
4482 * New members can only be added to the end of this structure.
4483 */
4484 struct bpf_cpumap_val {
4485 __u32 qsize; /* queue size to remote target CPU */
4486 union {
4487 int fd; /* prog fd on map write */
4488 __u32 id; /* prog id on map read */
4489 } bpf_prog;
4490 };
4491
4492 enum sk_action {
4493 SK_DROP = 0,
4494 SK_PASS,
4495 };
4496
4497 /* user accessible metadata for SK_MSG packet hook, new fields must
4498 * be added to the end of this structure
4499 */
4500 struct sk_msg_md {
4501 __bpf_md_ptr(void *, data);
4502 __bpf_md_ptr(void *, data_end);
4503
4504 __u32 family;
4505 __u32 remote_ip4; /* Stored in network byte order */
4506 __u32 local_ip4; /* Stored in network byte order */
4507 __u32 remote_ip6[4]; /* Stored in network byte order */
4508 __u32 local_ip6[4]; /* Stored in network byte order */
4509 __u32 remote_port; /* Stored in network byte order */
4510 __u32 local_port; /* stored in host byte order */
4511 __u32 size; /* Total size of sk_msg */
4512
4513 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
4514 };
4515
4516 struct sk_reuseport_md {
4517 /*
4518 * Start of directly accessible data. It begins from
4519 * the tcp/udp header.
4520 */
4521 __bpf_md_ptr(void *, data);
4522 /* End of directly accessible data */
4523 __bpf_md_ptr(void *, data_end);
4524 /*
4525 * Total length of packet (starting from the tcp/udp header).
4526 * Note that the directly accessible bytes (data_end - data)
4527 * could be less than this "len". Those bytes could be
4528 * indirectly read by a helper "bpf_skb_load_bytes()".
4529 */
4530 __u32 len;
4531 /*
4532 * Eth protocol in the mac header (network byte order). e.g.
4533 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
4534 */
4535 __u32 eth_protocol;
4536 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
4537 __u32 bind_inany; /* Is sock bound to an INANY address? */
4538 __u32 hash; /* A hash of the packet 4 tuples */
4539 };
4540
4541 #define BPF_TAG_SIZE 8
4542
4543 struct bpf_prog_info {
4544 __u32 type;
4545 __u32 id;
4546 __u8 tag[BPF_TAG_SIZE];
4547 __u32 jited_prog_len;
4548 __u32 xlated_prog_len;
4549 __aligned_u64 jited_prog_insns;
4550 __aligned_u64 xlated_prog_insns;
4551 __u64 load_time; /* ns since boottime */
4552 __u32 created_by_uid;
4553 __u32 nr_map_ids;
4554 __aligned_u64 map_ids;
4555 char name[BPF_OBJ_NAME_LEN];
4556 __u32 ifindex;
4557 __u32 gpl_compatible:1;
4558 __u32 :31; /* alignment pad */
4559 __u64 netns_dev;
4560 __u64 netns_ino;
4561 __u32 nr_jited_ksyms;
4562 __u32 nr_jited_func_lens;
4563 __aligned_u64 jited_ksyms;
4564 __aligned_u64 jited_func_lens;
4565 __u32 btf_id;
4566 __u32 func_info_rec_size;
4567 __aligned_u64 func_info;
4568 __u32 nr_func_info;
4569 __u32 nr_line_info;
4570 __aligned_u64 line_info;
4571 __aligned_u64 jited_line_info;
4572 __u32 nr_jited_line_info;
4573 __u32 line_info_rec_size;
4574 __u32 jited_line_info_rec_size;
4575 __u32 nr_prog_tags;
4576 __aligned_u64 prog_tags;
4577 __u64 run_time_ns;
4578 __u64 run_cnt;
4579 __u64 recursion_misses;
4580 } __attribute__((aligned(8)));
4581
4582 struct bpf_map_info {
4583 __u32 type;
4584 __u32 id;
4585 __u32 key_size;
4586 __u32 value_size;
4587 __u32 max_entries;
4588 __u32 map_flags;
4589 char name[BPF_OBJ_NAME_LEN];
4590 __u32 ifindex;
4591 __u32 btf_vmlinux_value_type_id;
4592 __u64 netns_dev;
4593 __u64 netns_ino;
4594 __u32 btf_id;
4595 __u32 btf_key_type_id;
4596 __u32 btf_value_type_id;
4597 } __attribute__((aligned(8)));
4598
4599 struct bpf_btf_info {
4600 __aligned_u64 btf;
4601 __u32 btf_size;
4602 __u32 id;
4603 __aligned_u64 name;
4604 __u32 name_len;
4605 __u32 kernel_btf;
4606 } __attribute__((aligned(8)));
4607
4608 struct bpf_link_info {
4609 __u32 type;
4610 __u32 id;
4611 __u32 prog_id;
4612 union {
4613 struct {
4614 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
4615 __u32 tp_name_len; /* in/out: tp_name buffer len */
4616 } raw_tracepoint;
4617 struct {
4618 __u32 attach_type;
4619 } tracing;
4620 struct {
4621 __u64 cgroup_id;
4622 __u32 attach_type;
4623 } cgroup;
4624 struct {
4625 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
4626 __u32 target_name_len; /* in/out: target_name buffer len */
4627 union {
4628 struct {
4629 __u32 map_id;
4630 } map;
4631 };
4632 } iter;
4633 struct {
4634 __u32 netns_ino;
4635 __u32 attach_type;
4636 } netns;
4637 struct {
4638 __u32 ifindex;
4639 } xdp;
4640 };
4641 } __attribute__((aligned(8)));
4642
4643 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
4644 * by user and intended to be used by socket (e.g. to bind to, depends on
4645 * attach type).
4646 */
4647 struct bpf_sock_addr {
4648 __u32 user_family; /* Allows 4-byte read, but no write. */
4649 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4650 * Stored in network byte order.
4651 */
4652 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4653 * Stored in network byte order.
4654 */
4655 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
4656 * Stored in network byte order
4657 */
4658 __u32 family; /* Allows 4-byte read, but no write */
4659 __u32 type; /* Allows 4-byte read, but no write */
4660 __u32 protocol; /* Allows 4-byte read, but no write */
4661 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4662 * Stored in network byte order.
4663 */
4664 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4665 * Stored in network byte order.
4666 */
4667 __bpf_md_ptr(struct bpf_sock *, sk);
4668 };
4669
4670 /* User bpf_sock_ops struct to access socket values and specify request ops
4671 * and their replies.
4672 * Some of this fields are in network (bigendian) byte order and may need
4673 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
4674 * New fields can only be added at the end of this structure
4675 */
4676 struct bpf_sock_ops {
4677 __u32 op;
4678 union {
4679 __u32 args[4]; /* Optionally passed to bpf program */
4680 __u32 reply; /* Returned by bpf program */
4681 __u32 replylong[4]; /* Optionally returned by bpf prog */
4682 };
4683 __u32 family;
4684 __u32 remote_ip4; /* Stored in network byte order */
4685 __u32 local_ip4; /* Stored in network byte order */
4686 __u32 remote_ip6[4]; /* Stored in network byte order */
4687 __u32 local_ip6[4]; /* Stored in network byte order */
4688 __u32 remote_port; /* Stored in network byte order */
4689 __u32 local_port; /* stored in host byte order */
4690 __u32 is_fullsock; /* Some TCP fields are only valid if
4691 * there is a full socket. If not, the
4692 * fields read as zero.
4693 */
4694 __u32 snd_cwnd;
4695 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
4696 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
4697 __u32 state;
4698 __u32 rtt_min;
4699 __u32 snd_ssthresh;
4700 __u32 rcv_nxt;
4701 __u32 snd_nxt;
4702 __u32 snd_una;
4703 __u32 mss_cache;
4704 __u32 ecn_flags;
4705 __u32 rate_delivered;
4706 __u32 rate_interval_us;
4707 __u32 packets_out;
4708 __u32 retrans_out;
4709 __u32 total_retrans;
4710 __u32 segs_in;
4711 __u32 data_segs_in;
4712 __u32 segs_out;
4713 __u32 data_segs_out;
4714 __u32 lost_out;
4715 __u32 sacked_out;
4716 __u32 sk_txhash;
4717 __u64 bytes_received;
4718 __u64 bytes_acked;
4719 __bpf_md_ptr(struct bpf_sock *, sk);
4720 /* [skb_data, skb_data_end) covers the whole TCP header.
4721 *
4722 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
4723 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
4724 * header has not been written.
4725 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
4726 * been written so far.
4727 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
4728 * the 3WHS.
4729 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
4730 * the 3WHS.
4731 *
4732 * bpf_load_hdr_opt() can also be used to read a particular option.
4733 */
4734 __bpf_md_ptr(void *, skb_data);
4735 __bpf_md_ptr(void *, skb_data_end);
4736 __u32 skb_len; /* The total length of a packet.
4737 * It includes the header, options,
4738 * and payload.
4739 */
4740 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
4741 * an easy way to check for tcp_flags
4742 * without parsing skb_data.
4743 *
4744 * In particular, the skb_tcp_flags
4745 * will still be available in
4746 * BPF_SOCK_OPS_HDR_OPT_LEN even though
4747 * the outgoing header has not
4748 * been written yet.
4749 */
4750 };
4751
4752 /* Definitions for bpf_sock_ops_cb_flags */
4753 enum {
4754 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
4755 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
4756 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
4757 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
4758 /* Call bpf for all received TCP headers. The bpf prog will be
4759 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4760 *
4761 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4762 * for the header option related helpers that will be useful
4763 * to the bpf programs.
4764 *
4765 * It could be used at the client/active side (i.e. connect() side)
4766 * when the server told it that the server was in syncookie
4767 * mode and required the active side to resend the bpf-written
4768 * options. The active side can keep writing the bpf-options until
4769 * it received a valid packet from the server side to confirm
4770 * the earlier packet (and options) has been received. The later
4771 * example patch is using it like this at the active side when the
4772 * server is in syncookie mode.
4773 *
4774 * The bpf prog will usually turn this off in the common cases.
4775 */
4776 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
4777 /* Call bpf when kernel has received a header option that
4778 * the kernel cannot handle. The bpf prog will be called under
4779 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
4780 *
4781 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4782 * for the header option related helpers that will be useful
4783 * to the bpf programs.
4784 */
4785 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
4786 /* Call bpf when the kernel is writing header options for the
4787 * outgoing packet. The bpf prog will first be called
4788 * to reserve space in a skb under
4789 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
4790 * the bpf prog will be called to write the header option(s)
4791 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4792 *
4793 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
4794 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
4795 * related helpers that will be useful to the bpf programs.
4796 *
4797 * The kernel gets its chance to reserve space and write
4798 * options first before the BPF program does.
4799 */
4800 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
4801 /* Mask of all currently supported cb flags */
4802 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
4803 };
4804
4805 /* List of known BPF sock_ops operators.
4806 * New entries can only be added at the end
4807 */
4808 enum {
4809 BPF_SOCK_OPS_VOID,
4810 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
4811 * -1 if default value should be used
4812 */
4813 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
4814 * window (in packets) or -1 if default
4815 * value should be used
4816 */
4817 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
4818 * active connection is initialized
4819 */
4820 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
4821 * active connection is
4822 * established
4823 */
4824 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
4825 * passive connection is
4826 * established
4827 */
4828 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
4829 * needs ECN
4830 */
4831 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
4832 * based on the path and may be
4833 * dependent on the congestion control
4834 * algorithm. In general it indicates
4835 * a congestion threshold. RTTs above
4836 * this indicate congestion
4837 */
4838 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
4839 * Arg1: value of icsk_retransmits
4840 * Arg2: value of icsk_rto
4841 * Arg3: whether RTO has expired
4842 */
4843 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
4844 * Arg1: sequence number of 1st byte
4845 * Arg2: # segments
4846 * Arg3: return value of
4847 * tcp_transmit_skb (0 => success)
4848 */
4849 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
4850 * Arg1: old_state
4851 * Arg2: new_state
4852 */
4853 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
4854 * socket transition to LISTEN state.
4855 */
4856 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
4857 */
4858 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
4859 * It will be called to handle
4860 * the packets received at
4861 * an already established
4862 * connection.
4863 *
4864 * sock_ops->skb_data:
4865 * Referring to the received skb.
4866 * It covers the TCP header only.
4867 *
4868 * bpf_load_hdr_opt() can also
4869 * be used to search for a
4870 * particular option.
4871 */
4872 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
4873 * header option later in
4874 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4875 * Arg1: bool want_cookie. (in
4876 * writing SYNACK only)
4877 *
4878 * sock_ops->skb_data:
4879 * Not available because no header has
4880 * been written yet.
4881 *
4882 * sock_ops->skb_tcp_flags:
4883 * The tcp_flags of the
4884 * outgoing skb. (e.g. SYN, ACK, FIN).
4885 *
4886 * bpf_reserve_hdr_opt() should
4887 * be used to reserve space.
4888 */
4889 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
4890 * Arg1: bool want_cookie. (in
4891 * writing SYNACK only)
4892 *
4893 * sock_ops->skb_data:
4894 * Referring to the outgoing skb.
4895 * It covers the TCP header
4896 * that has already been written
4897 * by the kernel and the
4898 * earlier bpf-progs.
4899 *
4900 * sock_ops->skb_tcp_flags:
4901 * The tcp_flags of the outgoing
4902 * skb. (e.g. SYN, ACK, FIN).
4903 *
4904 * bpf_store_hdr_opt() should
4905 * be used to write the
4906 * option.
4907 *
4908 * bpf_load_hdr_opt() can also
4909 * be used to search for a
4910 * particular option that
4911 * has already been written
4912 * by the kernel or the
4913 * earlier bpf-progs.
4914 */
4915 };
4916
4917 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
4918 * changes between the TCP and BPF versions. Ideally this should never happen.
4919 * If it does, we need to add code to convert them before calling
4920 * the BPF sock_ops function.
4921 */
4922 enum {
4923 BPF_TCP_ESTABLISHED = 1,
4924 BPF_TCP_SYN_SENT,
4925 BPF_TCP_SYN_RECV,
4926 BPF_TCP_FIN_WAIT1,
4927 BPF_TCP_FIN_WAIT2,
4928 BPF_TCP_TIME_WAIT,
4929 BPF_TCP_CLOSE,
4930 BPF_TCP_CLOSE_WAIT,
4931 BPF_TCP_LAST_ACK,
4932 BPF_TCP_LISTEN,
4933 BPF_TCP_CLOSING, /* Now a valid state */
4934 BPF_TCP_NEW_SYN_RECV,
4935
4936 BPF_TCP_MAX_STATES /* Leave at the end! */
4937 };
4938
4939 enum {
4940 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4941 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
4942 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
4943 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
4944 /* Copy the SYN pkt to optval
4945 *
4946 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
4947 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
4948 * to only getting from the saved_syn. It can either get the
4949 * syn packet from:
4950 *
4951 * 1. the just-received SYN packet (only available when writing the
4952 * SYNACK). It will be useful when it is not necessary to
4953 * save the SYN packet for latter use. It is also the only way
4954 * to get the SYN during syncookie mode because the syn
4955 * packet cannot be saved during syncookie.
4956 *
4957 * OR
4958 *
4959 * 2. the earlier saved syn which was done by
4960 * bpf_setsockopt(TCP_SAVE_SYN).
4961 *
4962 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
4963 * SYN packet is obtained.
4964 *
4965 * If the bpf-prog does not need the IP[46] header, the
4966 * bpf-prog can avoid parsing the IP header by using
4967 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
4968 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
4969 *
4970 * >0: Total number of bytes copied
4971 * -ENOSPC: Not enough space in optval. Only optlen number of
4972 * bytes is copied.
4973 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
4974 * is not saved by setsockopt(TCP_SAVE_SYN).
4975 */
4976 TCP_BPF_SYN = 1005, /* Copy the TCP header */
4977 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
4978 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
4979 };
4980
4981 enum {
4982 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
4983 };
4984
4985 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
4986 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4987 */
4988 enum {
4989 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
4990 * total option spaces
4991 * required for an established
4992 * sk in order to calculate the
4993 * MSS. No skb is actually
4994 * sent.
4995 */
4996 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
4997 * when sending a SYN.
4998 */
4999 };
5000
5001 struct bpf_perf_event_value {
5002 __u64 counter;
5003 __u64 enabled;
5004 __u64 running;
5005 };
5006
5007 enum {
5008 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
5009 BPF_DEVCG_ACC_READ = (1ULL << 1),
5010 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
5011 };
5012
5013 enum {
5014 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
5015 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
5016 };
5017
5018 struct bpf_cgroup_dev_ctx {
5019 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
5020 __u32 access_type;
5021 __u32 major;
5022 __u32 minor;
5023 };
5024
5025 struct bpf_raw_tracepoint_args {
5026 __u64 args[0];
5027 };
5028
5029 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
5030 * OUTPUT: Do lookup from egress perspective; default is ingress
5031 */
5032 enum {
5033 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
5034 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
5035 };
5036
5037 enum {
5038 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
5039 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
5040 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
5041 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
5042 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
5043 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
5044 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
5045 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
5046 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
5047 };
5048
5049 struct bpf_fib_lookup {
5050 /* input: network family for lookup (AF_INET, AF_INET6)
5051 * output: network family of egress nexthop
5052 */
5053 __u8 family;
5054
5055 /* set if lookup is to consider L4 data - e.g., FIB rules */
5056 __u8 l4_protocol;
5057 __be16 sport;
5058 __be16 dport;
5059
5060 union { /* used for MTU check */
5061 /* input to lookup */
5062 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */
5063
5064 /* output: MTU value */
5065 __u16 mtu_result;
5066 };
5067 /* input: L3 device index for lookup
5068 * output: device index from FIB lookup
5069 */
5070 __u32 ifindex;
5071
5072 union {
5073 /* inputs to lookup */
5074 __u8 tos; /* AF_INET */
5075 __be32 flowinfo; /* AF_INET6, flow_label + priority */
5076
5077 /* output: metric of fib result (IPv4/IPv6 only) */
5078 __u32 rt_metric;
5079 };
5080
5081 union {
5082 __be32 ipv4_src;
5083 __u32 ipv6_src[4]; /* in6_addr; network order */
5084 };
5085
5086 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
5087 * network header. output: bpf_fib_lookup sets to gateway address
5088 * if FIB lookup returns gateway route
5089 */
5090 union {
5091 __be32 ipv4_dst;
5092 __u32 ipv6_dst[4]; /* in6_addr; network order */
5093 };
5094
5095 /* output */
5096 __be16 h_vlan_proto;
5097 __be16 h_vlan_TCI;
5098 __u8 smac[6]; /* ETH_ALEN */
5099 __u8 dmac[6]; /* ETH_ALEN */
5100 };
5101
5102 struct bpf_redir_neigh {
5103 /* network family for lookup (AF_INET, AF_INET6) */
5104 __u32 nh_family;
5105 /* network address of nexthop; skips fib lookup to find gateway */
5106 union {
5107 __be32 ipv4_nh;
5108 __u32 ipv6_nh[4]; /* in6_addr; network order */
5109 };
5110 };
5111
5112 /* bpf_check_mtu flags*/
5113 enum bpf_check_mtu_flags {
5114 BPF_MTU_CHK_SEGS = (1U << 0),
5115 };
5116
5117 enum bpf_check_mtu_ret {
5118 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */
5119 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */
5120 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */
5121 };
5122
5123 enum bpf_task_fd_type {
5124 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
5125 BPF_FD_TYPE_TRACEPOINT, /* tp name */
5126 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
5127 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
5128 BPF_FD_TYPE_UPROBE, /* filename + offset */
5129 BPF_FD_TYPE_URETPROBE, /* filename + offset */
5130 };
5131
5132 enum {
5133 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
5134 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
5135 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
5136 };
5137
5138 struct bpf_flow_keys {
5139 __u16 nhoff;
5140 __u16 thoff;
5141 __u16 addr_proto; /* ETH_P_* of valid addrs */
5142 __u8 is_frag;
5143 __u8 is_first_frag;
5144 __u8 is_encap;
5145 __u8 ip_proto;
5146 __be16 n_proto;
5147 __be16 sport;
5148 __be16 dport;
5149 union {
5150 struct {
5151 __be32 ipv4_src;
5152 __be32 ipv4_dst;
5153 };
5154 struct {
5155 __u32 ipv6_src[4]; /* in6_addr; network order */
5156 __u32 ipv6_dst[4]; /* in6_addr; network order */
5157 };
5158 };
5159 __u32 flags;
5160 __be32 flow_label;
5161 };
5162
5163 struct bpf_func_info {
5164 __u32 insn_off;
5165 __u32 type_id;
5166 };
5167
5168 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
5169 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
5170
5171 struct bpf_line_info {
5172 __u32 insn_off;
5173 __u32 file_name_off;
5174 __u32 line_off;
5175 __u32 line_col;
5176 };
5177
5178 struct bpf_spin_lock {
5179 __u32 val;
5180 };
5181
5182 struct bpf_sysctl {
5183 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
5184 * Allows 1,2,4-byte read, but no write.
5185 */
5186 __u32 file_pos; /* Sysctl file position to read from, write to.
5187 * Allows 1,2,4-byte read an 4-byte write.
5188 */
5189 };
5190
5191 struct bpf_sockopt {
5192 __bpf_md_ptr(struct bpf_sock *, sk);
5193 __bpf_md_ptr(void *, optval);
5194 __bpf_md_ptr(void *, optval_end);
5195
5196 __s32 level;
5197 __s32 optname;
5198 __s32 optlen;
5199 __s32 retval;
5200 };
5201
5202 struct bpf_pidns_info {
5203 __u32 pid;
5204 __u32 tgid;
5205 };
5206
5207 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
5208 struct bpf_sk_lookup {
5209 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
5210
5211 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
5212 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
5213 __u32 remote_ip4; /* Network byte order */
5214 __u32 remote_ip6[4]; /* Network byte order */
5215 __u32 remote_port; /* Network byte order */
5216 __u32 local_ip4; /* Network byte order */
5217 __u32 local_ip6[4]; /* Network byte order */
5218 __u32 local_port; /* Host byte order */
5219 };
5220
5221 /*
5222 * struct btf_ptr is used for typed pointer representation; the
5223 * type id is used to render the pointer data as the appropriate type
5224 * via the bpf_snprintf_btf() helper described above. A flags field -
5225 * potentially to specify additional details about the BTF pointer
5226 * (rather than its mode of display) - is included for future use.
5227 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
5228 */
5229 struct btf_ptr {
5230 void *ptr;
5231 __u32 type_id;
5232 __u32 flags; /* BTF ptr flags; unused at present. */
5233 };
5234
5235 /*
5236 * Flags to control bpf_snprintf_btf() behaviour.
5237 * - BTF_F_COMPACT: no formatting around type information
5238 * - BTF_F_NONAME: no struct/union member names/types
5239 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
5240 * equivalent to %px.
5241 * - BTF_F_ZERO: show zero-valued struct/union members; they
5242 * are not displayed by default
5243 */
5244 enum {
5245 BTF_F_COMPACT = (1ULL << 0),
5246 BTF_F_NONAME = (1ULL << 1),
5247 BTF_F_PTR_RAW = (1ULL << 2),
5248 BTF_F_ZERO = (1ULL << 3),
5249 };
5250
5251 #endif /* __LINUX_BPF_H__ */