]>
Commit | Line | Data |
---|---|---|
e2be04c7 | 1 | /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ |
daedfb22 AS |
2 | /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of version 2 of the GNU General Public | |
6 | * License as published by the Free Software Foundation. | |
7 | */ | |
8 | #ifndef _UAPI__LINUX_BPF_H__ | |
9 | #define _UAPI__LINUX_BPF_H__ | |
10 | ||
11 | #include <linux/types.h> | |
c15952dc | 12 | #include <linux/bpf_common.h> |
daedfb22 AS |
13 | |
14 | /* Extended instruction set based on top of classic BPF */ | |
15 | ||
16 | /* instruction classes */ | |
17 | #define BPF_ALU64 0x07 /* alu mode in double word width */ | |
18 | ||
19 | /* ld/ldx fields */ | |
cb5f7334 | 20 | #define BPF_DW 0x18 /* double word (64-bit) */ |
daedfb22 AS |
21 | #define BPF_XADD 0xc0 /* exclusive add */ |
22 | ||
23 | /* alu/jmp fields */ | |
24 | #define BPF_MOV 0xb0 /* mov reg to reg */ | |
25 | #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ | |
26 | ||
27 | /* change endianness of a register */ | |
28 | #define BPF_END 0xd0 /* flags for endianness conversion: */ | |
29 | #define BPF_TO_LE 0x00 /* convert to little-endian */ | |
30 | #define BPF_TO_BE 0x08 /* convert to big-endian */ | |
31 | #define BPF_FROM_LE BPF_TO_LE | |
32 | #define BPF_FROM_BE BPF_TO_BE | |
33 | ||
92b31a9a | 34 | /* jmp encodings */ |
daedfb22 | 35 | #define BPF_JNE 0x50 /* jump != */ |
92b31a9a DB |
36 | #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ |
37 | #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ | |
daedfb22 AS |
38 | #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ |
39 | #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ | |
92b31a9a DB |
40 | #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ |
41 | #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ | |
daedfb22 AS |
42 | #define BPF_CALL 0x80 /* function call */ |
43 | #define BPF_EXIT 0x90 /* function return */ | |
44 | ||
45 | /* Register numbers */ | |
46 | enum { | |
47 | BPF_REG_0 = 0, | |
48 | BPF_REG_1, | |
49 | BPF_REG_2, | |
50 | BPF_REG_3, | |
51 | BPF_REG_4, | |
52 | BPF_REG_5, | |
53 | BPF_REG_6, | |
54 | BPF_REG_7, | |
55 | BPF_REG_8, | |
56 | BPF_REG_9, | |
57 | BPF_REG_10, | |
58 | __MAX_BPF_REG, | |
59 | }; | |
60 | ||
61 | /* BPF has 10 general purpose 64-bit registers and stack frame. */ | |
62 | #define MAX_BPF_REG __MAX_BPF_REG | |
63 | ||
64 | struct bpf_insn { | |
65 | __u8 code; /* opcode */ | |
66 | __u8 dst_reg:4; /* dest register */ | |
67 | __u8 src_reg:4; /* source register */ | |
68 | __s16 off; /* signed offset */ | |
69 | __s32 imm; /* signed immediate constant */ | |
70 | }; | |
71 | ||
b95a5c4d DM |
72 | /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ |
73 | struct bpf_lpm_trie_key { | |
74 | __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ | |
75 | __u8 data[0]; /* Arbitrary size */ | |
76 | }; | |
77 | ||
b2197755 | 78 | /* BPF syscall commands, see bpf(2) man-page for details. */ |
99c55f7d | 79 | enum bpf_cmd { |
99c55f7d | 80 | BPF_MAP_CREATE, |
db20fd2b | 81 | BPF_MAP_LOOKUP_ELEM, |
db20fd2b | 82 | BPF_MAP_UPDATE_ELEM, |
db20fd2b | 83 | BPF_MAP_DELETE_ELEM, |
db20fd2b | 84 | BPF_MAP_GET_NEXT_KEY, |
09756af4 | 85 | BPF_PROG_LOAD, |
b2197755 DB |
86 | BPF_OBJ_PIN, |
87 | BPF_OBJ_GET, | |
f4324551 DM |
88 | BPF_PROG_ATTACH, |
89 | BPF_PROG_DETACH, | |
1cf1cae9 | 90 | BPF_PROG_TEST_RUN, |
34ad5580 MKL |
91 | BPF_PROG_GET_NEXT_ID, |
92 | BPF_MAP_GET_NEXT_ID, | |
b16d9aa4 | 93 | BPF_PROG_GET_FD_BY_ID, |
bd5f5f4e | 94 | BPF_MAP_GET_FD_BY_ID, |
1e270976 | 95 | BPF_OBJ_GET_INFO_BY_FD, |
468e2f64 | 96 | BPF_PROG_QUERY, |
c4f6699d | 97 | BPF_RAW_TRACEPOINT_OPEN, |
f56a653c | 98 | BPF_BTF_LOAD, |
78958fca | 99 | BPF_BTF_GET_FD_BY_ID, |
99c55f7d AS |
100 | }; |
101 | ||
102 | enum bpf_map_type { | |
103 | BPF_MAP_TYPE_UNSPEC, | |
0f8e4bd8 | 104 | BPF_MAP_TYPE_HASH, |
28fbcfa0 | 105 | BPF_MAP_TYPE_ARRAY, |
04fd61ab | 106 | BPF_MAP_TYPE_PROG_ARRAY, |
ea317b26 | 107 | BPF_MAP_TYPE_PERF_EVENT_ARRAY, |
824bd0ce | 108 | BPF_MAP_TYPE_PERCPU_HASH, |
a10423b8 | 109 | BPF_MAP_TYPE_PERCPU_ARRAY, |
d5a3b1f6 | 110 | BPF_MAP_TYPE_STACK_TRACE, |
4ed8ec52 | 111 | BPF_MAP_TYPE_CGROUP_ARRAY, |
29ba732a | 112 | BPF_MAP_TYPE_LRU_HASH, |
8f844938 | 113 | BPF_MAP_TYPE_LRU_PERCPU_HASH, |
b95a5c4d | 114 | BPF_MAP_TYPE_LPM_TRIE, |
56f668df | 115 | BPF_MAP_TYPE_ARRAY_OF_MAPS, |
bcc6b1b7 | 116 | BPF_MAP_TYPE_HASH_OF_MAPS, |
546ac1ff | 117 | BPF_MAP_TYPE_DEVMAP, |
174a79ff | 118 | BPF_MAP_TYPE_SOCKMAP, |
6710e112 | 119 | BPF_MAP_TYPE_CPUMAP, |
fbfc504a | 120 | BPF_MAP_TYPE_XSKMAP, |
81110384 | 121 | BPF_MAP_TYPE_SOCKHASH, |
99c55f7d AS |
122 | }; |
123 | ||
09756af4 AS |
124 | enum bpf_prog_type { |
125 | BPF_PROG_TYPE_UNSPEC, | |
ddd872bc | 126 | BPF_PROG_TYPE_SOCKET_FILTER, |
2541517c | 127 | BPF_PROG_TYPE_KPROBE, |
96be4325 | 128 | BPF_PROG_TYPE_SCHED_CLS, |
94caee8c | 129 | BPF_PROG_TYPE_SCHED_ACT, |
98b5c2c6 | 130 | BPF_PROG_TYPE_TRACEPOINT, |
6a773a15 | 131 | BPF_PROG_TYPE_XDP, |
0515e599 | 132 | BPF_PROG_TYPE_PERF_EVENT, |
0e33661d | 133 | BPF_PROG_TYPE_CGROUP_SKB, |
61023658 | 134 | BPF_PROG_TYPE_CGROUP_SOCK, |
3a0af8fd TG |
135 | BPF_PROG_TYPE_LWT_IN, |
136 | BPF_PROG_TYPE_LWT_OUT, | |
137 | BPF_PROG_TYPE_LWT_XMIT, | |
40304b2a | 138 | BPF_PROG_TYPE_SOCK_OPS, |
b005fd18 | 139 | BPF_PROG_TYPE_SK_SKB, |
ebc614f6 | 140 | BPF_PROG_TYPE_CGROUP_DEVICE, |
4f738adb | 141 | BPF_PROG_TYPE_SK_MSG, |
c4f6699d | 142 | BPF_PROG_TYPE_RAW_TRACEPOINT, |
4fbac77d | 143 | BPF_PROG_TYPE_CGROUP_SOCK_ADDR, |
09756af4 AS |
144 | }; |
145 | ||
0e33661d DM |
146 | enum bpf_attach_type { |
147 | BPF_CGROUP_INET_INGRESS, | |
148 | BPF_CGROUP_INET_EGRESS, | |
61023658 | 149 | BPF_CGROUP_INET_SOCK_CREATE, |
40304b2a | 150 | BPF_CGROUP_SOCK_OPS, |
464bc0fd JF |
151 | BPF_SK_SKB_STREAM_PARSER, |
152 | BPF_SK_SKB_STREAM_VERDICT, | |
ebc614f6 | 153 | BPF_CGROUP_DEVICE, |
4f738adb | 154 | BPF_SK_MSG_VERDICT, |
4fbac77d AI |
155 | BPF_CGROUP_INET4_BIND, |
156 | BPF_CGROUP_INET6_BIND, | |
d74bad4e AI |
157 | BPF_CGROUP_INET4_CONNECT, |
158 | BPF_CGROUP_INET6_CONNECT, | |
aac3fc32 AI |
159 | BPF_CGROUP_INET4_POST_BIND, |
160 | BPF_CGROUP_INET6_POST_BIND, | |
0e33661d DM |
161 | __MAX_BPF_ATTACH_TYPE |
162 | }; | |
163 | ||
164 | #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE | |
165 | ||
324bda9e AS |
166 | /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command |
167 | * | |
168 | * NONE(default): No further bpf programs allowed in the subtree. | |
169 | * | |
170 | * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, | |
171 | * the program in this cgroup yields to sub-cgroup program. | |
172 | * | |
173 | * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, | |
174 | * that cgroup program gets run in addition to the program in this cgroup. | |
175 | * | |
176 | * Only one program is allowed to be attached to a cgroup with | |
177 | * NONE or BPF_F_ALLOW_OVERRIDE flag. | |
178 | * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will | |
179 | * release old program and attach the new one. Attach flags has to match. | |
180 | * | |
181 | * Multiple programs are allowed to be attached to a cgroup with | |
182 | * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order | |
183 | * (those that were attached first, run first) | |
184 | * The programs of sub-cgroup are executed first, then programs of | |
185 | * this cgroup and then programs of parent cgroup. | |
186 | * When children program makes decision (like picking TCP CA or sock bind) | |
187 | * parent program has a chance to override it. | |
188 | * | |
189 | * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. | |
190 | * A cgroup with NONE doesn't allow any programs in sub-cgroups. | |
191 | * Ex1: | |
192 | * cgrp1 (MULTI progs A, B) -> | |
193 | * cgrp2 (OVERRIDE prog C) -> | |
194 | * cgrp3 (MULTI prog D) -> | |
195 | * cgrp4 (OVERRIDE prog E) -> | |
196 | * cgrp5 (NONE prog F) | |
197 | * the event in cgrp5 triggers execution of F,D,A,B in that order. | |
198 | * if prog F is detached, the execution is E,D,A,B | |
199 | * if prog F and D are detached, the execution is E,A,B | |
200 | * if prog F, E and D are detached, the execution is C,A,B | |
201 | * | |
202 | * All eligible programs are executed regardless of return code from | |
203 | * earlier programs. | |
7f677633 AS |
204 | */ |
205 | #define BPF_F_ALLOW_OVERRIDE (1U << 0) | |
324bda9e | 206 | #define BPF_F_ALLOW_MULTI (1U << 1) |
7f677633 | 207 | |
e07b98d9 DM |
208 | /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the |
209 | * verifier will perform strict alignment checking as if the kernel | |
210 | * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, | |
211 | * and NET_IP_ALIGN defined to 2. | |
212 | */ | |
213 | #define BPF_F_STRICT_ALIGNMENT (1U << 0) | |
214 | ||
cc8b0b92 | 215 | /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */ |
f1a66f85 DB |
216 | #define BPF_PSEUDO_MAP_FD 1 |
217 | ||
cc8b0b92 AS |
218 | /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative |
219 | * offset to another bpf function | |
220 | */ | |
221 | #define BPF_PSEUDO_CALL 1 | |
222 | ||
3274f520 AS |
223 | /* flags for BPF_MAP_UPDATE_ELEM command */ |
224 | #define BPF_ANY 0 /* create new element or update existing */ | |
225 | #define BPF_NOEXIST 1 /* create new element if it didn't exist */ | |
226 | #define BPF_EXIST 2 /* update existing element */ | |
227 | ||
96eabe7a | 228 | /* flags for BPF_MAP_CREATE command */ |
6c905981 | 229 | #define BPF_F_NO_PREALLOC (1U << 0) |
29ba732a | 230 | /* Instead of having one common LRU list in the |
8f844938 | 231 | * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list |
29ba732a MKL |
232 | * which can scale and perform better. |
233 | * Note, the LRU nodes (including free nodes) cannot be moved | |
234 | * across different LRU lists. | |
235 | */ | |
236 | #define BPF_F_NO_COMMON_LRU (1U << 1) | |
96eabe7a MKL |
237 | /* Specify numa node during map creation */ |
238 | #define BPF_F_NUMA_NODE (1U << 2) | |
6c905981 | 239 | |
468e2f64 AS |
240 | /* flags for BPF_PROG_QUERY */ |
241 | #define BPF_F_QUERY_EFFECTIVE (1U << 0) | |
242 | ||
cb4d2b3f MKL |
243 | #define BPF_OBJ_NAME_LEN 16U |
244 | ||
6e71b04a CF |
245 | /* Flags for accessing BPF object */ |
246 | #define BPF_F_RDONLY (1U << 3) | |
247 | #define BPF_F_WRONLY (1U << 4) | |
248 | ||
615755a7 SL |
249 | /* Flag for stack_map, store build_id+offset instead of pointer */ |
250 | #define BPF_F_STACK_BUILD_ID (1U << 5) | |
251 | ||
252 | enum bpf_stack_build_id_status { | |
253 | /* user space need an empty entry to identify end of a trace */ | |
254 | BPF_STACK_BUILD_ID_EMPTY = 0, | |
255 | /* with valid build_id and offset */ | |
256 | BPF_STACK_BUILD_ID_VALID = 1, | |
257 | /* couldn't get build_id, fallback to ip */ | |
258 | BPF_STACK_BUILD_ID_IP = 2, | |
259 | }; | |
260 | ||
261 | #define BPF_BUILD_ID_SIZE 20 | |
262 | struct bpf_stack_build_id { | |
263 | __s32 status; | |
264 | unsigned char build_id[BPF_BUILD_ID_SIZE]; | |
265 | union { | |
266 | __u64 offset; | |
267 | __u64 ip; | |
268 | }; | |
269 | }; | |
270 | ||
99c55f7d AS |
271 | union bpf_attr { |
272 | struct { /* anonymous struct used by BPF_MAP_CREATE command */ | |
273 | __u32 map_type; /* one of enum bpf_map_type */ | |
274 | __u32 key_size; /* size of key in bytes */ | |
275 | __u32 value_size; /* size of value in bytes */ | |
276 | __u32 max_entries; /* max number of entries in a map */ | |
96eabe7a MKL |
277 | __u32 map_flags; /* BPF_MAP_CREATE related |
278 | * flags defined above. | |
279 | */ | |
56f668df | 280 | __u32 inner_map_fd; /* fd pointing to the inner map */ |
96eabe7a MKL |
281 | __u32 numa_node; /* numa node (effective only if |
282 | * BPF_F_NUMA_NODE is set). | |
283 | */ | |
067cae47 | 284 | char map_name[BPF_OBJ_NAME_LEN]; |
a3884572 | 285 | __u32 map_ifindex; /* ifindex of netdev to create on */ |
a26ca7c9 MKL |
286 | __u32 btf_fd; /* fd pointing to a BTF type data */ |
287 | __u32 btf_key_id; /* BTF type_id of the key */ | |
288 | __u32 btf_value_id; /* BTF type_id of the value */ | |
99c55f7d | 289 | }; |
db20fd2b AS |
290 | |
291 | struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ | |
292 | __u32 map_fd; | |
293 | __aligned_u64 key; | |
294 | union { | |
295 | __aligned_u64 value; | |
296 | __aligned_u64 next_key; | |
297 | }; | |
3274f520 | 298 | __u64 flags; |
db20fd2b | 299 | }; |
09756af4 AS |
300 | |
301 | struct { /* anonymous struct used by BPF_PROG_LOAD command */ | |
302 | __u32 prog_type; /* one of enum bpf_prog_type */ | |
303 | __u32 insn_cnt; | |
304 | __aligned_u64 insns; | |
305 | __aligned_u64 license; | |
cbd35700 AS |
306 | __u32 log_level; /* verbosity level of verifier */ |
307 | __u32 log_size; /* size of user buffer */ | |
308 | __aligned_u64 log_buf; /* user supplied buffer */ | |
2541517c | 309 | __u32 kern_version; /* checked when prog_type=kprobe */ |
e07b98d9 | 310 | __u32 prog_flags; |
067cae47 | 311 | char prog_name[BPF_OBJ_NAME_LEN]; |
1f6f4cb7 | 312 | __u32 prog_ifindex; /* ifindex of netdev to prep for */ |
5e43f899 AI |
313 | /* For some prog types expected attach type must be known at |
314 | * load time to verify attach type specific parts of prog | |
315 | * (context accesses, allowed helpers, etc). | |
316 | */ | |
317 | __u32 expected_attach_type; | |
09756af4 | 318 | }; |
b2197755 DB |
319 | |
320 | struct { /* anonymous struct used by BPF_OBJ_* commands */ | |
321 | __aligned_u64 pathname; | |
322 | __u32 bpf_fd; | |
6e71b04a | 323 | __u32 file_flags; |
b2197755 | 324 | }; |
f4324551 DM |
325 | |
326 | struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ | |
327 | __u32 target_fd; /* container object to attach to */ | |
328 | __u32 attach_bpf_fd; /* eBPF program to attach */ | |
329 | __u32 attach_type; | |
7f677633 | 330 | __u32 attach_flags; |
f4324551 | 331 | }; |
1cf1cae9 AS |
332 | |
333 | struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ | |
334 | __u32 prog_fd; | |
335 | __u32 retval; | |
336 | __u32 data_size_in; | |
337 | __u32 data_size_out; | |
338 | __aligned_u64 data_in; | |
339 | __aligned_u64 data_out; | |
340 | __u32 repeat; | |
341 | __u32 duration; | |
342 | } test; | |
34ad5580 | 343 | |
b16d9aa4 MKL |
344 | struct { /* anonymous struct used by BPF_*_GET_*_ID */ |
345 | union { | |
346 | __u32 start_id; | |
347 | __u32 prog_id; | |
bd5f5f4e | 348 | __u32 map_id; |
78958fca | 349 | __u32 btf_id; |
b16d9aa4 | 350 | }; |
34ad5580 | 351 | __u32 next_id; |
6e71b04a | 352 | __u32 open_flags; |
34ad5580 | 353 | }; |
1e270976 MKL |
354 | |
355 | struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ | |
356 | __u32 bpf_fd; | |
357 | __u32 info_len; | |
358 | __aligned_u64 info; | |
359 | } info; | |
468e2f64 AS |
360 | |
361 | struct { /* anonymous struct used by BPF_PROG_QUERY command */ | |
362 | __u32 target_fd; /* container object to query */ | |
363 | __u32 attach_type; | |
364 | __u32 query_flags; | |
365 | __u32 attach_flags; | |
366 | __aligned_u64 prog_ids; | |
367 | __u32 prog_cnt; | |
368 | } query; | |
c4f6699d AS |
369 | |
370 | struct { | |
371 | __u64 name; | |
372 | __u32 prog_fd; | |
373 | } raw_tracepoint; | |
f56a653c MKL |
374 | |
375 | struct { /* anonymous struct for BPF_BTF_LOAD */ | |
376 | __aligned_u64 btf; | |
377 | __aligned_u64 btf_log_buf; | |
378 | __u32 btf_size; | |
379 | __u32 btf_log_size; | |
380 | __u32 btf_log_level; | |
381 | }; | |
99c55f7d AS |
382 | } __attribute__((aligned(8))); |
383 | ||
56a092c8 QM |
384 | /* The description below is an attempt at providing documentation to eBPF |
385 | * developers about the multiple available eBPF helper functions. It can be | |
386 | * parsed and used to produce a manual page. The workflow is the following, | |
387 | * and requires the rst2man utility: | |
388 | * | |
389 | * $ ./scripts/bpf_helpers_doc.py \ | |
390 | * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst | |
391 | * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 | |
392 | * $ man /tmp/bpf-helpers.7 | |
393 | * | |
394 | * Note that in order to produce this external documentation, some RST | |
395 | * formatting is used in the descriptions to get "bold" and "italics" in | |
396 | * manual pages. Also note that the few trailing white spaces are | |
397 | * intentional, removing them would break paragraphs for rst2man. | |
398 | * | |
399 | * Start of BPF helper function descriptions: | |
ad4a5223 QM |
400 | * |
401 | * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) | |
402 | * Description | |
403 | * Perform a lookup in *map* for an entry associated to *key*. | |
404 | * Return | |
405 | * Map value associated to *key*, or **NULL** if no entry was | |
406 | * found. | |
407 | * | |
408 | * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) | |
409 | * Description | |
410 | * Add or update the value of the entry associated to *key* in | |
411 | * *map* with *value*. *flags* is one of: | |
412 | * | |
413 | * **BPF_NOEXIST** | |
414 | * The entry for *key* must not exist in the map. | |
415 | * **BPF_EXIST** | |
416 | * The entry for *key* must already exist in the map. | |
417 | * **BPF_ANY** | |
418 | * No condition on the existence of the entry for *key*. | |
419 | * | |
420 | * Flag value **BPF_NOEXIST** cannot be used for maps of types | |
421 | * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all | |
422 | * elements always exist), the helper would return an error. | |
423 | * Return | |
424 | * 0 on success, or a negative error in case of failure. | |
425 | * | |
426 | * int bpf_map_delete_elem(struct bpf_map *map, const void *key) | |
427 | * Description | |
428 | * Delete entry with *key* from *map*. | |
429 | * Return | |
430 | * 0 on success, or a negative error in case of failure. | |
431 | * | |
432 | * int bpf_probe_read(void *dst, u32 size, const void *src) | |
433 | * Description | |
434 | * For tracing programs, safely attempt to read *size* bytes from | |
435 | * address *src* and store the data in *dst*. | |
436 | * Return | |
437 | * 0 on success, or a negative error in case of failure. | |
438 | * | |
439 | * u64 bpf_ktime_get_ns(void) | |
440 | * Description | |
441 | * Return the time elapsed since system boot, in nanoseconds. | |
442 | * Return | |
443 | * Current *ktime*. | |
444 | * | |
445 | * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) | |
446 | * Description | |
447 | * This helper is a "printk()-like" facility for debugging. It | |
448 | * prints a message defined by format *fmt* (of size *fmt_size*) | |
449 | * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if | |
450 | * available. It can take up to three additional **u64** | |
451 | * arguments (as an eBPF helpers, the total number of arguments is | |
452 | * limited to five). | |
453 | * | |
454 | * Each time the helper is called, it appends a line to the trace. | |
455 | * The format of the trace is customizable, and the exact output | |
456 | * one will get depends on the options set in | |
457 | * *\/sys/kernel/debug/tracing/trace_options* (see also the | |
458 | * *README* file under the same directory). However, it usually | |
459 | * defaults to something like: | |
460 | * | |
461 | * :: | |
462 | * | |
463 | * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> | |
464 | * | |
465 | * In the above: | |
466 | * | |
467 | * * ``telnet`` is the name of the current task. | |
468 | * * ``470`` is the PID of the current task. | |
469 | * * ``001`` is the CPU number on which the task is | |
470 | * running. | |
471 | * * In ``.N..``, each character refers to a set of | |
472 | * options (whether irqs are enabled, scheduling | |
473 | * options, whether hard/softirqs are running, level of | |
474 | * preempt_disabled respectively). **N** means that | |
475 | * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** | |
476 | * are set. | |
477 | * * ``419421.045894`` is a timestamp. | |
478 | * * ``0x00000001`` is a fake value used by BPF for the | |
479 | * instruction pointer register. | |
480 | * * ``<formatted msg>`` is the message formatted with | |
481 | * *fmt*. | |
482 | * | |
483 | * The conversion specifiers supported by *fmt* are similar, but | |
484 | * more limited than for printk(). They are **%d**, **%i**, | |
485 | * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, | |
486 | * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size | |
487 | * of field, padding with zeroes, etc.) is available, and the | |
488 | * helper will return **-EINVAL** (but print nothing) if it | |
489 | * encounters an unknown specifier. | |
490 | * | |
491 | * Also, note that **bpf_trace_printk**\ () is slow, and should | |
492 | * only be used for debugging purposes. For this reason, a notice | |
493 | * bloc (spanning several lines) is printed to kernel logs and | |
494 | * states that the helper should not be used "for production use" | |
495 | * the first time this helper is used (or more precisely, when | |
496 | * **trace_printk**\ () buffers are allocated). For passing values | |
497 | * to user space, perf events should be preferred. | |
498 | * Return | |
499 | * The number of bytes written to the buffer, or a negative error | |
500 | * in case of failure. | |
501 | * | |
1fdd08be QM |
502 | * u32 bpf_get_prandom_u32(void) |
503 | * Description | |
504 | * Get a pseudo-random number. | |
505 | * | |
506 | * From a security point of view, this helper uses its own | |
507 | * pseudo-random internal state, and cannot be used to infer the | |
508 | * seed of other random functions in the kernel. However, it is | |
509 | * essential to note that the generator used by the helper is not | |
510 | * cryptographically secure. | |
511 | * Return | |
512 | * A random 32-bit unsigned value. | |
513 | * | |
514 | * u32 bpf_get_smp_processor_id(void) | |
515 | * Description | |
516 | * Get the SMP (symmetric multiprocessing) processor id. Note that | |
517 | * all programs run with preemption disabled, which means that the | |
518 | * SMP processor id is stable during all the execution of the | |
519 | * program. | |
520 | * Return | |
521 | * The SMP id of the processor running the program. | |
522 | * | |
ad4a5223 QM |
523 | * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) |
524 | * Description | |
525 | * Store *len* bytes from address *from* into the packet | |
526 | * associated to *skb*, at *offset*. *flags* are a combination of | |
527 | * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the | |
528 | * checksum for the packet after storing the bytes) and | |
529 | * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ | |
530 | * **->swhash** and *skb*\ **->l4hash** to 0). | |
531 | * | |
532 | * A call to this helper is susceptible to change the underlaying | |
533 | * packet buffer. Therefore, at load time, all checks on pointers | |
534 | * previously done by the verifier are invalidated and must be | |
535 | * performed again, if the helper is used in combination with | |
536 | * direct packet access. | |
537 | * Return | |
538 | * 0 on success, or a negative error in case of failure. | |
539 | * | |
540 | * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) | |
541 | * Description | |
542 | * Recompute the layer 3 (e.g. IP) checksum for the packet | |
543 | * associated to *skb*. Computation is incremental, so the helper | |
544 | * must know the former value of the header field that was | |
545 | * modified (*from*), the new value of this field (*to*), and the | |
546 | * number of bytes (2 or 4) for this field, stored in *size*. | |
547 | * Alternatively, it is possible to store the difference between | |
548 | * the previous and the new values of the header field in *to*, by | |
549 | * setting *from* and *size* to 0. For both methods, *offset* | |
550 | * indicates the location of the IP checksum within the packet. | |
551 | * | |
552 | * This helper works in combination with **bpf_csum_diff**\ (), | |
553 | * which does not update the checksum in-place, but offers more | |
554 | * flexibility and can handle sizes larger than 2 or 4 for the | |
555 | * checksum to update. | |
556 | * | |
557 | * A call to this helper is susceptible to change the underlaying | |
558 | * packet buffer. Therefore, at load time, all checks on pointers | |
559 | * previously done by the verifier are invalidated and must be | |
560 | * performed again, if the helper is used in combination with | |
561 | * direct packet access. | |
562 | * Return | |
563 | * 0 on success, or a negative error in case of failure. | |
564 | * | |
565 | * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) | |
566 | * Description | |
567 | * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the | |
568 | * packet associated to *skb*. Computation is incremental, so the | |
569 | * helper must know the former value of the header field that was | |
570 | * modified (*from*), the new value of this field (*to*), and the | |
571 | * number of bytes (2 or 4) for this field, stored on the lowest | |
572 | * four bits of *flags*. Alternatively, it is possible to store | |
573 | * the difference between the previous and the new values of the | |
574 | * header field in *to*, by setting *from* and the four lowest | |
575 | * bits of *flags* to 0. For both methods, *offset* indicates the | |
576 | * location of the IP checksum within the packet. In addition to | |
577 | * the size of the field, *flags* can be added (bitwise OR) actual | |
578 | * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left | |
579 | * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and | |
580 | * for updates resulting in a null checksum the value is set to | |
581 | * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates | |
582 | * the checksum is to be computed against a pseudo-header. | |
583 | * | |
584 | * This helper works in combination with **bpf_csum_diff**\ (), | |
585 | * which does not update the checksum in-place, but offers more | |
586 | * flexibility and can handle sizes larger than 2 or 4 for the | |
587 | * checksum to update. | |
588 | * | |
589 | * A call to this helper is susceptible to change the underlaying | |
590 | * packet buffer. Therefore, at load time, all checks on pointers | |
591 | * previously done by the verifier are invalidated and must be | |
592 | * performed again, if the helper is used in combination with | |
593 | * direct packet access. | |
594 | * Return | |
595 | * 0 on success, or a negative error in case of failure. | |
596 | * | |
597 | * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) | |
598 | * Description | |
599 | * This special helper is used to trigger a "tail call", or in | |
600 | * other words, to jump into another eBPF program. The same stack | |
601 | * frame is used (but values on stack and in registers for the | |
602 | * caller are not accessible to the callee). This mechanism allows | |
603 | * for program chaining, either for raising the maximum number of | |
604 | * available eBPF instructions, or to execute given programs in | |
605 | * conditional blocks. For security reasons, there is an upper | |
606 | * limit to the number of successive tail calls that can be | |
607 | * performed. | |
608 | * | |
609 | * Upon call of this helper, the program attempts to jump into a | |
610 | * program referenced at index *index* in *prog_array_map*, a | |
611 | * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes | |
612 | * *ctx*, a pointer to the context. | |
613 | * | |
614 | * If the call succeeds, the kernel immediately runs the first | |
615 | * instruction of the new program. This is not a function call, | |
616 | * and it never returns to the previous program. If the call | |
617 | * fails, then the helper has no effect, and the caller continues | |
618 | * to run its subsequent instructions. A call can fail if the | |
619 | * destination program for the jump does not exist (i.e. *index* | |
620 | * is superior to the number of entries in *prog_array_map*), or | |
621 | * if the maximum number of tail calls has been reached for this | |
622 | * chain of programs. This limit is defined in the kernel by the | |
623 | * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), | |
624 | * which is currently set to 32. | |
625 | * Return | |
626 | * 0 on success, or a negative error in case of failure. | |
627 | * | |
628 | * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) | |
629 | * Description | |
630 | * Clone and redirect the packet associated to *skb* to another | |
631 | * net device of index *ifindex*. Both ingress and egress | |
632 | * interfaces can be used for redirection. The **BPF_F_INGRESS** | |
633 | * value in *flags* is used to make the distinction (ingress path | |
634 | * is selected if the flag is present, egress path otherwise). | |
635 | * This is the only flag supported for now. | |
636 | * | |
637 | * In comparison with **bpf_redirect**\ () helper, | |
638 | * **bpf_clone_redirect**\ () has the associated cost of | |
639 | * duplicating the packet buffer, but this can be executed out of | |
640 | * the eBPF program. Conversely, **bpf_redirect**\ () is more | |
641 | * efficient, but it is handled through an action code where the | |
642 | * redirection happens only after the eBPF program has returned. | |
643 | * | |
644 | * A call to this helper is susceptible to change the underlaying | |
645 | * packet buffer. Therefore, at load time, all checks on pointers | |
646 | * previously done by the verifier are invalidated and must be | |
647 | * performed again, if the helper is used in combination with | |
648 | * direct packet access. | |
649 | * Return | |
650 | * 0 on success, or a negative error in case of failure. | |
c456dec4 QM |
651 | * |
652 | * u64 bpf_get_current_pid_tgid(void) | |
653 | * Return | |
654 | * A 64-bit integer containing the current tgid and pid, and | |
655 | * created as such: | |
656 | * *current_task*\ **->tgid << 32 \|** | |
657 | * *current_task*\ **->pid**. | |
658 | * | |
659 | * u64 bpf_get_current_uid_gid(void) | |
660 | * Return | |
661 | * A 64-bit integer containing the current GID and UID, and | |
662 | * created as such: *current_gid* **<< 32 \|** *current_uid*. | |
663 | * | |
664 | * int bpf_get_current_comm(char *buf, u32 size_of_buf) | |
665 | * Description | |
666 | * Copy the **comm** attribute of the current task into *buf* of | |
667 | * *size_of_buf*. The **comm** attribute contains the name of | |
668 | * the executable (excluding the path) for the current task. The | |
669 | * *size_of_buf* must be strictly positive. On success, the | |
670 | * helper makes sure that the *buf* is NUL-terminated. On failure, | |
671 | * it is filled with zeroes. | |
672 | * Return | |
673 | * 0 on success, or a negative error in case of failure. | |
674 | * | |
1fdd08be QM |
675 | * u32 bpf_get_cgroup_classid(struct sk_buff *skb) |
676 | * Description | |
677 | * Retrieve the classid for the current task, i.e. for the net_cls | |
678 | * cgroup to which *skb* belongs. | |
679 | * | |
680 | * This helper can be used on TC egress path, but not on ingress. | |
681 | * | |
682 | * The net_cls cgroup provides an interface to tag network packets | |
683 | * based on a user-provided identifier for all traffic coming from | |
684 | * the tasks belonging to the related cgroup. See also the related | |
685 | * kernel documentation, available from the Linux sources in file | |
686 | * *Documentation/cgroup-v1/net_cls.txt*. | |
687 | * | |
688 | * The Linux kernel has two versions for cgroups: there are | |
689 | * cgroups v1 and cgroups v2. Both are available to users, who can | |
690 | * use a mixture of them, but note that the net_cls cgroup is for | |
691 | * cgroup v1 only. This makes it incompatible with BPF programs | |
692 | * run on cgroups, which is a cgroup-v2-only feature (a socket can | |
693 | * only hold data for one version of cgroups at a time). | |
694 | * | |
695 | * This helper is only available is the kernel was compiled with | |
696 | * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to | |
697 | * "**y**" or to "**m**". | |
698 | * Return | |
699 | * The classid, or 0 for the default unconfigured classid. | |
700 | * | |
c456dec4 QM |
701 | * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) |
702 | * Description | |
703 | * Push a *vlan_tci* (VLAN tag control information) of protocol | |
704 | * *vlan_proto* to the packet associated to *skb*, then update | |
705 | * the checksum. Note that if *vlan_proto* is different from | |
706 | * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to | |
707 | * be **ETH_P_8021Q**. | |
708 | * | |
709 | * A call to this helper is susceptible to change the underlaying | |
710 | * packet buffer. Therefore, at load time, all checks on pointers | |
711 | * previously done by the verifier are invalidated and must be | |
712 | * performed again, if the helper is used in combination with | |
713 | * direct packet access. | |
714 | * Return | |
715 | * 0 on success, or a negative error in case of failure. | |
716 | * | |
717 | * int bpf_skb_vlan_pop(struct sk_buff *skb) | |
718 | * Description | |
719 | * Pop a VLAN header from the packet associated to *skb*. | |
720 | * | |
721 | * A call to this helper is susceptible to change the underlaying | |
722 | * packet buffer. Therefore, at load time, all checks on pointers | |
723 | * previously done by the verifier are invalidated and must be | |
724 | * performed again, if the helper is used in combination with | |
725 | * direct packet access. | |
726 | * Return | |
727 | * 0 on success, or a negative error in case of failure. | |
728 | * | |
729 | * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) | |
730 | * Description | |
731 | * Get tunnel metadata. This helper takes a pointer *key* to an | |
732 | * empty **struct bpf_tunnel_key** of **size**, that will be | |
733 | * filled with tunnel metadata for the packet associated to *skb*. | |
734 | * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which | |
735 | * indicates that the tunnel is based on IPv6 protocol instead of | |
736 | * IPv4. | |
737 | * | |
738 | * The **struct bpf_tunnel_key** is an object that generalizes the | |
739 | * principal parameters used by various tunneling protocols into a | |
740 | * single struct. This way, it can be used to easily make a | |
741 | * decision based on the contents of the encapsulation header, | |
742 | * "summarized" in this struct. In particular, it holds the IP | |
743 | * address of the remote end (IPv4 or IPv6, depending on the case) | |
744 | * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, | |
745 | * this struct exposes the *key*\ **->tunnel_id**, which is | |
746 | * generally mapped to a VNI (Virtual Network Identifier), making | |
747 | * it programmable together with the **bpf_skb_set_tunnel_key**\ | |
748 | * () helper. | |
749 | * | |
750 | * Let's imagine that the following code is part of a program | |
751 | * attached to the TC ingress interface, on one end of a GRE | |
752 | * tunnel, and is supposed to filter out all messages coming from | |
753 | * remote ends with IPv4 address other than 10.0.0.1: | |
754 | * | |
755 | * :: | |
756 | * | |
757 | * int ret; | |
758 | * struct bpf_tunnel_key key = {}; | |
759 | * | |
760 | * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); | |
761 | * if (ret < 0) | |
762 | * return TC_ACT_SHOT; // drop packet | |
763 | * | |
764 | * if (key.remote_ipv4 != 0x0a000001) | |
765 | * return TC_ACT_SHOT; // drop packet | |
766 | * | |
767 | * return TC_ACT_OK; // accept packet | |
768 | * | |
769 | * This interface can also be used with all encapsulation devices | |
770 | * that can operate in "collect metadata" mode: instead of having | |
771 | * one network device per specific configuration, the "collect | |
772 | * metadata" mode only requires a single device where the | |
773 | * configuration can be extracted from this helper. | |
774 | * | |
775 | * This can be used together with various tunnels such as VXLan, | |
776 | * Geneve, GRE or IP in IP (IPIP). | |
777 | * Return | |
778 | * 0 on success, or a negative error in case of failure. | |
779 | * | |
780 | * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) | |
781 | * Description | |
782 | * Populate tunnel metadata for packet associated to *skb.* The | |
783 | * tunnel metadata is set to the contents of *key*, of *size*. The | |
784 | * *flags* can be set to a combination of the following values: | |
785 | * | |
786 | * **BPF_F_TUNINFO_IPV6** | |
787 | * Indicate that the tunnel is based on IPv6 protocol | |
788 | * instead of IPv4. | |
789 | * **BPF_F_ZERO_CSUM_TX** | |
790 | * For IPv4 packets, add a flag to tunnel metadata | |
791 | * indicating that checksum computation should be skipped | |
792 | * and checksum set to zeroes. | |
793 | * **BPF_F_DONT_FRAGMENT** | |
794 | * Add a flag to tunnel metadata indicating that the | |
795 | * packet should not be fragmented. | |
796 | * **BPF_F_SEQ_NUMBER** | |
797 | * Add a flag to tunnel metadata indicating that a | |
798 | * sequence number should be added to tunnel header before | |
799 | * sending the packet. This flag was added for GRE | |
800 | * encapsulation, but might be used with other protocols | |
801 | * as well in the future. | |
802 | * | |
803 | * Here is a typical usage on the transmit path: | |
804 | * | |
805 | * :: | |
806 | * | |
807 | * struct bpf_tunnel_key key; | |
808 | * populate key ... | |
809 | * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); | |
810 | * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); | |
811 | * | |
812 | * See also the description of the **bpf_skb_get_tunnel_key**\ () | |
813 | * helper for additional information. | |
814 | * Return | |
815 | * 0 on success, or a negative error in case of failure. | |
816 | * | |
c6b5fb86 QM |
817 | * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) |
818 | * Description | |
819 | * Read the value of a perf event counter. This helper relies on a | |
820 | * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of | |
821 | * the perf event counter is selected when *map* is updated with | |
822 | * perf event file descriptors. The *map* is an array whose size | |
823 | * is the number of available CPUs, and each cell contains a value | |
824 | * relative to one CPU. The value to retrieve is indicated by | |
825 | * *flags*, that contains the index of the CPU to look up, masked | |
826 | * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to | |
827 | * **BPF_F_CURRENT_CPU** to indicate that the value for the | |
828 | * current CPU should be retrieved. | |
829 | * | |
830 | * Note that before Linux 4.13, only hardware perf event can be | |
831 | * retrieved. | |
832 | * | |
833 | * Also, be aware that the newer helper | |
834 | * **bpf_perf_event_read_value**\ () is recommended over | |
3bd5a09b | 835 | * **bpf_perf_event_read**\ () in general. The latter has some ABI |
c6b5fb86 QM |
836 | * quirks where error and counter value are used as a return code |
837 | * (which is wrong to do since ranges may overlap). This issue is | |
3bd5a09b QM |
838 | * fixed with **bpf_perf_event_read_value**\ (), which at the same |
839 | * time provides more features over the **bpf_perf_event_read**\ | |
840 | * () interface. Please refer to the description of | |
c6b5fb86 QM |
841 | * **bpf_perf_event_read_value**\ () for details. |
842 | * Return | |
843 | * The value of the perf event counter read from the map, or a | |
844 | * negative error code in case of failure. | |
845 | * | |
c456dec4 QM |
846 | * int bpf_redirect(u32 ifindex, u64 flags) |
847 | * Description | |
848 | * Redirect the packet to another net device of index *ifindex*. | |
849 | * This helper is somewhat similar to **bpf_clone_redirect**\ | |
850 | * (), except that the packet is not cloned, which provides | |
851 | * increased performance. | |
852 | * | |
853 | * Except for XDP, both ingress and egress interfaces can be used | |
854 | * for redirection. The **BPF_F_INGRESS** value in *flags* is used | |
855 | * to make the distinction (ingress path is selected if the flag | |
856 | * is present, egress path otherwise). Currently, XDP only | |
857 | * supports redirection to the egress interface, and accepts no | |
858 | * flag at all. | |
859 | * | |
860 | * The same effect can be attained with the more generic | |
861 | * **bpf_redirect_map**\ (), which requires specific maps to be | |
862 | * used but offers better performance. | |
863 | * Return | |
864 | * For XDP, the helper returns **XDP_REDIRECT** on success or | |
865 | * **XDP_ABORTED** on error. For other program types, the values | |
866 | * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on | |
867 | * error. | |
868 | * | |
1fdd08be QM |
869 | * u32 bpf_get_route_realm(struct sk_buff *skb) |
870 | * Description | |
871 | * Retrieve the realm or the route, that is to say the | |
872 | * **tclassid** field of the destination for the *skb*. The | |
873 | * indentifier retrieved is a user-provided tag, similar to the | |
874 | * one used with the net_cls cgroup (see description for | |
875 | * **bpf_get_cgroup_classid**\ () helper), but here this tag is | |
876 | * held by a route (a destination entry), not by a task. | |
877 | * | |
878 | * Retrieving this identifier works with the clsact TC egress hook | |
879 | * (see also **tc-bpf(8)**), or alternatively on conventional | |
880 | * classful egress qdiscs, but not on TC ingress path. In case of | |
881 | * clsact TC egress hook, this has the advantage that, internally, | |
882 | * the destination entry has not been dropped yet in the transmit | |
883 | * path. Therefore, the destination entry does not need to be | |
884 | * artificially held via **netif_keep_dst**\ () for a classful | |
885 | * qdisc until the *skb* is freed. | |
886 | * | |
887 | * This helper is available only if the kernel was compiled with | |
888 | * **CONFIG_IP_ROUTE_CLASSID** configuration option. | |
889 | * Return | |
890 | * The realm of the route for the packet associated to *skb*, or 0 | |
891 | * if none was found. | |
892 | * | |
c456dec4 QM |
893 | * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) |
894 | * Description | |
895 | * Write raw *data* blob into a special BPF perf event held by | |
896 | * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf | |
897 | * event must have the following attributes: **PERF_SAMPLE_RAW** | |
898 | * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and | |
899 | * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. | |
900 | * | |
901 | * The *flags* are used to indicate the index in *map* for which | |
902 | * the value must be put, masked with **BPF_F_INDEX_MASK**. | |
903 | * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** | |
904 | * to indicate that the index of the current CPU core should be | |
905 | * used. | |
906 | * | |
907 | * The value to write, of *size*, is passed through eBPF stack and | |
908 | * pointed by *data*. | |
909 | * | |
910 | * The context of the program *ctx* needs also be passed to the | |
911 | * helper. | |
912 | * | |
913 | * On user space, a program willing to read the values needs to | |
914 | * call **perf_event_open**\ () on the perf event (either for | |
915 | * one or for all CPUs) and to store the file descriptor into the | |
916 | * *map*. This must be done before the eBPF program can send data | |
917 | * into it. An example is available in file | |
918 | * *samples/bpf/trace_output_user.c* in the Linux kernel source | |
919 | * tree (the eBPF program counterpart is in | |
920 | * *samples/bpf/trace_output_kern.c*). | |
921 | * | |
922 | * **bpf_perf_event_output**\ () achieves better performance | |
923 | * than **bpf_trace_printk**\ () for sharing data with user | |
924 | * space, and is much better suitable for streaming data from eBPF | |
925 | * programs. | |
926 | * | |
927 | * Note that this helper is not restricted to tracing use cases | |
928 | * and can be used with programs attached to TC or XDP as well, | |
929 | * where it allows for passing data to user space listeners. Data | |
930 | * can be: | |
931 | * | |
932 | * * Only custom structs, | |
933 | * * Only the packet payload, or | |
934 | * * A combination of both. | |
935 | * Return | |
936 | * 0 on success, or a negative error in case of failure. | |
937 | * | |
1fdd08be QM |
938 | * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) |
939 | * Description | |
940 | * This helper was provided as an easy way to load data from a | |
941 | * packet. It can be used to load *len* bytes from *offset* from | |
942 | * the packet associated to *skb*, into the buffer pointed by | |
943 | * *to*. | |
944 | * | |
945 | * Since Linux 4.7, usage of this helper has mostly been replaced | |
946 | * by "direct packet access", enabling packet data to be | |
947 | * manipulated with *skb*\ **->data** and *skb*\ **->data_end** | |
948 | * pointing respectively to the first byte of packet data and to | |
949 | * the byte after the last byte of packet data. However, it | |
950 | * remains useful if one wishes to read large quantities of data | |
951 | * at once from a packet into the eBPF stack. | |
952 | * Return | |
953 | * 0 on success, or a negative error in case of failure. | |
954 | * | |
c456dec4 QM |
955 | * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags) |
956 | * Description | |
957 | * Walk a user or a kernel stack and return its id. To achieve | |
958 | * this, the helper needs *ctx*, which is a pointer to the context | |
959 | * on which the tracing program is executed, and a pointer to a | |
960 | * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. | |
961 | * | |
962 | * The last argument, *flags*, holds the number of stack frames to | |
963 | * skip (from 0 to 255), masked with | |
964 | * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set | |
965 | * a combination of the following flags: | |
966 | * | |
967 | * **BPF_F_USER_STACK** | |
968 | * Collect a user space stack instead of a kernel stack. | |
969 | * **BPF_F_FAST_STACK_CMP** | |
970 | * Compare stacks by hash only. | |
971 | * **BPF_F_REUSE_STACKID** | |
972 | * If two different stacks hash into the same *stackid*, | |
973 | * discard the old one. | |
974 | * | |
975 | * The stack id retrieved is a 32 bit long integer handle which | |
976 | * can be further combined with other data (including other stack | |
977 | * ids) and used as a key into maps. This can be useful for | |
978 | * generating a variety of graphs (such as flame graphs or off-cpu | |
979 | * graphs). | |
980 | * | |
981 | * For walking a stack, this helper is an improvement over | |
982 | * **bpf_probe_read**\ (), which can be used with unrolled loops | |
983 | * but is not efficient and consumes a lot of eBPF instructions. | |
984 | * Instead, **bpf_get_stackid**\ () can collect up to | |
985 | * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that | |
986 | * this limit can be controlled with the **sysctl** program, and | |
987 | * that it should be manually increased in order to profile long | |
988 | * user stacks (such as stacks for Java programs). To do so, use: | |
989 | * | |
990 | * :: | |
991 | * | |
992 | * # sysctl kernel.perf_event_max_stack=<new value> | |
993 | * | |
994 | * Return | |
995 | * The positive or null stack id on success, or a negative error | |
996 | * in case of failure. | |
997 | * | |
1fdd08be QM |
998 | * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) |
999 | * Description | |
1000 | * Compute a checksum difference, from the raw buffer pointed by | |
1001 | * *from*, of length *from_size* (that must be a multiple of 4), | |
1002 | * towards the raw buffer pointed by *to*, of size *to_size* | |
1003 | * (same remark). An optional *seed* can be added to the value | |
1004 | * (this can be cascaded, the seed may come from a previous call | |
1005 | * to the helper). | |
1006 | * | |
1007 | * This is flexible enough to be used in several ways: | |
1008 | * | |
1009 | * * With *from_size* == 0, *to_size* > 0 and *seed* set to | |
1010 | * checksum, it can be used when pushing new data. | |
1011 | * * With *from_size* > 0, *to_size* == 0 and *seed* set to | |
1012 | * checksum, it can be used when removing data from a packet. | |
1013 | * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it | |
1014 | * can be used to compute a diff. Note that *from_size* and | |
1015 | * *to_size* do not need to be equal. | |
1016 | * | |
1017 | * This helper can be used in combination with | |
1018 | * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to | |
1019 | * which one can feed in the difference computed with | |
1020 | * **bpf_csum_diff**\ (). | |
1021 | * Return | |
1022 | * The checksum result, or a negative error code in case of | |
1023 | * failure. | |
1024 | * | |
1025 | * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) | |
1026 | * Description | |
1027 | * Retrieve tunnel options metadata for the packet associated to | |
1028 | * *skb*, and store the raw tunnel option data to the buffer *opt* | |
1029 | * of *size*. | |
1030 | * | |
1031 | * This helper can be used with encapsulation devices that can | |
1032 | * operate in "collect metadata" mode (please refer to the related | |
1033 | * note in the description of **bpf_skb_get_tunnel_key**\ () for | |
1034 | * more details). A particular example where this can be used is | |
1035 | * in combination with the Geneve encapsulation protocol, where it | |
1036 | * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) | |
1037 | * and retrieving arbitrary TLVs (Type-Length-Value headers) from | |
1038 | * the eBPF program. This allows for full customization of these | |
1039 | * headers. | |
1040 | * Return | |
1041 | * The size of the option data retrieved. | |
1042 | * | |
1043 | * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) | |
1044 | * Description | |
1045 | * Set tunnel options metadata for the packet associated to *skb* | |
1046 | * to the option data contained in the raw buffer *opt* of *size*. | |
1047 | * | |
1048 | * See also the description of the **bpf_skb_get_tunnel_opt**\ () | |
1049 | * helper for additional information. | |
1050 | * Return | |
1051 | * 0 on success, or a negative error in case of failure. | |
1052 | * | |
1053 | * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) | |
1054 | * Description | |
1055 | * Change the protocol of the *skb* to *proto*. Currently | |
1056 | * supported are transition from IPv4 to IPv6, and from IPv6 to | |
1057 | * IPv4. The helper takes care of the groundwork for the | |
1058 | * transition, including resizing the socket buffer. The eBPF | |
1059 | * program is expected to fill the new headers, if any, via | |
1060 | * **skb_store_bytes**\ () and to recompute the checksums with | |
1061 | * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ | |
1062 | * (). The main case for this helper is to perform NAT64 | |
1063 | * operations out of an eBPF program. | |
1064 | * | |
1065 | * Internally, the GSO type is marked as dodgy so that headers are | |
1066 | * checked and segments are recalculated by the GSO/GRO engine. | |
1067 | * The size for GSO target is adapted as well. | |
1068 | * | |
1069 | * All values for *flags* are reserved for future usage, and must | |
1070 | * be left at zero. | |
1071 | * | |
1072 | * A call to this helper is susceptible to change the underlaying | |
1073 | * packet buffer. Therefore, at load time, all checks on pointers | |
1074 | * previously done by the verifier are invalidated and must be | |
1075 | * performed again, if the helper is used in combination with | |
1076 | * direct packet access. | |
1077 | * Return | |
1078 | * 0 on success, or a negative error in case of failure. | |
1079 | * | |
1080 | * int bpf_skb_change_type(struct sk_buff *skb, u32 type) | |
1081 | * Description | |
1082 | * Change the packet type for the packet associated to *skb*. This | |
1083 | * comes down to setting *skb*\ **->pkt_type** to *type*, except | |
1084 | * the eBPF program does not have a write access to *skb*\ | |
1085 | * **->pkt_type** beside this helper. Using a helper here allows | |
1086 | * for graceful handling of errors. | |
1087 | * | |
1088 | * The major use case is to change incoming *skb*s to | |
1089 | * **PACKET_HOST** in a programmatic way instead of having to | |
1090 | * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for | |
1091 | * example. | |
1092 | * | |
1093 | * Note that *type* only allows certain values. At this time, they | |
1094 | * are: | |
1095 | * | |
1096 | * **PACKET_HOST** | |
1097 | * Packet is for us. | |
1098 | * **PACKET_BROADCAST** | |
1099 | * Send packet to all. | |
1100 | * **PACKET_MULTICAST** | |
1101 | * Send packet to group. | |
1102 | * **PACKET_OTHERHOST** | |
1103 | * Send packet to someone else. | |
1104 | * Return | |
1105 | * 0 on success, or a negative error in case of failure. | |
1106 | * | |
c6b5fb86 QM |
1107 | * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) |
1108 | * Description | |
1109 | * Check whether *skb* is a descendant of the cgroup2 held by | |
1110 | * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. | |
1111 | * Return | |
1112 | * The return value depends on the result of the test, and can be: | |
1113 | * | |
1114 | * * 0, if the *skb* failed the cgroup2 descendant test. | |
1115 | * * 1, if the *skb* succeeded the cgroup2 descendant test. | |
1116 | * * A negative error code, if an error occurred. | |
1117 | * | |
fa15601a QM |
1118 | * u32 bpf_get_hash_recalc(struct sk_buff *skb) |
1119 | * Description | |
1120 | * Retrieve the hash of the packet, *skb*\ **->hash**. If it is | |
1121 | * not set, in particular if the hash was cleared due to mangling, | |
1122 | * recompute this hash. Later accesses to the hash can be done | |
1123 | * directly with *skb*\ **->hash**. | |
1124 | * | |
1125 | * Calling **bpf_set_hash_invalid**\ (), changing a packet | |
1126 | * prototype with **bpf_skb_change_proto**\ (), or calling | |
1127 | * **bpf_skb_store_bytes**\ () with the | |
1128 | * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear | |
1129 | * the hash and to trigger a new computation for the next call to | |
1130 | * **bpf_get_hash_recalc**\ (). | |
1131 | * Return | |
1132 | * The 32-bit hash. | |
1133 | * | |
c456dec4 QM |
1134 | * u64 bpf_get_current_task(void) |
1135 | * Return | |
1136 | * A pointer to the current task struct. | |
fa15601a | 1137 | * |
c6b5fb86 QM |
1138 | * int bpf_probe_write_user(void *dst, const void *src, u32 len) |
1139 | * Description | |
1140 | * Attempt in a safe way to write *len* bytes from the buffer | |
1141 | * *src* to *dst* in memory. It only works for threads that are in | |
1142 | * user context, and *dst* must be a valid user space address. | |
1143 | * | |
1144 | * This helper should not be used to implement any kind of | |
1145 | * security mechanism because of TOC-TOU attacks, but rather to | |
1146 | * debug, divert, and manipulate execution of semi-cooperative | |
1147 | * processes. | |
1148 | * | |
1149 | * Keep in mind that this feature is meant for experiments, and it | |
1150 | * has a risk of crashing the system and running programs. | |
1151 | * Therefore, when an eBPF program using this helper is attached, | |
1152 | * a warning including PID and process name is printed to kernel | |
1153 | * logs. | |
1154 | * Return | |
1155 | * 0 on success, or a negative error in case of failure. | |
1156 | * | |
1157 | * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) | |
1158 | * Description | |
1159 | * Check whether the probe is being run is the context of a given | |
1160 | * subset of the cgroup2 hierarchy. The cgroup2 to test is held by | |
1161 | * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. | |
1162 | * Return | |
1163 | * The return value depends on the result of the test, and can be: | |
1164 | * | |
1165 | * * 0, if the *skb* task belongs to the cgroup2. | |
1166 | * * 1, if the *skb* task does not belong to the cgroup2. | |
1167 | * * A negative error code, if an error occurred. | |
1168 | * | |
fa15601a QM |
1169 | * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) |
1170 | * Description | |
1171 | * Resize (trim or grow) the packet associated to *skb* to the | |
1172 | * new *len*. The *flags* are reserved for future usage, and must | |
1173 | * be left at zero. | |
1174 | * | |
1175 | * The basic idea is that the helper performs the needed work to | |
1176 | * change the size of the packet, then the eBPF program rewrites | |
1177 | * the rest via helpers like **bpf_skb_store_bytes**\ (), | |
1178 | * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () | |
1179 | * and others. This helper is a slow path utility intended for | |
1180 | * replies with control messages. And because it is targeted for | |
1181 | * slow path, the helper itself can afford to be slow: it | |
1182 | * implicitly linearizes, unclones and drops offloads from the | |
1183 | * *skb*. | |
1184 | * | |
1185 | * A call to this helper is susceptible to change the underlaying | |
1186 | * packet buffer. Therefore, at load time, all checks on pointers | |
1187 | * previously done by the verifier are invalidated and must be | |
1188 | * performed again, if the helper is used in combination with | |
1189 | * direct packet access. | |
1190 | * Return | |
1191 | * 0 on success, or a negative error in case of failure. | |
1192 | * | |
1193 | * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) | |
1194 | * Description | |
1195 | * Pull in non-linear data in case the *skb* is non-linear and not | |
1196 | * all of *len* are part of the linear section. Make *len* bytes | |
1197 | * from *skb* readable and writable. If a zero value is passed for | |
1198 | * *len*, then the whole length of the *skb* is pulled. | |
1199 | * | |
1200 | * This helper is only needed for reading and writing with direct | |
1201 | * packet access. | |
1202 | * | |
1203 | * For direct packet access, testing that offsets to access | |
1204 | * are within packet boundaries (test on *skb*\ **->data_end**) is | |
1205 | * susceptible to fail if offsets are invalid, or if the requested | |
1206 | * data is in non-linear parts of the *skb*. On failure the | |
1207 | * program can just bail out, or in the case of a non-linear | |
1208 | * buffer, use a helper to make the data available. The | |
1209 | * **bpf_skb_load_bytes**\ () helper is a first solution to access | |
1210 | * the data. Another one consists in using **bpf_skb_pull_data** | |
1211 | * to pull in once the non-linear parts, then retesting and | |
1212 | * eventually access the data. | |
1213 | * | |
1214 | * At the same time, this also makes sure the *skb* is uncloned, | |
1215 | * which is a necessary condition for direct write. As this needs | |
1216 | * to be an invariant for the write part only, the verifier | |
1217 | * detects writes and adds a prologue that is calling | |
1218 | * **bpf_skb_pull_data()** to effectively unclone the *skb* from | |
1219 | * the very beginning in case it is indeed cloned. | |
1220 | * | |
1221 | * A call to this helper is susceptible to change the underlaying | |
1222 | * packet buffer. Therefore, at load time, all checks on pointers | |
1223 | * previously done by the verifier are invalidated and must be | |
1224 | * performed again, if the helper is used in combination with | |
1225 | * direct packet access. | |
1226 | * Return | |
1227 | * 0 on success, or a negative error in case of failure. | |
1228 | * | |
1229 | * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) | |
1230 | * Description | |
1231 | * Add the checksum *csum* into *skb*\ **->csum** in case the | |
1232 | * driver has supplied a checksum for the entire packet into that | |
1233 | * field. Return an error otherwise. This helper is intended to be | |
1234 | * used in combination with **bpf_csum_diff**\ (), in particular | |
1235 | * when the checksum needs to be updated after data has been | |
1236 | * written into the packet through direct packet access. | |
1237 | * Return | |
1238 | * The checksum on success, or a negative error code in case of | |
1239 | * failure. | |
1240 | * | |
1241 | * void bpf_set_hash_invalid(struct sk_buff *skb) | |
1242 | * Description | |
1243 | * Invalidate the current *skb*\ **->hash**. It can be used after | |
1244 | * mangling on headers through direct packet access, in order to | |
1245 | * indicate that the hash is outdated and to trigger a | |
1246 | * recalculation the next time the kernel tries to access this | |
1247 | * hash or when the **bpf_get_hash_recalc**\ () helper is called. | |
1248 | * | |
1249 | * int bpf_get_numa_node_id(void) | |
1250 | * Description | |
1251 | * Return the id of the current NUMA node. The primary use case | |
1252 | * for this helper is the selection of sockets for the local NUMA | |
1253 | * node, when the program is attached to sockets using the | |
1254 | * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), | |
1255 | * but the helper is also available to other eBPF program types, | |
1256 | * similarly to **bpf_get_smp_processor_id**\ (). | |
1257 | * Return | |
1258 | * The id of current NUMA node. | |
1259 | * | |
c6b5fb86 QM |
1260 | * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) |
1261 | * Description | |
1262 | * Grows headroom of packet associated to *skb* and adjusts the | |
1263 | * offset of the MAC header accordingly, adding *len* bytes of | |
1264 | * space. It automatically extends and reallocates memory as | |
1265 | * required. | |
1266 | * | |
1267 | * This helper can be used on a layer 3 *skb* to push a MAC header | |
1268 | * for redirection into a layer 2 device. | |
1269 | * | |
1270 | * All values for *flags* are reserved for future usage, and must | |
1271 | * be left at zero. | |
1272 | * | |
1273 | * A call to this helper is susceptible to change the underlaying | |
1274 | * packet buffer. Therefore, at load time, all checks on pointers | |
1275 | * previously done by the verifier are invalidated and must be | |
1276 | * performed again, if the helper is used in combination with | |
1277 | * direct packet access. | |
1278 | * Return | |
1279 | * 0 on success, or a negative error in case of failure. | |
1280 | * | |
1281 | * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) | |
1282 | * Description | |
1283 | * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that | |
1284 | * it is possible to use a negative value for *delta*. This helper | |
1285 | * can be used to prepare the packet for pushing or popping | |
1286 | * headers. | |
1287 | * | |
1288 | * A call to this helper is susceptible to change the underlaying | |
1289 | * packet buffer. Therefore, at load time, all checks on pointers | |
1290 | * previously done by the verifier are invalidated and must be | |
1291 | * performed again, if the helper is used in combination with | |
1292 | * direct packet access. | |
1293 | * Return | |
1294 | * 0 on success, or a negative error in case of failure. | |
1295 | * | |
1296 | * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) | |
1297 | * Description | |
1298 | * Copy a NUL terminated string from an unsafe address | |
1299 | * *unsafe_ptr* to *dst*. The *size* should include the | |
1300 | * terminating NUL byte. In case the string length is smaller than | |
1301 | * *size*, the target is not padded with further NUL bytes. If the | |
1302 | * string length is larger than *size*, just *size*-1 bytes are | |
1303 | * copied and the last byte is set to NUL. | |
1304 | * | |
1305 | * On success, the length of the copied string is returned. This | |
1306 | * makes this helper useful in tracing programs for reading | |
1307 | * strings, and more importantly to get its length at runtime. See | |
1308 | * the following snippet: | |
1309 | * | |
1310 | * :: | |
1311 | * | |
1312 | * SEC("kprobe/sys_open") | |
1313 | * void bpf_sys_open(struct pt_regs *ctx) | |
1314 | * { | |
1315 | * char buf[PATHLEN]; // PATHLEN is defined to 256 | |
1316 | * int res = bpf_probe_read_str(buf, sizeof(buf), | |
1317 | * ctx->di); | |
1318 | * | |
1319 | * // Consume buf, for example push it to | |
1320 | * // userspace via bpf_perf_event_output(); we | |
1321 | * // can use res (the string length) as event | |
1322 | * // size, after checking its boundaries. | |
1323 | * } | |
1324 | * | |
1325 | * In comparison, using **bpf_probe_read()** helper here instead | |
1326 | * to read the string would require to estimate the length at | |
1327 | * compile time, and would often result in copying more memory | |
1328 | * than necessary. | |
1329 | * | |
1330 | * Another useful use case is when parsing individual process | |
1331 | * arguments or individual environment variables navigating | |
1332 | * *current*\ **->mm->arg_start** and *current*\ | |
1333 | * **->mm->env_start**: using this helper and the return value, | |
1334 | * one can quickly iterate at the right offset of the memory area. | |
1335 | * Return | |
1336 | * On success, the strictly positive length of the string, | |
1337 | * including the trailing NUL character. On error, a negative | |
1338 | * value. | |
1339 | * | |
1340 | * u64 bpf_get_socket_cookie(struct sk_buff *skb) | |
1341 | * Description | |
1342 | * If the **struct sk_buff** pointed by *skb* has a known socket, | |
1343 | * retrieve the cookie (generated by the kernel) of this socket. | |
1344 | * If no cookie has been set yet, generate a new cookie. Once | |
1345 | * generated, the socket cookie remains stable for the life of the | |
1346 | * socket. This helper can be useful for monitoring per socket | |
1347 | * networking traffic statistics as it provides a unique socket | |
1348 | * identifier per namespace. | |
1349 | * Return | |
1350 | * A 8-byte long non-decreasing number on success, or 0 if the | |
1351 | * socket field is missing inside *skb*. | |
1352 | * | |
1353 | * u32 bpf_get_socket_uid(struct sk_buff *skb) | |
1354 | * Return | |
1355 | * The owner UID of the socket associated to *skb*. If the socket | |
1356 | * is **NULL**, or if it is not a full socket (i.e. if it is a | |
1357 | * time-wait or a request socket instead), **overflowuid** value | |
1358 | * is returned (note that **overflowuid** might also be the actual | |
1359 | * UID value for the socket). | |
1360 | * | |
fa15601a QM |
1361 | * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) |
1362 | * Description | |
1363 | * Set the full hash for *skb* (set the field *skb*\ **->hash**) | |
1364 | * to value *hash*. | |
1365 | * Return | |
1366 | * 0 | |
1367 | * | |
a3ef8e9a | 1368 | * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) |
7aa79a86 QM |
1369 | * Description |
1370 | * Emulate a call to **setsockopt()** on the socket associated to | |
1371 | * *bpf_socket*, which must be a full socket. The *level* at | |
1372 | * which the option resides and the name *optname* of the option | |
1373 | * must be specified, see **setsockopt(2)** for more information. | |
1374 | * The option value of length *optlen* is pointed by *optval*. | |
1375 | * | |
1376 | * This helper actually implements a subset of **setsockopt()**. | |
1377 | * It supports the following *level*\ s: | |
1378 | * | |
1379 | * * **SOL_SOCKET**, which supports the following *optname*\ s: | |
1380 | * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, | |
1381 | * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. | |
1382 | * * **IPPROTO_TCP**, which supports the following *optname*\ s: | |
1383 | * **TCP_CONGESTION**, **TCP_BPF_IW**, | |
1384 | * **TCP_BPF_SNDCWND_CLAMP**. | |
1385 | * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. | |
1386 | * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. | |
1387 | * Return | |
1388 | * 0 on success, or a negative error in case of failure. | |
1389 | * | |
fa15601a QM |
1390 | * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags) |
1391 | * Description | |
1392 | * Grow or shrink the room for data in the packet associated to | |
1393 | * *skb* by *len_diff*, and according to the selected *mode*. | |
1394 | * | |
1395 | * There is a single supported mode at this time: | |
1396 | * | |
1397 | * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer | |
1398 | * (room space is added or removed below the layer 3 header). | |
1399 | * | |
1400 | * All values for *flags* are reserved for future usage, and must | |
1401 | * be left at zero. | |
1402 | * | |
1403 | * A call to this helper is susceptible to change the underlaying | |
1404 | * packet buffer. Therefore, at load time, all checks on pointers | |
1405 | * previously done by the verifier are invalidated and must be | |
1406 | * performed again, if the helper is used in combination with | |
1407 | * direct packet access. | |
1408 | * Return | |
1409 | * 0 on success, or a negative error in case of failure. | |
1410 | * | |
ab127040 QM |
1411 | * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) |
1412 | * Description | |
1413 | * Redirect the packet to the endpoint referenced by *map* at | |
1414 | * index *key*. Depending on its type, this *map* can contain | |
1415 | * references to net devices (for forwarding packets through other | |
1416 | * ports), or to CPUs (for redirecting XDP frames to another CPU; | |
1417 | * but this is only implemented for native XDP (with driver | |
1418 | * support) as of this writing). | |
1419 | * | |
1420 | * All values for *flags* are reserved for future usage, and must | |
1421 | * be left at zero. | |
1422 | * | |
1423 | * When used to redirect packets to net devices, this helper | |
1424 | * provides a high performance increase over **bpf_redirect**\ (). | |
1425 | * This is due to various implementation details of the underlying | |
1426 | * mechanisms, one of which is the fact that **bpf_redirect_map**\ | |
1427 | * () tries to send packet as a "bulk" to the device. | |
1428 | * Return | |
1429 | * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. | |
1430 | * | |
1431 | * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) | |
1432 | * Description | |
1433 | * Redirect the packet to the socket referenced by *map* (of type | |
1434 | * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and | |
1435 | * egress interfaces can be used for redirection. The | |
1436 | * **BPF_F_INGRESS** value in *flags* is used to make the | |
1437 | * distinction (ingress path is selected if the flag is present, | |
1438 | * egress path otherwise). This is the only flag supported for now. | |
1439 | * Return | |
1440 | * **SK_PASS** on success, or **SK_DROP** on error. | |
1441 | * | |
a3ef8e9a | 1442 | * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) |
ab127040 QM |
1443 | * Description |
1444 | * Add an entry to, or update a *map* referencing sockets. The | |
1445 | * *skops* is used as a new value for the entry associated to | |
1446 | * *key*. *flags* is one of: | |
1447 | * | |
1448 | * **BPF_NOEXIST** | |
1449 | * The entry for *key* must not exist in the map. | |
1450 | * **BPF_EXIST** | |
1451 | * The entry for *key* must already exist in the map. | |
1452 | * **BPF_ANY** | |
1453 | * No condition on the existence of the entry for *key*. | |
1454 | * | |
1455 | * If the *map* has eBPF programs (parser and verdict), those will | |
1456 | * be inherited by the socket being added. If the socket is | |
1457 | * already attached to eBPF programs, this results in an error. | |
1458 | * Return | |
1459 | * 0 on success, or a negative error in case of failure. | |
1460 | * | |
fa15601a QM |
1461 | * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) |
1462 | * Description | |
1463 | * Adjust the address pointed by *xdp_md*\ **->data_meta** by | |
1464 | * *delta* (which can be positive or negative). Note that this | |
1465 | * operation modifies the address stored in *xdp_md*\ **->data**, | |
1466 | * so the latter must be loaded only after the helper has been | |
1467 | * called. | |
1468 | * | |
1469 | * The use of *xdp_md*\ **->data_meta** is optional and programs | |
1470 | * are not required to use it. The rationale is that when the | |
1471 | * packet is processed with XDP (e.g. as DoS filter), it is | |
1472 | * possible to push further meta data along with it before passing | |
1473 | * to the stack, and to give the guarantee that an ingress eBPF | |
1474 | * program attached as a TC classifier on the same device can pick | |
1475 | * this up for further post-processing. Since TC works with socket | |
1476 | * buffers, it remains possible to set from XDP the **mark** or | |
1477 | * **priority** pointers, or other pointers for the socket buffer. | |
1478 | * Having this scratch space generic and programmable allows for | |
1479 | * more flexibility as the user is free to store whatever meta | |
1480 | * data they need. | |
1481 | * | |
1482 | * A call to this helper is susceptible to change the underlaying | |
1483 | * packet buffer. Therefore, at load time, all checks on pointers | |
1484 | * previously done by the verifier are invalidated and must be | |
1485 | * performed again, if the helper is used in combination with | |
1486 | * direct packet access. | |
1487 | * Return | |
1488 | * 0 on success, or a negative error in case of failure. | |
7aa79a86 QM |
1489 | * |
1490 | * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) | |
1491 | * Description | |
1492 | * Read the value of a perf event counter, and store it into *buf* | |
1493 | * of size *buf_size*. This helper relies on a *map* of type | |
1494 | * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event | |
1495 | * counter is selected when *map* is updated with perf event file | |
1496 | * descriptors. The *map* is an array whose size is the number of | |
1497 | * available CPUs, and each cell contains a value relative to one | |
1498 | * CPU. The value to retrieve is indicated by *flags*, that | |
1499 | * contains the index of the CPU to look up, masked with | |
1500 | * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to | |
1501 | * **BPF_F_CURRENT_CPU** to indicate that the value for the | |
1502 | * current CPU should be retrieved. | |
1503 | * | |
1504 | * This helper behaves in a way close to | |
1505 | * **bpf_perf_event_read**\ () helper, save that instead of | |
1506 | * just returning the value observed, it fills the *buf* | |
1507 | * structure. This allows for additional data to be retrieved: in | |
1508 | * particular, the enabled and running times (in *buf*\ | |
1509 | * **->enabled** and *buf*\ **->running**, respectively) are | |
1510 | * copied. In general, **bpf_perf_event_read_value**\ () is | |
1511 | * recommended over **bpf_perf_event_read**\ (), which has some | |
1512 | * ABI issues and provides fewer functionalities. | |
1513 | * | |
1514 | * These values are interesting, because hardware PMU (Performance | |
1515 | * Monitoring Unit) counters are limited resources. When there are | |
1516 | * more PMU based perf events opened than available counters, | |
1517 | * kernel will multiplex these events so each event gets certain | |
1518 | * percentage (but not all) of the PMU time. In case that | |
1519 | * multiplexing happens, the number of samples or counter value | |
1520 | * will not reflect the case compared to when no multiplexing | |
1521 | * occurs. This makes comparison between different runs difficult. | |
1522 | * Typically, the counter value should be normalized before | |
1523 | * comparing to other experiments. The usual normalization is done | |
1524 | * as follows. | |
1525 | * | |
1526 | * :: | |
1527 | * | |
1528 | * normalized_counter = counter * t_enabled / t_running | |
1529 | * | |
1530 | * Where t_enabled is the time enabled for event and t_running is | |
1531 | * the time running for event since last normalization. The | |
1532 | * enabled and running times are accumulated since the perf event | |
1533 | * open. To achieve scaling factor between two invocations of an | |
1534 | * eBPF program, users can can use CPU id as the key (which is | |
1535 | * typical for perf array usage model) to remember the previous | |
1536 | * value and do the calculation inside the eBPF program. | |
1537 | * Return | |
1538 | * 0 on success, or a negative error in case of failure. | |
1539 | * | |
a3ef8e9a | 1540 | * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) |
7aa79a86 QM |
1541 | * Description |
1542 | * For en eBPF program attached to a perf event, retrieve the | |
1543 | * value of the event counter associated to *ctx* and store it in | |
1544 | * the structure pointed by *buf* and of size *buf_size*. Enabled | |
1545 | * and running times are also stored in the structure (see | |
1546 | * description of helper **bpf_perf_event_read_value**\ () for | |
1547 | * more details). | |
1548 | * Return | |
1549 | * 0 on success, or a negative error in case of failure. | |
1550 | * | |
a3ef8e9a | 1551 | * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) |
7aa79a86 QM |
1552 | * Description |
1553 | * Emulate a call to **getsockopt()** on the socket associated to | |
1554 | * *bpf_socket*, which must be a full socket. The *level* at | |
1555 | * which the option resides and the name *optname* of the option | |
1556 | * must be specified, see **getsockopt(2)** for more information. | |
1557 | * The retrieved value is stored in the structure pointed by | |
1558 | * *opval* and of length *optlen*. | |
1559 | * | |
1560 | * This helper actually implements a subset of **getsockopt()**. | |
1561 | * It supports the following *level*\ s: | |
1562 | * | |
1563 | * * **IPPROTO_TCP**, which supports *optname* | |
1564 | * **TCP_CONGESTION**. | |
1565 | * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. | |
1566 | * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. | |
1567 | * Return | |
1568 | * 0 on success, or a negative error in case of failure. | |
1569 | * | |
1570 | * int bpf_override_return(struct pt_reg *regs, u64 rc) | |
1571 | * Description | |
1572 | * Used for error injection, this helper uses kprobes to override | |
1573 | * the return value of the probed function, and to set it to *rc*. | |
1574 | * The first argument is the context *regs* on which the kprobe | |
1575 | * works. | |
1576 | * | |
1577 | * This helper works by setting setting the PC (program counter) | |
1578 | * to an override function which is run in place of the original | |
1579 | * probed function. This means the probed function is not run at | |
1580 | * all. The replacement function just returns with the required | |
1581 | * value. | |
1582 | * | |
1583 | * This helper has security implications, and thus is subject to | |
1584 | * restrictions. It is only available if the kernel was compiled | |
1585 | * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration | |
1586 | * option, and in this case it only works on functions tagged with | |
1587 | * **ALLOW_ERROR_INJECTION** in the kernel code. | |
1588 | * | |
1589 | * Also, the helper is only available for the architectures having | |
1590 | * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, | |
1591 | * x86 architecture is the only one to support this feature. | |
1592 | * Return | |
1593 | * 0 | |
1594 | * | |
a3ef8e9a | 1595 | * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) |
7aa79a86 QM |
1596 | * Description |
1597 | * Attempt to set the value of the **bpf_sock_ops_cb_flags** field | |
1598 | * for the full TCP socket associated to *bpf_sock_ops* to | |
1599 | * *argval*. | |
1600 | * | |
1601 | * The primary use of this field is to determine if there should | |
1602 | * be calls to eBPF programs of type | |
1603 | * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP | |
1604 | * code. A program of the same type can change its value, per | |
1605 | * connection and as necessary, when the connection is | |
1606 | * established. This field is directly accessible for reading, but | |
1607 | * this helper must be used for updates in order to return an | |
1608 | * error if an eBPF program tries to set a callback that is not | |
1609 | * supported in the current kernel. | |
1610 | * | |
1611 | * The supported callback values that *argval* can combine are: | |
1612 | * | |
1613 | * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) | |
1614 | * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) | |
1615 | * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) | |
1616 | * | |
1617 | * Here are some examples of where one could call such eBPF | |
1618 | * program: | |
1619 | * | |
1620 | * * When RTO fires. | |
1621 | * * When a packet is retransmitted. | |
1622 | * * When the connection terminates. | |
1623 | * * When a packet is sent. | |
1624 | * * When a packet is received. | |
1625 | * Return | |
1626 | * Code **-EINVAL** if the socket is not a full TCP socket; | |
1627 | * otherwise, a positive number containing the bits that could not | |
1628 | * be set is returned (which comes down to 0 if all bits were set | |
1629 | * as required). | |
1630 | * | |
ab127040 QM |
1631 | * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) |
1632 | * Description | |
1633 | * This helper is used in programs implementing policies at the | |
1634 | * socket level. If the message *msg* is allowed to pass (i.e. if | |
1635 | * the verdict eBPF program returns **SK_PASS**), redirect it to | |
1636 | * the socket referenced by *map* (of type | |
1637 | * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and | |
1638 | * egress interfaces can be used for redirection. The | |
1639 | * **BPF_F_INGRESS** value in *flags* is used to make the | |
1640 | * distinction (ingress path is selected if the flag is present, | |
1641 | * egress path otherwise). This is the only flag supported for now. | |
1642 | * Return | |
1643 | * **SK_PASS** on success, or **SK_DROP** on error. | |
1644 | * | |
1645 | * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) | |
1646 | * Description | |
1647 | * For socket policies, apply the verdict of the eBPF program to | |
1648 | * the next *bytes* (number of bytes) of message *msg*. | |
1649 | * | |
1650 | * For example, this helper can be used in the following cases: | |
1651 | * | |
1652 | * * A single **sendmsg**\ () or **sendfile**\ () system call | |
1653 | * contains multiple logical messages that the eBPF program is | |
1654 | * supposed to read and for which it should apply a verdict. | |
1655 | * * An eBPF program only cares to read the first *bytes* of a | |
1656 | * *msg*. If the message has a large payload, then setting up | |
1657 | * and calling the eBPF program repeatedly for all bytes, even | |
1658 | * though the verdict is already known, would create unnecessary | |
1659 | * overhead. | |
1660 | * | |
1661 | * When called from within an eBPF program, the helper sets a | |
1662 | * counter internal to the BPF infrastructure, that is used to | |
1663 | * apply the last verdict to the next *bytes*. If *bytes* is | |
1664 | * smaller than the current data being processed from a | |
1665 | * **sendmsg**\ () or **sendfile**\ () system call, the first | |
1666 | * *bytes* will be sent and the eBPF program will be re-run with | |
1667 | * the pointer for start of data pointing to byte number *bytes* | |
1668 | * **+ 1**. If *bytes* is larger than the current data being | |
1669 | * processed, then the eBPF verdict will be applied to multiple | |
1670 | * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are | |
1671 | * consumed. | |
1672 | * | |
1673 | * Note that if a socket closes with the internal counter holding | |
1674 | * a non-zero value, this is not a problem because data is not | |
1675 | * being buffered for *bytes* and is sent as it is received. | |
1676 | * Return | |
1677 | * 0 | |
1678 | * | |
1679 | * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) | |
1680 | * Description | |
1681 | * For socket policies, prevent the execution of the verdict eBPF | |
1682 | * program for message *msg* until *bytes* (byte number) have been | |
1683 | * accumulated. | |
1684 | * | |
1685 | * This can be used when one needs a specific number of bytes | |
1686 | * before a verdict can be assigned, even if the data spans | |
1687 | * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme | |
1688 | * case would be a user calling **sendmsg**\ () repeatedly with | |
1689 | * 1-byte long message segments. Obviously, this is bad for | |
1690 | * performance, but it is still valid. If the eBPF program needs | |
1691 | * *bytes* bytes to validate a header, this helper can be used to | |
1692 | * prevent the eBPF program to be called again until *bytes* have | |
1693 | * been accumulated. | |
1694 | * Return | |
1695 | * 0 | |
1696 | * | |
1697 | * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) | |
1698 | * Description | |
1699 | * For socket policies, pull in non-linear data from user space | |
1700 | * for *msg* and set pointers *msg*\ **->data** and *msg*\ | |
1701 | * **->data_end** to *start* and *end* bytes offsets into *msg*, | |
1702 | * respectively. | |
1703 | * | |
1704 | * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a | |
1705 | * *msg* it can only parse data that the (**data**, **data_end**) | |
1706 | * pointers have already consumed. For **sendmsg**\ () hooks this | |
1707 | * is likely the first scatterlist element. But for calls relying | |
1708 | * on the **sendpage** handler (e.g. **sendfile**\ ()) this will | |
1709 | * be the range (**0**, **0**) because the data is shared with | |
1710 | * user space and by default the objective is to avoid allowing | |
1711 | * user space to modify data while (or after) eBPF verdict is | |
1712 | * being decided. This helper can be used to pull in data and to | |
1713 | * set the start and end pointer to given values. Data will be | |
1714 | * copied if necessary (i.e. if data was not linear and if start | |
1715 | * and end pointers do not point to the same chunk). | |
1716 | * | |
1717 | * A call to this helper is susceptible to change the underlaying | |
1718 | * packet buffer. Therefore, at load time, all checks on pointers | |
1719 | * previously done by the verifier are invalidated and must be | |
1720 | * performed again, if the helper is used in combination with | |
1721 | * direct packet access. | |
1722 | * | |
1723 | * All values for *flags* are reserved for future usage, and must | |
1724 | * be left at zero. | |
1725 | * Return | |
1726 | * 0 on success, or a negative error in case of failure. | |
1727 | * | |
a3ef8e9a | 1728 | * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) |
7aa79a86 QM |
1729 | * Description |
1730 | * Bind the socket associated to *ctx* to the address pointed by | |
1731 | * *addr*, of length *addr_len*. This allows for making outgoing | |
1732 | * connection from the desired IP address, which can be useful for | |
1733 | * example when all processes inside a cgroup should use one | |
1734 | * single IP address on a host that has multiple IP configured. | |
1735 | * | |
1736 | * This helper works for IPv4 and IPv6, TCP and UDP sockets. The | |
1737 | * domain (*addr*\ **->sa_family**) must be **AF_INET** (or | |
1738 | * **AF_INET6**). Looking for a free port to bind to can be | |
1739 | * expensive, therefore binding to port is not permitted by the | |
1740 | * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) | |
1741 | * must be set to zero. | |
1742 | * Return | |
1743 | * 0 on success, or a negative error in case of failure. | |
2d020dd7 QM |
1744 | * |
1745 | * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) | |
1746 | * Description | |
1747 | * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is | |
1748 | * only possible to shrink the packet as of this writing, | |
1749 | * therefore *delta* must be a negative integer. | |
1750 | * | |
1751 | * A call to this helper is susceptible to change the underlaying | |
1752 | * packet buffer. Therefore, at load time, all checks on pointers | |
1753 | * previously done by the verifier are invalidated and must be | |
1754 | * performed again, if the helper is used in combination with | |
1755 | * direct packet access. | |
1756 | * Return | |
1757 | * 0 on success, or a negative error in case of failure. | |
1758 | * | |
1759 | * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) | |
1760 | * Description | |
1761 | * Retrieve the XFRM state (IP transform framework, see also | |
1762 | * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. | |
1763 | * | |
1764 | * The retrieved value is stored in the **struct bpf_xfrm_state** | |
1765 | * pointed by *xfrm_state* and of length *size*. | |
1766 | * | |
1767 | * All values for *flags* are reserved for future usage, and must | |
1768 | * be left at zero. | |
1769 | * | |
1770 | * This helper is available only if the kernel was compiled with | |
1771 | * **CONFIG_XFRM** configuration option. | |
1772 | * Return | |
1773 | * 0 on success, or a negative error in case of failure. | |
c195651e YS |
1774 | * |
1775 | * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) | |
1776 | * Description | |
79552fbc QM |
1777 | * Return a user or a kernel stack in bpf program provided buffer. |
1778 | * To achieve this, the helper needs *ctx*, which is a pointer | |
1779 | * to the context on which the tracing program is executed. | |
1780 | * To store the stacktrace, the bpf program provides *buf* with | |
1781 | * a nonnegative *size*. | |
1782 | * | |
1783 | * The last argument, *flags*, holds the number of stack frames to | |
1784 | * skip (from 0 to 255), masked with | |
1785 | * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set | |
1786 | * the following flags: | |
1787 | * | |
1788 | * **BPF_F_USER_STACK** | |
1789 | * Collect a user space stack instead of a kernel stack. | |
1790 | * **BPF_F_USER_BUILD_ID** | |
1791 | * Collect buildid+offset instead of ips for user stack, | |
1792 | * only valid if **BPF_F_USER_STACK** is also specified. | |
1793 | * | |
1794 | * **bpf_get_stack**\ () can collect up to | |
1795 | * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject | |
1796 | * to sufficient large buffer size. Note that | |
1797 | * this limit can be controlled with the **sysctl** program, and | |
1798 | * that it should be manually increased in order to profile long | |
1799 | * user stacks (such as stacks for Java programs). To do so, use: | |
1800 | * | |
1801 | * :: | |
1802 | * | |
1803 | * # sysctl kernel.perf_event_max_stack=<new value> | |
c195651e YS |
1804 | * |
1805 | * Return | |
1806 | * a non-negative value equal to or less than size on success, or | |
1807 | * a negative error in case of failure. | |
4e1ec56c DB |
1808 | * |
1809 | * int skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) | |
1810 | * Description | |
1811 | * This helper is similar to **bpf_skb_load_bytes**\ () in that | |
1812 | * it provides an easy way to load *len* bytes from *offset* | |
1813 | * from the packet associated to *skb*, into the buffer pointed | |
1814 | * by *to*. The difference to **bpf_skb_load_bytes**\ () is that | |
1815 | * a fifth argument *start_header* exists in order to select a | |
1816 | * base offset to start from. *start_header* can be one of: | |
1817 | * | |
1818 | * **BPF_HDR_START_MAC** | |
1819 | * Base offset to load data from is *skb*'s mac header. | |
1820 | * **BPF_HDR_START_NET** | |
1821 | * Base offset to load data from is *skb*'s network header. | |
1822 | * | |
1823 | * In general, "direct packet access" is the preferred method to | |
1824 | * access packet data, however, this helper is in particular useful | |
1825 | * in socket filters where *skb*\ **->data** does not always point | |
1826 | * to the start of the mac header and where "direct packet access" | |
1827 | * is not available. | |
1828 | * | |
1829 | * Return | |
1830 | * 0 on success, or a negative error in case of failure. | |
1831 | * | |
87f5fc7e DA |
1832 | * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) |
1833 | * Description | |
1834 | * Do FIB lookup in kernel tables using parameters in *params*. | |
1835 | * If lookup is successful and result shows packet is to be | |
1836 | * forwarded, the neighbor tables are searched for the nexthop. | |
1837 | * If successful (ie., FIB lookup shows forwarding and nexthop | |
1838 | * is resolved), the nexthop address is returned in ipv4_dst, | |
1839 | * ipv6_dst or mpls_out based on family, smac is set to mac | |
1840 | * address of egress device, dmac is set to nexthop mac address, | |
1841 | * rt_metric is set to metric from route. | |
1842 | * | |
1843 | * *plen* argument is the size of the passed in struct. | |
1844 | * *flags* argument can be one or more BPF_FIB_LOOKUP_ flags: | |
1845 | * | |
1846 | * **BPF_FIB_LOOKUP_DIRECT** means do a direct table lookup vs | |
1847 | * full lookup using FIB rules | |
1848 | * **BPF_FIB_LOOKUP_OUTPUT** means do lookup from an egress | |
1849 | * perspective (default is ingress) | |
1850 | * | |
1851 | * *ctx* is either **struct xdp_md** for XDP programs or | |
1852 | * **struct sk_buff** tc cls_act programs. | |
1853 | * | |
1854 | * Return | |
1855 | * Egress device index on success, 0 if packet needs to continue | |
1856 | * up the stack for further processing or a negative error in case | |
1857 | * of failure. | |
81110384 JF |
1858 | * |
1859 | * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags) | |
1860 | * Description | |
1861 | * Add an entry to, or update a sockhash *map* referencing sockets. | |
1862 | * The *skops* is used as a new value for the entry associated to | |
1863 | * *key*. *flags* is one of: | |
1864 | * | |
1865 | * **BPF_NOEXIST** | |
1866 | * The entry for *key* must not exist in the map. | |
1867 | * **BPF_EXIST** | |
1868 | * The entry for *key* must already exist in the map. | |
1869 | * **BPF_ANY** | |
1870 | * No condition on the existence of the entry for *key*. | |
1871 | * | |
1872 | * If the *map* has eBPF programs (parser and verdict), those will | |
1873 | * be inherited by the socket being added. If the socket is | |
1874 | * already attached to eBPF programs, this results in an error. | |
1875 | * Return | |
1876 | * 0 on success, or a negative error in case of failure. | |
1877 | * | |
1878 | * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) | |
1879 | * Description | |
1880 | * This helper is used in programs implementing policies at the | |
1881 | * socket level. If the message *msg* is allowed to pass (i.e. if | |
1882 | * the verdict eBPF program returns **SK_PASS**), redirect it to | |
1883 | * the socket referenced by *map* (of type | |
1884 | * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and | |
1885 | * egress interfaces can be used for redirection. The | |
1886 | * **BPF_F_INGRESS** value in *flags* is used to make the | |
1887 | * distinction (ingress path is selected if the flag is present, | |
1888 | * egress path otherwise). This is the only flag supported for now. | |
1889 | * Return | |
1890 | * **SK_PASS** on success, or **SK_DROP** on error. | |
1891 | * | |
1892 | * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) | |
1893 | * Description | |
1894 | * This helper is used in programs implementing policies at the | |
1895 | * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. | |
1896 | * if the verdeict eBPF program returns **SK_PASS**), redirect it | |
1897 | * to the socket referenced by *map* (of type | |
1898 | * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and | |
1899 | * egress interfaces can be used for redirection. The | |
1900 | * **BPF_F_INGRESS** value in *flags* is used to make the | |
1901 | * distinction (ingress path is selected if the flag is present, | |
1902 | * egress otherwise). This is the only flag supported for now. | |
1903 | * Return | |
1904 | * **SK_PASS** on success, or **SK_DROP** on error. | |
ebb676da TG |
1905 | */ |
1906 | #define __BPF_FUNC_MAPPER(FN) \ | |
1907 | FN(unspec), \ | |
1908 | FN(map_lookup_elem), \ | |
1909 | FN(map_update_elem), \ | |
1910 | FN(map_delete_elem), \ | |
1911 | FN(probe_read), \ | |
1912 | FN(ktime_get_ns), \ | |
1913 | FN(trace_printk), \ | |
1914 | FN(get_prandom_u32), \ | |
1915 | FN(get_smp_processor_id), \ | |
1916 | FN(skb_store_bytes), \ | |
1917 | FN(l3_csum_replace), \ | |
1918 | FN(l4_csum_replace), \ | |
1919 | FN(tail_call), \ | |
1920 | FN(clone_redirect), \ | |
1921 | FN(get_current_pid_tgid), \ | |
1922 | FN(get_current_uid_gid), \ | |
1923 | FN(get_current_comm), \ | |
1924 | FN(get_cgroup_classid), \ | |
1925 | FN(skb_vlan_push), \ | |
1926 | FN(skb_vlan_pop), \ | |
1927 | FN(skb_get_tunnel_key), \ | |
1928 | FN(skb_set_tunnel_key), \ | |
1929 | FN(perf_event_read), \ | |
1930 | FN(redirect), \ | |
1931 | FN(get_route_realm), \ | |
1932 | FN(perf_event_output), \ | |
1933 | FN(skb_load_bytes), \ | |
1934 | FN(get_stackid), \ | |
1935 | FN(csum_diff), \ | |
1936 | FN(skb_get_tunnel_opt), \ | |
1937 | FN(skb_set_tunnel_opt), \ | |
1938 | FN(skb_change_proto), \ | |
1939 | FN(skb_change_type), \ | |
1940 | FN(skb_under_cgroup), \ | |
1941 | FN(get_hash_recalc), \ | |
1942 | FN(get_current_task), \ | |
1943 | FN(probe_write_user), \ | |
1944 | FN(current_task_under_cgroup), \ | |
1945 | FN(skb_change_tail), \ | |
1946 | FN(skb_pull_data), \ | |
1947 | FN(csum_update), \ | |
1948 | FN(set_hash_invalid), \ | |
3a0af8fd | 1949 | FN(get_numa_node_id), \ |
17bedab2 | 1950 | FN(skb_change_head), \ |
a5e8c070 | 1951 | FN(xdp_adjust_head), \ |
91b8270f | 1952 | FN(probe_read_str), \ |
6acc5c29 | 1953 | FN(get_socket_cookie), \ |
ded092cd | 1954 | FN(get_socket_uid), \ |
8c4b4c7e | 1955 | FN(set_hash), \ |
2be7e212 | 1956 | FN(setsockopt), \ |
97f91a7c | 1957 | FN(skb_adjust_room), \ |
174a79ff JF |
1958 | FN(redirect_map), \ |
1959 | FN(sk_redirect_map), \ | |
1960 | FN(sock_map_update), \ | |
908432ca | 1961 | FN(xdp_adjust_meta), \ |
4bebdc7a | 1962 | FN(perf_event_read_value), \ |
cd86d1fd | 1963 | FN(perf_prog_read_value), \ |
9802d865 | 1964 | FN(getsockopt), \ |
b13d8807 | 1965 | FN(override_return), \ |
4f738adb | 1966 | FN(sock_ops_cb_flags_set), \ |
2a100317 | 1967 | FN(msg_redirect_map), \ |
91843d54 | 1968 | FN(msg_apply_bytes), \ |
015632bb | 1969 | FN(msg_cork_bytes), \ |
d74bad4e | 1970 | FN(msg_pull_data), \ |
b32cc5b9 | 1971 | FN(bind), \ |
12bed760 | 1972 | FN(xdp_adjust_tail), \ |
c195651e | 1973 | FN(skb_get_xfrm_state), \ |
4e1ec56c | 1974 | FN(get_stack), \ |
87f5fc7e | 1975 | FN(skb_load_bytes_relative), \ |
81110384 JF |
1976 | FN(fib_lookup), \ |
1977 | FN(sock_hash_update), \ | |
1978 | FN(msg_redirect_hash), \ | |
1979 | FN(sk_redirect_hash), | |
ebb676da | 1980 | |
09756af4 AS |
1981 | /* integer value in 'imm' field of BPF_CALL instruction selects which helper |
1982 | * function eBPF program intends to call | |
1983 | */ | |
ebb676da | 1984 | #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x |
09756af4 | 1985 | enum bpf_func_id { |
ebb676da | 1986 | __BPF_FUNC_MAPPER(__BPF_ENUM_FN) |
09756af4 AS |
1987 | __BPF_FUNC_MAX_ID, |
1988 | }; | |
ebb676da | 1989 | #undef __BPF_ENUM_FN |
09756af4 | 1990 | |
781c53bc DB |
1991 | /* All flags used by eBPF helper functions, placed here. */ |
1992 | ||
1993 | /* BPF_FUNC_skb_store_bytes flags. */ | |
1994 | #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) | |
8afd54c8 | 1995 | #define BPF_F_INVALIDATE_HASH (1ULL << 1) |
781c53bc DB |
1996 | |
1997 | /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. | |
1998 | * First 4 bits are for passing the header field size. | |
1999 | */ | |
2000 | #define BPF_F_HDR_FIELD_MASK 0xfULL | |
2001 | ||
2002 | /* BPF_FUNC_l4_csum_replace flags. */ | |
2003 | #define BPF_F_PSEUDO_HDR (1ULL << 4) | |
2f72959a | 2004 | #define BPF_F_MARK_MANGLED_0 (1ULL << 5) |
d1b662ad | 2005 | #define BPF_F_MARK_ENFORCE (1ULL << 6) |
781c53bc DB |
2006 | |
2007 | /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ | |
2008 | #define BPF_F_INGRESS (1ULL << 0) | |
2009 | ||
c6c33454 DB |
2010 | /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ |
2011 | #define BPF_F_TUNINFO_IPV6 (1ULL << 0) | |
2012 | ||
c195651e | 2013 | /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ |
d5a3b1f6 AS |
2014 | #define BPF_F_SKIP_FIELD_MASK 0xffULL |
2015 | #define BPF_F_USER_STACK (1ULL << 8) | |
c195651e | 2016 | /* flags used by BPF_FUNC_get_stackid only. */ |
d5a3b1f6 AS |
2017 | #define BPF_F_FAST_STACK_CMP (1ULL << 9) |
2018 | #define BPF_F_REUSE_STACKID (1ULL << 10) | |
c195651e YS |
2019 | /* flags used by BPF_FUNC_get_stack only. */ |
2020 | #define BPF_F_USER_BUILD_ID (1ULL << 11) | |
d5a3b1f6 | 2021 | |
2da897e5 DB |
2022 | /* BPF_FUNC_skb_set_tunnel_key flags. */ |
2023 | #define BPF_F_ZERO_CSUM_TX (1ULL << 1) | |
22080870 | 2024 | #define BPF_F_DONT_FRAGMENT (1ULL << 2) |
77a5196a | 2025 | #define BPF_F_SEQ_NUMBER (1ULL << 3) |
2da897e5 | 2026 | |
908432ca YS |
2027 | /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and |
2028 | * BPF_FUNC_perf_event_read_value flags. | |
2029 | */ | |
1e33759c DB |
2030 | #define BPF_F_INDEX_MASK 0xffffffffULL |
2031 | #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK | |
555c8a86 DB |
2032 | /* BPF_FUNC_perf_event_output for sk_buff input context. */ |
2033 | #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) | |
1e33759c | 2034 | |
2be7e212 DB |
2035 | /* Mode for BPF_FUNC_skb_adjust_room helper. */ |
2036 | enum bpf_adj_room_mode { | |
2037 | BPF_ADJ_ROOM_NET, | |
2038 | }; | |
2039 | ||
4e1ec56c DB |
2040 | /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ |
2041 | enum bpf_hdr_start_off { | |
2042 | BPF_HDR_START_MAC, | |
2043 | BPF_HDR_START_NET, | |
2044 | }; | |
2045 | ||
9bac3d6d AS |
2046 | /* user accessible mirror of in-kernel sk_buff. |
2047 | * new fields can only be added to the end of this structure | |
2048 | */ | |
2049 | struct __sk_buff { | |
2050 | __u32 len; | |
2051 | __u32 pkt_type; | |
2052 | __u32 mark; | |
2053 | __u32 queue_mapping; | |
c2497395 AS |
2054 | __u32 protocol; |
2055 | __u32 vlan_present; | |
2056 | __u32 vlan_tci; | |
27cd5452 | 2057 | __u32 vlan_proto; |
bcad5718 | 2058 | __u32 priority; |
37e82c2f AS |
2059 | __u32 ingress_ifindex; |
2060 | __u32 ifindex; | |
d691f9e8 AS |
2061 | __u32 tc_index; |
2062 | __u32 cb[5]; | |
ba7591d8 | 2063 | __u32 hash; |
045efa82 | 2064 | __u32 tc_classid; |
969bf05e AS |
2065 | __u32 data; |
2066 | __u32 data_end; | |
b1d9fc41 | 2067 | __u32 napi_id; |
8a31db56 | 2068 | |
de8f3a83 | 2069 | /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ |
8a31db56 JF |
2070 | __u32 family; |
2071 | __u32 remote_ip4; /* Stored in network byte order */ | |
2072 | __u32 local_ip4; /* Stored in network byte order */ | |
2073 | __u32 remote_ip6[4]; /* Stored in network byte order */ | |
2074 | __u32 local_ip6[4]; /* Stored in network byte order */ | |
2075 | __u32 remote_port; /* Stored in network byte order */ | |
2076 | __u32 local_port; /* stored in host byte order */ | |
de8f3a83 DB |
2077 | /* ... here. */ |
2078 | ||
2079 | __u32 data_meta; | |
9bac3d6d AS |
2080 | }; |
2081 | ||
d3aa45ce AS |
2082 | struct bpf_tunnel_key { |
2083 | __u32 tunnel_id; | |
c6c33454 DB |
2084 | union { |
2085 | __u32 remote_ipv4; | |
2086 | __u32 remote_ipv6[4]; | |
2087 | }; | |
2088 | __u8 tunnel_tos; | |
2089 | __u8 tunnel_ttl; | |
c0e760c9 | 2090 | __u16 tunnel_ext; |
4018ab18 | 2091 | __u32 tunnel_label; |
d3aa45ce AS |
2092 | }; |
2093 | ||
12bed760 EB |
2094 | /* user accessible mirror of in-kernel xfrm_state. |
2095 | * new fields can only be added to the end of this structure | |
2096 | */ | |
2097 | struct bpf_xfrm_state { | |
2098 | __u32 reqid; | |
2099 | __u32 spi; /* Stored in network byte order */ | |
2100 | __u16 family; | |
2101 | union { | |
2102 | __u32 remote_ipv4; /* Stored in network byte order */ | |
2103 | __u32 remote_ipv6[4]; /* Stored in network byte order */ | |
2104 | }; | |
2105 | }; | |
2106 | ||
3a0af8fd TG |
2107 | /* Generic BPF return codes which all BPF program types may support. |
2108 | * The values are binary compatible with their TC_ACT_* counter-part to | |
2109 | * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT | |
2110 | * programs. | |
2111 | * | |
2112 | * XDP is handled seprately, see XDP_*. | |
2113 | */ | |
2114 | enum bpf_ret_code { | |
2115 | BPF_OK = 0, | |
2116 | /* 1 reserved */ | |
2117 | BPF_DROP = 2, | |
2118 | /* 3-6 reserved */ | |
2119 | BPF_REDIRECT = 7, | |
2120 | /* >127 are reserved for prog type specific return codes */ | |
2121 | }; | |
2122 | ||
61023658 DA |
2123 | struct bpf_sock { |
2124 | __u32 bound_dev_if; | |
aa4c1037 DA |
2125 | __u32 family; |
2126 | __u32 type; | |
2127 | __u32 protocol; | |
482dca93 DA |
2128 | __u32 mark; |
2129 | __u32 priority; | |
aac3fc32 AI |
2130 | __u32 src_ip4; /* Allows 1,2,4-byte read. |
2131 | * Stored in network byte order. | |
2132 | */ | |
2133 | __u32 src_ip6[4]; /* Allows 1,2,4-byte read. | |
2134 | * Stored in network byte order. | |
2135 | */ | |
2136 | __u32 src_port; /* Allows 4-byte read. | |
2137 | * Stored in host byte order | |
2138 | */ | |
61023658 DA |
2139 | }; |
2140 | ||
17bedab2 MKL |
2141 | #define XDP_PACKET_HEADROOM 256 |
2142 | ||
6a773a15 BB |
2143 | /* User return codes for XDP prog type. |
2144 | * A valid XDP program must return one of these defined values. All other | |
9beb8bed DB |
2145 | * return codes are reserved for future use. Unknown return codes will |
2146 | * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). | |
6a773a15 BB |
2147 | */ |
2148 | enum xdp_action { | |
2149 | XDP_ABORTED = 0, | |
2150 | XDP_DROP, | |
2151 | XDP_PASS, | |
6ce96ca3 | 2152 | XDP_TX, |
814abfab | 2153 | XDP_REDIRECT, |
6a773a15 BB |
2154 | }; |
2155 | ||
2156 | /* user accessible metadata for XDP packet hook | |
2157 | * new fields must be added to the end of this structure | |
2158 | */ | |
2159 | struct xdp_md { | |
2160 | __u32 data; | |
2161 | __u32 data_end; | |
de8f3a83 | 2162 | __u32 data_meta; |
daaf24c6 | 2163 | /* Below access go through struct xdp_rxq_info */ |
02dd3291 JDB |
2164 | __u32 ingress_ifindex; /* rxq->dev->ifindex */ |
2165 | __u32 rx_queue_index; /* rxq->queue_index */ | |
6a773a15 BB |
2166 | }; |
2167 | ||
174a79ff | 2168 | enum sk_action { |
bfa64075 JF |
2169 | SK_DROP = 0, |
2170 | SK_PASS, | |
174a79ff JF |
2171 | }; |
2172 | ||
4f738adb JF |
2173 | /* user accessible metadata for SK_MSG packet hook, new fields must |
2174 | * be added to the end of this structure | |
2175 | */ | |
2176 | struct sk_msg_md { | |
2177 | void *data; | |
2178 | void *data_end; | |
2179 | }; | |
2180 | ||
1e270976 MKL |
2181 | #define BPF_TAG_SIZE 8 |
2182 | ||
2183 | struct bpf_prog_info { | |
2184 | __u32 type; | |
2185 | __u32 id; | |
2186 | __u8 tag[BPF_TAG_SIZE]; | |
2187 | __u32 jited_prog_len; | |
2188 | __u32 xlated_prog_len; | |
2189 | __aligned_u64 jited_prog_insns; | |
2190 | __aligned_u64 xlated_prog_insns; | |
cb4d2b3f MKL |
2191 | __u64 load_time; /* ns since boottime */ |
2192 | __u32 created_by_uid; | |
2193 | __u32 nr_map_ids; | |
2194 | __aligned_u64 map_ids; | |
067cae47 | 2195 | char name[BPF_OBJ_NAME_LEN]; |
675fc275 | 2196 | __u32 ifindex; |
b85fab0e | 2197 | __u32 gpl_compatible:1; |
675fc275 JK |
2198 | __u64 netns_dev; |
2199 | __u64 netns_ino; | |
1e270976 MKL |
2200 | } __attribute__((aligned(8))); |
2201 | ||
2202 | struct bpf_map_info { | |
2203 | __u32 type; | |
2204 | __u32 id; | |
2205 | __u32 key_size; | |
2206 | __u32 value_size; | |
2207 | __u32 max_entries; | |
2208 | __u32 map_flags; | |
067cae47 | 2209 | char name[BPF_OBJ_NAME_LEN]; |
52775b33 JK |
2210 | __u32 ifindex; |
2211 | __u64 netns_dev; | |
2212 | __u64 netns_ino; | |
78958fca MKL |
2213 | __u32 btf_id; |
2214 | __u32 btf_key_id; | |
2215 | __u32 btf_value_id; | |
1e270976 MKL |
2216 | } __attribute__((aligned(8))); |
2217 | ||
62dab84c MKL |
2218 | struct bpf_btf_info { |
2219 | __aligned_u64 btf; | |
2220 | __u32 btf_size; | |
2221 | __u32 id; | |
2222 | } __attribute__((aligned(8))); | |
2223 | ||
4fbac77d AI |
2224 | /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed |
2225 | * by user and intended to be used by socket (e.g. to bind to, depends on | |
2226 | * attach attach type). | |
2227 | */ | |
2228 | struct bpf_sock_addr { | |
2229 | __u32 user_family; /* Allows 4-byte read, but no write. */ | |
2230 | __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. | |
2231 | * Stored in network byte order. | |
2232 | */ | |
2233 | __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. | |
2234 | * Stored in network byte order. | |
2235 | */ | |
2236 | __u32 user_port; /* Allows 4-byte read and write. | |
2237 | * Stored in network byte order | |
2238 | */ | |
2239 | __u32 family; /* Allows 4-byte read, but no write */ | |
2240 | __u32 type; /* Allows 4-byte read, but no write */ | |
2241 | __u32 protocol; /* Allows 4-byte read, but no write */ | |
2242 | }; | |
2243 | ||
40304b2a LB |
2244 | /* User bpf_sock_ops struct to access socket values and specify request ops |
2245 | * and their replies. | |
2246 | * Some of this fields are in network (bigendian) byte order and may need | |
2247 | * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). | |
2248 | * New fields can only be added at the end of this structure | |
2249 | */ | |
2250 | struct bpf_sock_ops { | |
2251 | __u32 op; | |
2252 | union { | |
de525be2 LB |
2253 | __u32 args[4]; /* Optionally passed to bpf program */ |
2254 | __u32 reply; /* Returned by bpf program */ | |
2255 | __u32 replylong[4]; /* Optionally returned by bpf prog */ | |
40304b2a LB |
2256 | }; |
2257 | __u32 family; | |
2258 | __u32 remote_ip4; /* Stored in network byte order */ | |
2259 | __u32 local_ip4; /* Stored in network byte order */ | |
2260 | __u32 remote_ip6[4]; /* Stored in network byte order */ | |
2261 | __u32 local_ip6[4]; /* Stored in network byte order */ | |
2262 | __u32 remote_port; /* Stored in network byte order */ | |
2263 | __u32 local_port; /* stored in host byte order */ | |
f19397a5 LB |
2264 | __u32 is_fullsock; /* Some TCP fields are only valid if |
2265 | * there is a full socket. If not, the | |
2266 | * fields read as zero. | |
2267 | */ | |
2268 | __u32 snd_cwnd; | |
2269 | __u32 srtt_us; /* Averaged RTT << 3 in usecs */ | |
b13d8807 | 2270 | __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ |
44f0e430 LB |
2271 | __u32 state; |
2272 | __u32 rtt_min; | |
2273 | __u32 snd_ssthresh; | |
2274 | __u32 rcv_nxt; | |
2275 | __u32 snd_nxt; | |
2276 | __u32 snd_una; | |
2277 | __u32 mss_cache; | |
2278 | __u32 ecn_flags; | |
2279 | __u32 rate_delivered; | |
2280 | __u32 rate_interval_us; | |
2281 | __u32 packets_out; | |
2282 | __u32 retrans_out; | |
2283 | __u32 total_retrans; | |
2284 | __u32 segs_in; | |
2285 | __u32 data_segs_in; | |
2286 | __u32 segs_out; | |
2287 | __u32 data_segs_out; | |
2288 | __u32 lost_out; | |
2289 | __u32 sacked_out; | |
2290 | __u32 sk_txhash; | |
2291 | __u64 bytes_received; | |
2292 | __u64 bytes_acked; | |
40304b2a LB |
2293 | }; |
2294 | ||
b13d8807 | 2295 | /* Definitions for bpf_sock_ops_cb_flags */ |
f89013f6 | 2296 | #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) |
a31ad29e | 2297 | #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) |
d4487491 LB |
2298 | #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) |
2299 | #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently | |
b13d8807 LB |
2300 | * supported cb flags |
2301 | */ | |
2302 | ||
40304b2a LB |
2303 | /* List of known BPF sock_ops operators. |
2304 | * New entries can only be added at the end | |
2305 | */ | |
2306 | enum { | |
2307 | BPF_SOCK_OPS_VOID, | |
8550f328 LB |
2308 | BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or |
2309 | * -1 if default value should be used | |
2310 | */ | |
13d3b1eb LB |
2311 | BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized |
2312 | * window (in packets) or -1 if default | |
2313 | * value should be used | |
2314 | */ | |
9872a4bd LB |
2315 | BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an |
2316 | * active connection is initialized | |
2317 | */ | |
2318 | BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an | |
2319 | * active connection is | |
2320 | * established | |
2321 | */ | |
2322 | BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a | |
2323 | * passive connection is | |
2324 | * established | |
2325 | */ | |
91b5b21c LB |
2326 | BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control |
2327 | * needs ECN | |
2328 | */ | |
e6546ef6 LB |
2329 | BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is |
2330 | * based on the path and may be | |
2331 | * dependent on the congestion control | |
2332 | * algorithm. In general it indicates | |
2333 | * a congestion threshold. RTTs above | |
2334 | * this indicate congestion | |
2335 | */ | |
f89013f6 LB |
2336 | BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. |
2337 | * Arg1: value of icsk_retransmits | |
2338 | * Arg2: value of icsk_rto | |
2339 | * Arg3: whether RTO has expired | |
2340 | */ | |
a31ad29e LB |
2341 | BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. |
2342 | * Arg1: sequence number of 1st byte | |
2343 | * Arg2: # segments | |
2344 | * Arg3: return value of | |
2345 | * tcp_transmit_skb (0 => success) | |
2346 | */ | |
d4487491 LB |
2347 | BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. |
2348 | * Arg1: old_state | |
2349 | * Arg2: new_state | |
2350 | */ | |
2351 | }; | |
2352 | ||
2353 | /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect | |
2354 | * changes between the TCP and BPF versions. Ideally this should never happen. | |
2355 | * If it does, we need to add code to convert them before calling | |
2356 | * the BPF sock_ops function. | |
2357 | */ | |
2358 | enum { | |
2359 | BPF_TCP_ESTABLISHED = 1, | |
2360 | BPF_TCP_SYN_SENT, | |
2361 | BPF_TCP_SYN_RECV, | |
2362 | BPF_TCP_FIN_WAIT1, | |
2363 | BPF_TCP_FIN_WAIT2, | |
2364 | BPF_TCP_TIME_WAIT, | |
2365 | BPF_TCP_CLOSE, | |
2366 | BPF_TCP_CLOSE_WAIT, | |
2367 | BPF_TCP_LAST_ACK, | |
2368 | BPF_TCP_LISTEN, | |
2369 | BPF_TCP_CLOSING, /* Now a valid state */ | |
2370 | BPF_TCP_NEW_SYN_RECV, | |
2371 | ||
2372 | BPF_TCP_MAX_STATES /* Leave at the end! */ | |
40304b2a LB |
2373 | }; |
2374 | ||
fc747810 | 2375 | #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ |
13bf9641 | 2376 | #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ |
fc747810 | 2377 | |
908432ca YS |
2378 | struct bpf_perf_event_value { |
2379 | __u64 counter; | |
2380 | __u64 enabled; | |
2381 | __u64 running; | |
2382 | }; | |
2383 | ||
ebc614f6 RG |
2384 | #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) |
2385 | #define BPF_DEVCG_ACC_READ (1ULL << 1) | |
2386 | #define BPF_DEVCG_ACC_WRITE (1ULL << 2) | |
2387 | ||
2388 | #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) | |
2389 | #define BPF_DEVCG_DEV_CHAR (1ULL << 1) | |
2390 | ||
2391 | struct bpf_cgroup_dev_ctx { | |
06ef0ccb YS |
2392 | /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ |
2393 | __u32 access_type; | |
ebc614f6 RG |
2394 | __u32 major; |
2395 | __u32 minor; | |
2396 | }; | |
2397 | ||
c4f6699d AS |
2398 | struct bpf_raw_tracepoint_args { |
2399 | __u64 args[0]; | |
2400 | }; | |
2401 | ||
87f5fc7e DA |
2402 | /* DIRECT: Skip the FIB rules and go to FIB table associated with device |
2403 | * OUTPUT: Do lookup from egress perspective; default is ingress | |
2404 | */ | |
2405 | #define BPF_FIB_LOOKUP_DIRECT BIT(0) | |
2406 | #define BPF_FIB_LOOKUP_OUTPUT BIT(1) | |
2407 | ||
2408 | struct bpf_fib_lookup { | |
2409 | /* input */ | |
2410 | __u8 family; /* network family, AF_INET, AF_INET6, AF_MPLS */ | |
2411 | ||
2412 | /* set if lookup is to consider L4 data - e.g., FIB rules */ | |
2413 | __u8 l4_protocol; | |
2414 | __be16 sport; | |
2415 | __be16 dport; | |
2416 | ||
2417 | /* total length of packet from network header - used for MTU check */ | |
2418 | __u16 tot_len; | |
2419 | __u32 ifindex; /* L3 device index for lookup */ | |
2420 | ||
2421 | union { | |
2422 | /* inputs to lookup */ | |
2423 | __u8 tos; /* AF_INET */ | |
2424 | __be32 flowlabel; /* AF_INET6 */ | |
2425 | ||
2426 | /* output: metric of fib result */ | |
2427 | __u32 rt_metric; | |
2428 | }; | |
2429 | ||
2430 | union { | |
2431 | __be32 mpls_in; | |
2432 | __be32 ipv4_src; | |
2433 | __u32 ipv6_src[4]; /* in6_addr; network order */ | |
2434 | }; | |
2435 | ||
2436 | /* input to bpf_fib_lookup, *dst is destination address. | |
2437 | * output: bpf_fib_lookup sets to gateway address | |
2438 | */ | |
2439 | union { | |
2440 | /* return for MPLS lookups */ | |
2441 | __be32 mpls_out[4]; /* support up to 4 labels */ | |
2442 | __be32 ipv4_dst; | |
2443 | __u32 ipv6_dst[4]; /* in6_addr; network order */ | |
2444 | }; | |
2445 | ||
2446 | /* output */ | |
2447 | __be16 h_vlan_proto; | |
2448 | __be16 h_vlan_TCI; | |
2449 | __u8 smac[6]; /* ETH_ALEN */ | |
2450 | __u8 dmac[6]; /* ETH_ALEN */ | |
2451 | }; | |
2452 | ||
daedfb22 | 2453 | #endif /* _UAPI__LINUX_BPF_H__ */ |