Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
1da177e4 LT |
2 | /* |
3 | * Linux Socket Filter Data Structures | |
4 | */ | |
1da177e4 LT |
5 | #ifndef __LINUX_FILTER_H__ |
6 | #define __LINUX_FILTER_H__ | |
7 | ||
b954d834 DB |
8 | #include <stdarg.h> |
9 | ||
60063497 | 10 | #include <linux/atomic.h> |
4c355cdf | 11 | #include <linux/refcount.h> |
0c5fe1b4 | 12 | #include <linux/compat.h> |
9f12fbe6 | 13 | #include <linux/skbuff.h> |
b954d834 DB |
14 | #include <linux/linkage.h> |
15 | #include <linux/printk.h> | |
d45ed4a4 | 16 | #include <linux/workqueue.h> |
b13138ef | 17 | #include <linux/sched.h> |
4f3446bb | 18 | #include <linux/capability.h> |
7bd509e3 | 19 | #include <linux/cryptohash.h> |
820a0b24 | 20 | #include <linux/set_memory.h> |
7105e828 | 21 | #include <linux/kallsyms.h> |
6d5fc195 | 22 | #include <linux/if_vlan.h> |
d53d2f78 | 23 | #include <linux/vmalloc.h> |
4f3446bb | 24 | |
ff936a04 | 25 | #include <net/sch_generic.h> |
b954d834 | 26 | |
d9b8aada | 27 | #include <asm/byteorder.h> |
b954d834 | 28 | #include <uapi/linux/filter.h> |
daedfb22 | 29 | #include <uapi/linux/bpf.h> |
60a3b225 DB |
30 | |
31 | struct sk_buff; | |
32 | struct sock; | |
33 | struct seccomp_data; | |
09756af4 | 34 | struct bpf_prog_aux; |
297dd12c | 35 | struct xdp_rxq_info; |
106ca27f | 36 | struct xdp_buff; |
2dbb9b9e | 37 | struct sock_reuseport; |
7b146ceb AI |
38 | struct ctl_table; |
39 | struct ctl_table_header; | |
792d4b5c | 40 | |
30743837 DB |
41 | /* ArgX, context and stack frame pointer register positions. Note, |
42 | * Arg1, Arg2, Arg3, etc are used as argument mappings of function | |
43 | * calls in BPF_CALL instruction. | |
44 | */ | |
45 | #define BPF_REG_ARG1 BPF_REG_1 | |
46 | #define BPF_REG_ARG2 BPF_REG_2 | |
47 | #define BPF_REG_ARG3 BPF_REG_3 | |
48 | #define BPF_REG_ARG4 BPF_REG_4 | |
49 | #define BPF_REG_ARG5 BPF_REG_5 | |
50 | #define BPF_REG_CTX BPF_REG_6 | |
51 | #define BPF_REG_FP BPF_REG_10 | |
52 | ||
53 | /* Additional register mappings for converted user programs. */ | |
54 | #define BPF_REG_A BPF_REG_0 | |
55 | #define BPF_REG_X BPF_REG_7 | |
e0cea7ce DB |
56 | #define BPF_REG_TMP BPF_REG_2 /* scratch reg */ |
57 | #define BPF_REG_D BPF_REG_8 /* data, callee-saved */ | |
58 | #define BPF_REG_H BPF_REG_9 /* hlen, callee-saved */ | |
bd4cf0ed | 59 | |
9b73bfdd | 60 | /* Kernel hidden auxiliary/helper register. */ |
4f3446bb | 61 | #define BPF_REG_AX MAX_BPF_REG |
144cd91c DB |
62 | #define MAX_BPF_EXT_REG (MAX_BPF_REG + 1) |
63 | #define MAX_BPF_JIT_REG MAX_BPF_EXT_REG | |
4f3446bb | 64 | |
71189fa9 AS |
65 | /* unused opcode to mark special call to bpf_tail_call() helper */ |
66 | #define BPF_TAIL_CALL 0xf0 | |
67 | ||
1ea47e01 AS |
68 | /* unused opcode to mark call to interpreter with arguments */ |
69 | #define BPF_CALL_ARGS 0xe0 | |
70 | ||
74451e66 DB |
71 | /* As per nm, we expose JITed images as text (code) section for |
72 | * kallsyms. That way, tools like perf can find it to match | |
73 | * addresses. | |
74 | */ | |
75 | #define BPF_SYM_ELF_TYPE 't' | |
76 | ||
bd4cf0ed AS |
77 | /* BPF program can access up to 512 bytes of stack space. */ |
78 | #define MAX_BPF_STACK 512 | |
79 | ||
f8f6d679 DB |
80 | /* Helper macros for filter block array initializers. */ |
81 | ||
e430f34e | 82 | /* ALU ops on registers, bpf_add|sub|...: dst_reg += src_reg */ |
f8f6d679 | 83 | |
e430f34e | 84 | #define BPF_ALU64_REG(OP, DST, SRC) \ |
2695fb55 | 85 | ((struct bpf_insn) { \ |
f8f6d679 | 86 | .code = BPF_ALU64 | BPF_OP(OP) | BPF_X, \ |
e430f34e AS |
87 | .dst_reg = DST, \ |
88 | .src_reg = SRC, \ | |
f8f6d679 DB |
89 | .off = 0, \ |
90 | .imm = 0 }) | |
91 | ||
e430f34e | 92 | #define BPF_ALU32_REG(OP, DST, SRC) \ |
2695fb55 | 93 | ((struct bpf_insn) { \ |
f8f6d679 | 94 | .code = BPF_ALU | BPF_OP(OP) | BPF_X, \ |
e430f34e AS |
95 | .dst_reg = DST, \ |
96 | .src_reg = SRC, \ | |
f8f6d679 DB |
97 | .off = 0, \ |
98 | .imm = 0 }) | |
99 | ||
e430f34e | 100 | /* ALU ops on immediates, bpf_add|sub|...: dst_reg += imm32 */ |
f8f6d679 | 101 | |
e430f34e | 102 | #define BPF_ALU64_IMM(OP, DST, IMM) \ |
2695fb55 | 103 | ((struct bpf_insn) { \ |
f8f6d679 | 104 | .code = BPF_ALU64 | BPF_OP(OP) | BPF_K, \ |
e430f34e AS |
105 | .dst_reg = DST, \ |
106 | .src_reg = 0, \ | |
f8f6d679 DB |
107 | .off = 0, \ |
108 | .imm = IMM }) | |
109 | ||
e430f34e | 110 | #define BPF_ALU32_IMM(OP, DST, IMM) \ |
2695fb55 | 111 | ((struct bpf_insn) { \ |
f8f6d679 | 112 | .code = BPF_ALU | BPF_OP(OP) | BPF_K, \ |
e430f34e AS |
113 | .dst_reg = DST, \ |
114 | .src_reg = 0, \ | |
f8f6d679 DB |
115 | .off = 0, \ |
116 | .imm = IMM }) | |
117 | ||
118 | /* Endianess conversion, cpu_to_{l,b}e(), {l,b}e_to_cpu() */ | |
119 | ||
e430f34e | 120 | #define BPF_ENDIAN(TYPE, DST, LEN) \ |
2695fb55 | 121 | ((struct bpf_insn) { \ |
f8f6d679 | 122 | .code = BPF_ALU | BPF_END | BPF_SRC(TYPE), \ |
e430f34e AS |
123 | .dst_reg = DST, \ |
124 | .src_reg = 0, \ | |
f8f6d679 DB |
125 | .off = 0, \ |
126 | .imm = LEN }) | |
127 | ||
e430f34e | 128 | /* Short form of mov, dst_reg = src_reg */ |
f8f6d679 | 129 | |
e430f34e | 130 | #define BPF_MOV64_REG(DST, SRC) \ |
2695fb55 | 131 | ((struct bpf_insn) { \ |
f8f6d679 | 132 | .code = BPF_ALU64 | BPF_MOV | BPF_X, \ |
e430f34e AS |
133 | .dst_reg = DST, \ |
134 | .src_reg = SRC, \ | |
f8f6d679 DB |
135 | .off = 0, \ |
136 | .imm = 0 }) | |
137 | ||
e430f34e | 138 | #define BPF_MOV32_REG(DST, SRC) \ |
2695fb55 | 139 | ((struct bpf_insn) { \ |
f8f6d679 | 140 | .code = BPF_ALU | BPF_MOV | BPF_X, \ |
e430f34e AS |
141 | .dst_reg = DST, \ |
142 | .src_reg = SRC, \ | |
f8f6d679 DB |
143 | .off = 0, \ |
144 | .imm = 0 }) | |
145 | ||
e430f34e | 146 | /* Short form of mov, dst_reg = imm32 */ |
f8f6d679 | 147 | |
e430f34e | 148 | #define BPF_MOV64_IMM(DST, IMM) \ |
2695fb55 | 149 | ((struct bpf_insn) { \ |
f8f6d679 | 150 | .code = BPF_ALU64 | BPF_MOV | BPF_K, \ |
e430f34e AS |
151 | .dst_reg = DST, \ |
152 | .src_reg = 0, \ | |
f8f6d679 DB |
153 | .off = 0, \ |
154 | .imm = IMM }) | |
155 | ||
e430f34e | 156 | #define BPF_MOV32_IMM(DST, IMM) \ |
2695fb55 | 157 | ((struct bpf_insn) { \ |
f8f6d679 | 158 | .code = BPF_ALU | BPF_MOV | BPF_K, \ |
e430f34e AS |
159 | .dst_reg = DST, \ |
160 | .src_reg = 0, \ | |
f8f6d679 DB |
161 | .off = 0, \ |
162 | .imm = IMM }) | |
163 | ||
7d134041 JW |
164 | /* Special form of mov32, used for doing explicit zero extension on dst. */ |
165 | #define BPF_ZEXT_REG(DST) \ | |
166 | ((struct bpf_insn) { \ | |
167 | .code = BPF_ALU | BPF_MOV | BPF_X, \ | |
168 | .dst_reg = DST, \ | |
169 | .src_reg = DST, \ | |
170 | .off = 0, \ | |
171 | .imm = 1 }) | |
172 | ||
173 | static inline bool insn_is_zext(const struct bpf_insn *insn) | |
174 | { | |
175 | return insn->code == (BPF_ALU | BPF_MOV | BPF_X) && insn->imm == 1; | |
176 | } | |
177 | ||
02ab695b AS |
178 | /* BPF_LD_IMM64 macro encodes single 'load 64-bit immediate' insn */ |
179 | #define BPF_LD_IMM64(DST, IMM) \ | |
180 | BPF_LD_IMM64_RAW(DST, 0, IMM) | |
181 | ||
182 | #define BPF_LD_IMM64_RAW(DST, SRC, IMM) \ | |
183 | ((struct bpf_insn) { \ | |
184 | .code = BPF_LD | BPF_DW | BPF_IMM, \ | |
185 | .dst_reg = DST, \ | |
186 | .src_reg = SRC, \ | |
187 | .off = 0, \ | |
188 | .imm = (__u32) (IMM) }), \ | |
189 | ((struct bpf_insn) { \ | |
190 | .code = 0, /* zero is reserved opcode */ \ | |
191 | .dst_reg = 0, \ | |
192 | .src_reg = 0, \ | |
193 | .off = 0, \ | |
194 | .imm = ((__u64) (IMM)) >> 32 }) | |
195 | ||
0246e64d AS |
196 | /* pseudo BPF_LD_IMM64 insn used to refer to process-local map_fd */ |
197 | #define BPF_LD_MAP_FD(DST, MAP_FD) \ | |
198 | BPF_LD_IMM64_RAW(DST, BPF_PSEUDO_MAP_FD, MAP_FD) | |
199 | ||
e430f34e | 200 | /* Short form of mov based on type, BPF_X: dst_reg = src_reg, BPF_K: dst_reg = imm32 */ |
f8f6d679 | 201 | |
e430f34e | 202 | #define BPF_MOV64_RAW(TYPE, DST, SRC, IMM) \ |
2695fb55 | 203 | ((struct bpf_insn) { \ |
f8f6d679 | 204 | .code = BPF_ALU64 | BPF_MOV | BPF_SRC(TYPE), \ |
e430f34e AS |
205 | .dst_reg = DST, \ |
206 | .src_reg = SRC, \ | |
f8f6d679 DB |
207 | .off = 0, \ |
208 | .imm = IMM }) | |
209 | ||
e430f34e | 210 | #define BPF_MOV32_RAW(TYPE, DST, SRC, IMM) \ |
2695fb55 | 211 | ((struct bpf_insn) { \ |
f8f6d679 | 212 | .code = BPF_ALU | BPF_MOV | BPF_SRC(TYPE), \ |
e430f34e AS |
213 | .dst_reg = DST, \ |
214 | .src_reg = SRC, \ | |
f8f6d679 DB |
215 | .off = 0, \ |
216 | .imm = IMM }) | |
217 | ||
e430f34e | 218 | /* Direct packet access, R0 = *(uint *) (skb->data + imm32) */ |
f8f6d679 | 219 | |
e430f34e | 220 | #define BPF_LD_ABS(SIZE, IMM) \ |
2695fb55 | 221 | ((struct bpf_insn) { \ |
f8f6d679 | 222 | .code = BPF_LD | BPF_SIZE(SIZE) | BPF_ABS, \ |
e430f34e AS |
223 | .dst_reg = 0, \ |
224 | .src_reg = 0, \ | |
f8f6d679 | 225 | .off = 0, \ |
e430f34e | 226 | .imm = IMM }) |
f8f6d679 | 227 | |
e430f34e | 228 | /* Indirect packet access, R0 = *(uint *) (skb->data + src_reg + imm32) */ |
f8f6d679 | 229 | |
e430f34e | 230 | #define BPF_LD_IND(SIZE, SRC, IMM) \ |
2695fb55 | 231 | ((struct bpf_insn) { \ |
f8f6d679 | 232 | .code = BPF_LD | BPF_SIZE(SIZE) | BPF_IND, \ |
e430f34e AS |
233 | .dst_reg = 0, \ |
234 | .src_reg = SRC, \ | |
f8f6d679 | 235 | .off = 0, \ |
e430f34e | 236 | .imm = IMM }) |
f8f6d679 | 237 | |
e430f34e | 238 | /* Memory load, dst_reg = *(uint *) (src_reg + off16) */ |
f8f6d679 | 239 | |
e430f34e | 240 | #define BPF_LDX_MEM(SIZE, DST, SRC, OFF) \ |
2695fb55 | 241 | ((struct bpf_insn) { \ |
f8f6d679 | 242 | .code = BPF_LDX | BPF_SIZE(SIZE) | BPF_MEM, \ |
e430f34e AS |
243 | .dst_reg = DST, \ |
244 | .src_reg = SRC, \ | |
f8f6d679 DB |
245 | .off = OFF, \ |
246 | .imm = 0 }) | |
247 | ||
e430f34e AS |
248 | /* Memory store, *(uint *) (dst_reg + off16) = src_reg */ |
249 | ||
250 | #define BPF_STX_MEM(SIZE, DST, SRC, OFF) \ | |
2695fb55 | 251 | ((struct bpf_insn) { \ |
f8f6d679 | 252 | .code = BPF_STX | BPF_SIZE(SIZE) | BPF_MEM, \ |
e430f34e AS |
253 | .dst_reg = DST, \ |
254 | .src_reg = SRC, \ | |
f8f6d679 DB |
255 | .off = OFF, \ |
256 | .imm = 0 }) | |
257 | ||
cffc642d MH |
258 | /* Atomic memory add, *(uint *)(dst_reg + off16) += src_reg */ |
259 | ||
260 | #define BPF_STX_XADD(SIZE, DST, SRC, OFF) \ | |
261 | ((struct bpf_insn) { \ | |
262 | .code = BPF_STX | BPF_SIZE(SIZE) | BPF_XADD, \ | |
263 | .dst_reg = DST, \ | |
264 | .src_reg = SRC, \ | |
265 | .off = OFF, \ | |
266 | .imm = 0 }) | |
267 | ||
e430f34e AS |
268 | /* Memory store, *(uint *) (dst_reg + off16) = imm32 */ |
269 | ||
270 | #define BPF_ST_MEM(SIZE, DST, OFF, IMM) \ | |
2695fb55 | 271 | ((struct bpf_insn) { \ |
e430f34e AS |
272 | .code = BPF_ST | BPF_SIZE(SIZE) | BPF_MEM, \ |
273 | .dst_reg = DST, \ | |
274 | .src_reg = 0, \ | |
275 | .off = OFF, \ | |
276 | .imm = IMM }) | |
277 | ||
278 | /* Conditional jumps against registers, if (dst_reg 'op' src_reg) goto pc + off16 */ | |
f8f6d679 | 279 | |
e430f34e | 280 | #define BPF_JMP_REG(OP, DST, SRC, OFF) \ |
2695fb55 | 281 | ((struct bpf_insn) { \ |
f8f6d679 | 282 | .code = BPF_JMP | BPF_OP(OP) | BPF_X, \ |
e430f34e AS |
283 | .dst_reg = DST, \ |
284 | .src_reg = SRC, \ | |
f8f6d679 DB |
285 | .off = OFF, \ |
286 | .imm = 0 }) | |
287 | ||
e430f34e | 288 | /* Conditional jumps against immediates, if (dst_reg 'op' imm32) goto pc + off16 */ |
f8f6d679 | 289 | |
e430f34e | 290 | #define BPF_JMP_IMM(OP, DST, IMM, OFF) \ |
2695fb55 | 291 | ((struct bpf_insn) { \ |
f8f6d679 | 292 | .code = BPF_JMP | BPF_OP(OP) | BPF_K, \ |
e430f34e AS |
293 | .dst_reg = DST, \ |
294 | .src_reg = 0, \ | |
f8f6d679 | 295 | .off = OFF, \ |
a7b76c88 JW |
296 | .imm = IMM }) |
297 | ||
298 | /* Like BPF_JMP_REG, but with 32-bit wide operands for comparison. */ | |
299 | ||
300 | #define BPF_JMP32_REG(OP, DST, SRC, OFF) \ | |
301 | ((struct bpf_insn) { \ | |
302 | .code = BPF_JMP32 | BPF_OP(OP) | BPF_X, \ | |
303 | .dst_reg = DST, \ | |
304 | .src_reg = SRC, \ | |
305 | .off = OFF, \ | |
306 | .imm = 0 }) | |
307 | ||
308 | /* Like BPF_JMP_IMM, but with 32-bit wide operands for comparison. */ | |
309 | ||
310 | #define BPF_JMP32_IMM(OP, DST, IMM, OFF) \ | |
311 | ((struct bpf_insn) { \ | |
312 | .code = BPF_JMP32 | BPF_OP(OP) | BPF_K, \ | |
313 | .dst_reg = DST, \ | |
314 | .src_reg = 0, \ | |
315 | .off = OFF, \ | |
f8f6d679 DB |
316 | .imm = IMM }) |
317 | ||
614d0d77 DB |
318 | /* Unconditional jumps, goto pc + off16 */ |
319 | ||
320 | #define BPF_JMP_A(OFF) \ | |
321 | ((struct bpf_insn) { \ | |
322 | .code = BPF_JMP | BPF_JA, \ | |
323 | .dst_reg = 0, \ | |
324 | .src_reg = 0, \ | |
325 | .off = OFF, \ | |
326 | .imm = 0 }) | |
327 | ||
06be0864 DB |
328 | /* Relative call */ |
329 | ||
330 | #define BPF_CALL_REL(TGT) \ | |
331 | ((struct bpf_insn) { \ | |
332 | .code = BPF_JMP | BPF_CALL, \ | |
333 | .dst_reg = 0, \ | |
334 | .src_reg = BPF_PSEUDO_CALL, \ | |
335 | .off = 0, \ | |
336 | .imm = TGT }) | |
337 | ||
f8f6d679 DB |
338 | /* Function call */ |
339 | ||
09772d92 DB |
340 | #define BPF_CAST_CALL(x) \ |
341 | ((u64 (*)(u64, u64, u64, u64, u64))(x)) | |
342 | ||
f8f6d679 | 343 | #define BPF_EMIT_CALL(FUNC) \ |
2695fb55 | 344 | ((struct bpf_insn) { \ |
f8f6d679 | 345 | .code = BPF_JMP | BPF_CALL, \ |
e430f34e AS |
346 | .dst_reg = 0, \ |
347 | .src_reg = 0, \ | |
f8f6d679 DB |
348 | .off = 0, \ |
349 | .imm = ((FUNC) - __bpf_call_base) }) | |
350 | ||
351 | /* Raw code statement block */ | |
352 | ||
e430f34e | 353 | #define BPF_RAW_INSN(CODE, DST, SRC, OFF, IMM) \ |
2695fb55 | 354 | ((struct bpf_insn) { \ |
f8f6d679 | 355 | .code = CODE, \ |
e430f34e AS |
356 | .dst_reg = DST, \ |
357 | .src_reg = SRC, \ | |
f8f6d679 DB |
358 | .off = OFF, \ |
359 | .imm = IMM }) | |
360 | ||
361 | /* Program exit */ | |
362 | ||
363 | #define BPF_EXIT_INSN() \ | |
2695fb55 | 364 | ((struct bpf_insn) { \ |
f8f6d679 | 365 | .code = BPF_JMP | BPF_EXIT, \ |
e430f34e AS |
366 | .dst_reg = 0, \ |
367 | .src_reg = 0, \ | |
f8f6d679 DB |
368 | .off = 0, \ |
369 | .imm = 0 }) | |
370 | ||
a4afd37b DB |
371 | /* Internal classic blocks for direct assignment */ |
372 | ||
373 | #define __BPF_STMT(CODE, K) \ | |
374 | ((struct sock_filter) BPF_STMT(CODE, K)) | |
375 | ||
376 | #define __BPF_JUMP(CODE, K, JT, JF) \ | |
377 | ((struct sock_filter) BPF_JUMP(CODE, K, JT, JF)) | |
378 | ||
f8f6d679 DB |
379 | #define bytes_to_bpf_size(bytes) \ |
380 | ({ \ | |
381 | int bpf_size = -EINVAL; \ | |
382 | \ | |
383 | if (bytes == sizeof(u8)) \ | |
384 | bpf_size = BPF_B; \ | |
385 | else if (bytes == sizeof(u16)) \ | |
386 | bpf_size = BPF_H; \ | |
387 | else if (bytes == sizeof(u32)) \ | |
388 | bpf_size = BPF_W; \ | |
389 | else if (bytes == sizeof(u64)) \ | |
390 | bpf_size = BPF_DW; \ | |
391 | \ | |
392 | bpf_size; \ | |
393 | }) | |
9739eef1 | 394 | |
f96da094 DB |
395 | #define bpf_size_to_bytes(bpf_size) \ |
396 | ({ \ | |
397 | int bytes = -EINVAL; \ | |
398 | \ | |
399 | if (bpf_size == BPF_B) \ | |
400 | bytes = sizeof(u8); \ | |
401 | else if (bpf_size == BPF_H) \ | |
402 | bytes = sizeof(u16); \ | |
403 | else if (bpf_size == BPF_W) \ | |
404 | bytes = sizeof(u32); \ | |
405 | else if (bpf_size == BPF_DW) \ | |
406 | bytes = sizeof(u64); \ | |
407 | \ | |
408 | bytes; \ | |
409 | }) | |
410 | ||
f035a515 DB |
411 | #define BPF_SIZEOF(type) \ |
412 | ({ \ | |
413 | const int __size = bytes_to_bpf_size(sizeof(type)); \ | |
414 | BUILD_BUG_ON(__size < 0); \ | |
415 | __size; \ | |
416 | }) | |
417 | ||
418 | #define BPF_FIELD_SIZEOF(type, field) \ | |
419 | ({ \ | |
420 | const int __size = bytes_to_bpf_size(FIELD_SIZEOF(type, field)); \ | |
421 | BUILD_BUG_ON(__size < 0); \ | |
422 | __size; \ | |
423 | }) | |
424 | ||
f96da094 DB |
425 | #define BPF_LDST_BYTES(insn) \ |
426 | ({ \ | |
e59ac634 | 427 | const int __size = bpf_size_to_bytes(BPF_SIZE((insn)->code)); \ |
f96da094 DB |
428 | WARN_ON(__size < 0); \ |
429 | __size; \ | |
430 | }) | |
431 | ||
f3694e00 DB |
432 | #define __BPF_MAP_0(m, v, ...) v |
433 | #define __BPF_MAP_1(m, v, t, a, ...) m(t, a) | |
434 | #define __BPF_MAP_2(m, v, t, a, ...) m(t, a), __BPF_MAP_1(m, v, __VA_ARGS__) | |
435 | #define __BPF_MAP_3(m, v, t, a, ...) m(t, a), __BPF_MAP_2(m, v, __VA_ARGS__) | |
436 | #define __BPF_MAP_4(m, v, t, a, ...) m(t, a), __BPF_MAP_3(m, v, __VA_ARGS__) | |
437 | #define __BPF_MAP_5(m, v, t, a, ...) m(t, a), __BPF_MAP_4(m, v, __VA_ARGS__) | |
438 | ||
439 | #define __BPF_REG_0(...) __BPF_PAD(5) | |
440 | #define __BPF_REG_1(...) __BPF_MAP(1, __VA_ARGS__), __BPF_PAD(4) | |
441 | #define __BPF_REG_2(...) __BPF_MAP(2, __VA_ARGS__), __BPF_PAD(3) | |
442 | #define __BPF_REG_3(...) __BPF_MAP(3, __VA_ARGS__), __BPF_PAD(2) | |
443 | #define __BPF_REG_4(...) __BPF_MAP(4, __VA_ARGS__), __BPF_PAD(1) | |
444 | #define __BPF_REG_5(...) __BPF_MAP(5, __VA_ARGS__) | |
445 | ||
446 | #define __BPF_MAP(n, ...) __BPF_MAP_##n(__VA_ARGS__) | |
447 | #define __BPF_REG(n, ...) __BPF_REG_##n(__VA_ARGS__) | |
448 | ||
449 | #define __BPF_CAST(t, a) \ | |
450 | (__force t) \ | |
451 | (__force \ | |
452 | typeof(__builtin_choose_expr(sizeof(t) == sizeof(unsigned long), \ | |
453 | (unsigned long)0, (t)0))) a | |
454 | #define __BPF_V void | |
455 | #define __BPF_N | |
456 | ||
457 | #define __BPF_DECL_ARGS(t, a) t a | |
458 | #define __BPF_DECL_REGS(t, a) u64 a | |
459 | ||
460 | #define __BPF_PAD(n) \ | |
461 | __BPF_MAP(n, __BPF_DECL_ARGS, __BPF_N, u64, __ur_1, u64, __ur_2, \ | |
462 | u64, __ur_3, u64, __ur_4, u64, __ur_5) | |
463 | ||
464 | #define BPF_CALL_x(x, name, ...) \ | |
465 | static __always_inline \ | |
466 | u64 ____##name(__BPF_MAP(x, __BPF_DECL_ARGS, __BPF_V, __VA_ARGS__)); \ | |
467 | u64 name(__BPF_REG(x, __BPF_DECL_REGS, __BPF_N, __VA_ARGS__)); \ | |
468 | u64 name(__BPF_REG(x, __BPF_DECL_REGS, __BPF_N, __VA_ARGS__)) \ | |
469 | { \ | |
470 | return ____##name(__BPF_MAP(x,__BPF_CAST,__BPF_N,__VA_ARGS__));\ | |
471 | } \ | |
472 | static __always_inline \ | |
473 | u64 ____##name(__BPF_MAP(x, __BPF_DECL_ARGS, __BPF_V, __VA_ARGS__)) | |
474 | ||
475 | #define BPF_CALL_0(name, ...) BPF_CALL_x(0, name, __VA_ARGS__) | |
476 | #define BPF_CALL_1(name, ...) BPF_CALL_x(1, name, __VA_ARGS__) | |
477 | #define BPF_CALL_2(name, ...) BPF_CALL_x(2, name, __VA_ARGS__) | |
478 | #define BPF_CALL_3(name, ...) BPF_CALL_x(3, name, __VA_ARGS__) | |
479 | #define BPF_CALL_4(name, ...) BPF_CALL_x(4, name, __VA_ARGS__) | |
480 | #define BPF_CALL_5(name, ...) BPF_CALL_x(5, name, __VA_ARGS__) | |
481 | ||
f96da094 DB |
482 | #define bpf_ctx_range(TYPE, MEMBER) \ |
483 | offsetof(TYPE, MEMBER) ... offsetofend(TYPE, MEMBER) - 1 | |
484 | #define bpf_ctx_range_till(TYPE, MEMBER1, MEMBER2) \ | |
485 | offsetof(TYPE, MEMBER1) ... offsetofend(TYPE, MEMBER2) - 1 | |
b7df9ada DB |
486 | #if BITS_PER_LONG == 64 |
487 | # define bpf_ctx_range_ptr(TYPE, MEMBER) \ | |
488 | offsetof(TYPE, MEMBER) ... offsetofend(TYPE, MEMBER) - 1 | |
489 | #else | |
490 | # define bpf_ctx_range_ptr(TYPE, MEMBER) \ | |
491 | offsetof(TYPE, MEMBER) ... offsetof(TYPE, MEMBER) + 8 - 1 | |
492 | #endif /* BITS_PER_LONG == 64 */ | |
f96da094 DB |
493 | |
494 | #define bpf_target_off(TYPE, MEMBER, SIZE, PTR_SIZE) \ | |
495 | ({ \ | |
496 | BUILD_BUG_ON(FIELD_SIZEOF(TYPE, MEMBER) != (SIZE)); \ | |
497 | *(PTR_SIZE) = (SIZE); \ | |
498 | offsetof(TYPE, MEMBER); \ | |
499 | }) | |
500 | ||
bd4cf0ed AS |
501 | #ifdef CONFIG_COMPAT |
502 | /* A struct sock_filter is architecture independent. */ | |
0c5fe1b4 WD |
503 | struct compat_sock_fprog { |
504 | u16 len; | |
bd4cf0ed | 505 | compat_uptr_t filter; /* struct sock_filter * */ |
0c5fe1b4 WD |
506 | }; |
507 | #endif | |
508 | ||
a3ea269b DB |
509 | struct sock_fprog_kern { |
510 | u16 len; | |
511 | struct sock_filter *filter; | |
512 | }; | |
513 | ||
738cbe72 | 514 | struct bpf_binary_header { |
85782e03 | 515 | u32 pages; |
92624782 ED |
516 | /* Some arches need word alignment for their instructions */ |
517 | u8 image[] __aligned(4); | |
738cbe72 DB |
518 | }; |
519 | ||
7ae457c1 | 520 | struct bpf_prog { |
286aad3c | 521 | u16 pages; /* Number of allocated pages */ |
a91263d5 | 522 | u16 jited:1, /* Is our filter JIT'ed? */ |
60b58afc | 523 | jit_requested:1,/* archs need to JIT the prog */ |
c46646d0 | 524 | gpl_compatible:1, /* Is filter GPL compatible? */ |
ff936a04 | 525 | cb_access:1, /* Is control block accessed? */ |
9802d865 | 526 | dst_needed:1, /* Do we need dst entry? */ |
1c2a088a AS |
527 | blinded:1, /* Was blinded */ |
528 | is_func:1, /* program is a bpf function */ | |
c195651e | 529 | kprobe_override:1, /* Do we override a kprobe? */ |
5cf1e914 | 530 | has_callchain_buf:1, /* callchain buffer allocated? */ |
531 | enforce_expected_attach_type:1; /* Enforce expected_attach_type checking at attach time */ | |
24701ece | 532 | enum bpf_prog_type type; /* Type of BPF program */ |
5e43f899 | 533 | enum bpf_attach_type expected_attach_type; /* For some prog types */ |
7bd509e3 | 534 | u32 len; /* Number of filter blocks */ |
783d28dd | 535 | u32 jited_len; /* Size of jited insns in bytes */ |
f1f7714e | 536 | u8 tag[BPF_TAG_SIZE]; |
09756af4 | 537 | struct bpf_prog_aux *aux; /* Auxiliary fields */ |
24701ece | 538 | struct sock_fprog_kern *orig_prog; /* Original BPF program */ |
88575199 DB |
539 | unsigned int (*bpf_func)(const void *ctx, |
540 | const struct bpf_insn *insn); | |
60a3b225 | 541 | /* Instructions for interpreter */ |
d45ed4a4 | 542 | union { |
bd4cf0ed | 543 | struct sock_filter insns[0]; |
2695fb55 | 544 | struct bpf_insn insnsi[0]; |
d45ed4a4 | 545 | }; |
b715631f SH |
546 | }; |
547 | ||
7ae457c1 | 548 | struct sk_filter { |
4c355cdf | 549 | refcount_t refcnt; |
7ae457c1 AS |
550 | struct rcu_head rcu; |
551 | struct bpf_prog *prog; | |
552 | }; | |
553 | ||
492ecee8 AS |
554 | DECLARE_STATIC_KEY_FALSE(bpf_stats_enabled_key); |
555 | ||
556 | #define BPF_PROG_RUN(prog, ctx) ({ \ | |
557 | u32 ret; \ | |
558 | cant_sleep(); \ | |
559 | if (static_branch_unlikely(&bpf_stats_enabled_key)) { \ | |
560 | struct bpf_prog_stats *stats; \ | |
561 | u64 start = sched_clock(); \ | |
562 | ret = (*(prog)->bpf_func)(ctx, (prog)->insnsi); \ | |
563 | stats = this_cpu_ptr(prog->aux->stats); \ | |
564 | u64_stats_update_begin(&stats->syncp); \ | |
565 | stats->cnt++; \ | |
566 | stats->nsecs += sched_clock() - start; \ | |
567 | u64_stats_update_end(&stats->syncp); \ | |
568 | } else { \ | |
569 | ret = (*(prog)->bpf_func)(ctx, (prog)->insnsi); \ | |
570 | } \ | |
571 | ret; }) | |
7ae457c1 | 572 | |
01dd194c DB |
573 | #define BPF_SKB_CB_LEN QDISC_CB_PRIV_LEN |
574 | ||
db58ba45 AS |
575 | struct bpf_skb_data_end { |
576 | struct qdisc_skb_cb qdisc_cb; | |
de8f3a83 | 577 | void *data_meta; |
db58ba45 AS |
578 | void *data_end; |
579 | }; | |
580 | ||
0b19cc0a | 581 | struct bpf_redirect_info { |
0b19cc0a | 582 | u32 flags; |
4b55cf29 | 583 | u32 tgt_index; |
43e74c02 | 584 | void *tgt_value; |
0b19cc0a TM |
585 | struct bpf_map *map; |
586 | struct bpf_map *map_to_flush; | |
2539650f | 587 | u32 kern_flags; |
0b19cc0a TM |
588 | }; |
589 | ||
590 | DECLARE_PER_CPU(struct bpf_redirect_info, bpf_redirect_info); | |
591 | ||
2539650f TM |
592 | /* flags for bpf_redirect_info kern_flags */ |
593 | #define BPF_RI_F_RF_NO_DIRECT BIT(0) /* no napi_direct on return_frame */ | |
594 | ||
6aaae2b6 DB |
595 | /* Compute the linear packet data range [data, data_end) which |
596 | * will be accessed by various program types (cls_bpf, act_bpf, | |
597 | * lwt, ...). Subsystems allowing direct data access must (!) | |
598 | * ensure that cb[] area can be written to when BPF program is | |
599 | * invoked (otherwise cb[] save/restore is necessary). | |
db58ba45 | 600 | */ |
6aaae2b6 | 601 | static inline void bpf_compute_data_pointers(struct sk_buff *skb) |
db58ba45 AS |
602 | { |
603 | struct bpf_skb_data_end *cb = (struct bpf_skb_data_end *)skb->cb; | |
604 | ||
605 | BUILD_BUG_ON(sizeof(*cb) > FIELD_SIZEOF(struct sk_buff, cb)); | |
de8f3a83 DB |
606 | cb->data_meta = skb->data - skb_metadata_len(skb); |
607 | cb->data_end = skb->data + skb_headlen(skb); | |
db58ba45 AS |
608 | } |
609 | ||
b39b5f41 SL |
610 | /* Similar to bpf_compute_data_pointers(), except that save orginal |
611 | * data in cb->data and cb->meta_data for restore. | |
612 | */ | |
613 | static inline void bpf_compute_and_save_data_end( | |
614 | struct sk_buff *skb, void **saved_data_end) | |
615 | { | |
616 | struct bpf_skb_data_end *cb = (struct bpf_skb_data_end *)skb->cb; | |
617 | ||
618 | *saved_data_end = cb->data_end; | |
619 | cb->data_end = skb->data + skb_headlen(skb); | |
620 | } | |
621 | ||
622 | /* Restore data saved by bpf_compute_data_pointers(). */ | |
623 | static inline void bpf_restore_data_end( | |
624 | struct sk_buff *skb, void *saved_data_end) | |
625 | { | |
626 | struct bpf_skb_data_end *cb = (struct bpf_skb_data_end *)skb->cb; | |
627 | ||
628 | cb->data_end = saved_data_end; | |
629 | } | |
630 | ||
01dd194c DB |
631 | static inline u8 *bpf_skb_cb(struct sk_buff *skb) |
632 | { | |
633 | /* eBPF programs may read/write skb->cb[] area to transfer meta | |
634 | * data between tail calls. Since this also needs to work with | |
635 | * tc, that scratch memory is mapped to qdisc_skb_cb's data area. | |
636 | * | |
637 | * In some socket filter cases, the cb unfortunately needs to be | |
638 | * saved/restored so that protocol specific skb->cb[] data won't | |
639 | * be lost. In any case, due to unpriviledged eBPF programs | |
640 | * attached to sockets, we need to clear the bpf_skb_cb() area | |
641 | * to not leak previous contents to user space. | |
642 | */ | |
643 | BUILD_BUG_ON(FIELD_SIZEOF(struct __sk_buff, cb) != BPF_SKB_CB_LEN); | |
644 | BUILD_BUG_ON(FIELD_SIZEOF(struct __sk_buff, cb) != | |
645 | FIELD_SIZEOF(struct qdisc_skb_cb, data)); | |
646 | ||
647 | return qdisc_skb_cb(skb)->data; | |
648 | } | |
649 | ||
6cab5e90 AS |
650 | static inline u32 __bpf_prog_run_save_cb(const struct bpf_prog *prog, |
651 | struct sk_buff *skb) | |
ff936a04 | 652 | { |
01dd194c DB |
653 | u8 *cb_data = bpf_skb_cb(skb); |
654 | u8 cb_saved[BPF_SKB_CB_LEN]; | |
ff936a04 AS |
655 | u32 res; |
656 | ||
ff936a04 | 657 | if (unlikely(prog->cb_access)) { |
01dd194c DB |
658 | memcpy(cb_saved, cb_data, sizeof(cb_saved)); |
659 | memset(cb_data, 0, sizeof(cb_saved)); | |
ff936a04 AS |
660 | } |
661 | ||
662 | res = BPF_PROG_RUN(prog, skb); | |
663 | ||
664 | if (unlikely(prog->cb_access)) | |
01dd194c | 665 | memcpy(cb_data, cb_saved, sizeof(cb_saved)); |
ff936a04 AS |
666 | |
667 | return res; | |
668 | } | |
669 | ||
6cab5e90 AS |
670 | static inline u32 bpf_prog_run_save_cb(const struct bpf_prog *prog, |
671 | struct sk_buff *skb) | |
672 | { | |
673 | u32 res; | |
674 | ||
675 | preempt_disable(); | |
676 | res = __bpf_prog_run_save_cb(prog, skb); | |
677 | preempt_enable(); | |
678 | return res; | |
679 | } | |
680 | ||
ff936a04 AS |
681 | static inline u32 bpf_prog_run_clear_cb(const struct bpf_prog *prog, |
682 | struct sk_buff *skb) | |
683 | { | |
01dd194c | 684 | u8 *cb_data = bpf_skb_cb(skb); |
6cab5e90 | 685 | u32 res; |
ff936a04 AS |
686 | |
687 | if (unlikely(prog->cb_access)) | |
01dd194c DB |
688 | memset(cb_data, 0, BPF_SKB_CB_LEN); |
689 | ||
6cab5e90 AS |
690 | preempt_disable(); |
691 | res = BPF_PROG_RUN(prog, skb); | |
692 | preempt_enable(); | |
693 | return res; | |
ff936a04 AS |
694 | } |
695 | ||
366cbf2f DB |
696 | static __always_inline u32 bpf_prog_run_xdp(const struct bpf_prog *prog, |
697 | struct xdp_buff *xdp) | |
6a773a15 | 698 | { |
366cbf2f DB |
699 | /* Caller needs to hold rcu_read_lock() (!), otherwise program |
700 | * can be released while still running, or map elements could be | |
701 | * freed early while still having concurrent users. XDP fastpath | |
702 | * already takes rcu_read_lock() when fetching the program, so | |
703 | * it's not necessary here anymore. | |
704 | */ | |
705 | return BPF_PROG_RUN(prog, xdp); | |
6a773a15 BB |
706 | } |
707 | ||
aafe6ae9 DB |
708 | static inline u32 bpf_prog_insn_size(const struct bpf_prog *prog) |
709 | { | |
710 | return prog->len * sizeof(struct bpf_insn); | |
711 | } | |
712 | ||
f1f7714e | 713 | static inline u32 bpf_prog_tag_scratch_size(const struct bpf_prog *prog) |
aafe6ae9 DB |
714 | { |
715 | return round_up(bpf_prog_insn_size(prog) + | |
716 | sizeof(__be64) + 1, SHA_MESSAGE_BYTES); | |
717 | } | |
718 | ||
7ae457c1 | 719 | static inline unsigned int bpf_prog_size(unsigned int proglen) |
b715631f | 720 | { |
7ae457c1 AS |
721 | return max(sizeof(struct bpf_prog), |
722 | offsetof(struct bpf_prog, insns[proglen])); | |
b715631f SH |
723 | } |
724 | ||
7b36f929 DB |
725 | static inline bool bpf_prog_was_classic(const struct bpf_prog *prog) |
726 | { | |
727 | /* When classic BPF programs have been loaded and the arch | |
728 | * does not have a classic BPF JIT (anymore), they have been | |
729 | * converted via bpf_migrate_filter() to eBPF and thus always | |
730 | * have an unspec program type. | |
731 | */ | |
732 | return prog->type == BPF_PROG_TYPE_UNSPEC; | |
733 | } | |
734 | ||
bc23105c DB |
735 | static inline u32 bpf_ctx_off_adjust_machine(u32 size) |
736 | { | |
737 | const u32 size_machine = sizeof(unsigned long); | |
738 | ||
739 | if (size > size_machine && size % size_machine == 0) | |
740 | size = size_machine; | |
741 | ||
742 | return size; | |
743 | } | |
744 | ||
bc23105c DB |
745 | static inline bool |
746 | bpf_ctx_narrow_access_ok(u32 off, u32 size, u32 size_default) | |
747 | { | |
46f53a65 | 748 | return size <= size_default && (size & (size - 1)) == 0; |
f96da094 DB |
749 | } |
750 | ||
d9b8aada IL |
751 | static inline u8 |
752 | bpf_ctx_narrow_load_shift(u32 off, u32 size, u32 size_default) | |
753 | { | |
754 | u8 load_off = off & (size_default - 1); | |
755 | ||
756 | #ifdef __LITTLE_ENDIAN | |
757 | return load_off * 8; | |
758 | #else | |
759 | return (size_default - (load_off + size)) * 8; | |
760 | #endif | |
761 | } | |
762 | ||
b4399546 | 763 | #define bpf_ctx_wide_access_ok(off, size, type, field) \ |
600c70ba SF |
764 | (size == sizeof(__u64) && \ |
765 | off >= offsetof(type, field) && \ | |
766 | off + sizeof(__u64) <= offsetofend(type, field) && \ | |
767 | off % sizeof(__u64) == 0) | |
768 | ||
009937e7 | 769 | #define bpf_classic_proglen(fprog) (fprog->len * sizeof(fprog->filter[0])) |
a3ea269b | 770 | |
60a3b225 DB |
771 | static inline void bpf_prog_lock_ro(struct bpf_prog *fp) |
772 | { | |
d53d2f78 | 773 | set_vm_flush_reset_perms(fp); |
85782e03 | 774 | set_memory_ro((unsigned long)fp, fp->pages); |
60a3b225 DB |
775 | } |
776 | ||
9d876e79 DB |
777 | static inline void bpf_jit_binary_lock_ro(struct bpf_binary_header *hdr) |
778 | { | |
d53d2f78 | 779 | set_vm_flush_reset_perms(hdr); |
85782e03 | 780 | set_memory_ro((unsigned long)hdr, hdr->pages); |
f2c65fb3 | 781 | set_memory_x((unsigned long)hdr, hdr->pages); |
9d876e79 DB |
782 | } |
783 | ||
74451e66 DB |
784 | static inline struct bpf_binary_header * |
785 | bpf_jit_binary_hdr(const struct bpf_prog *fp) | |
786 | { | |
787 | unsigned long real_start = (unsigned long)fp->bpf_func; | |
788 | unsigned long addr = real_start & PAGE_MASK; | |
789 | ||
790 | return (void *)addr; | |
791 | } | |
792 | ||
f4979fce WB |
793 | int sk_filter_trim_cap(struct sock *sk, struct sk_buff *skb, unsigned int cap); |
794 | static inline int sk_filter(struct sock *sk, struct sk_buff *skb) | |
795 | { | |
796 | return sk_filter_trim_cap(sk, skb, 1); | |
797 | } | |
bd4cf0ed | 798 | |
d1c55ab5 | 799 | struct bpf_prog *bpf_prog_select_runtime(struct bpf_prog *fp, int *err); |
7ae457c1 | 800 | void bpf_prog_free(struct bpf_prog *fp); |
bd4cf0ed | 801 | |
5e581dad DB |
802 | bool bpf_opcode_in_insntable(u8 code); |
803 | ||
c454a46b MKL |
804 | void bpf_prog_free_linfo(struct bpf_prog *prog); |
805 | void bpf_prog_fill_jited_linfo(struct bpf_prog *prog, | |
806 | const u32 *insn_to_jit_off); | |
807 | int bpf_prog_alloc_jited_linfo(struct bpf_prog *prog); | |
808 | void bpf_prog_free_jited_linfo(struct bpf_prog *prog); | |
809 | void bpf_prog_free_unused_jited_linfo(struct bpf_prog *prog); | |
810 | ||
60a3b225 | 811 | struct bpf_prog *bpf_prog_alloc(unsigned int size, gfp_t gfp_extra_flags); |
492ecee8 | 812 | struct bpf_prog *bpf_prog_alloc_no_stats(unsigned int size, gfp_t gfp_extra_flags); |
60a3b225 DB |
813 | struct bpf_prog *bpf_prog_realloc(struct bpf_prog *fp_old, unsigned int size, |
814 | gfp_t gfp_extra_flags); | |
815 | void __bpf_prog_free(struct bpf_prog *fp); | |
816 | ||
817 | static inline void bpf_prog_unlock_free(struct bpf_prog *fp) | |
818 | { | |
60a3b225 DB |
819 | __bpf_prog_free(fp); |
820 | } | |
821 | ||
ac67eb2c DB |
822 | typedef int (*bpf_aux_classic_check_t)(struct sock_filter *filter, |
823 | unsigned int flen); | |
824 | ||
7ae457c1 | 825 | int bpf_prog_create(struct bpf_prog **pfp, struct sock_fprog_kern *fprog); |
ac67eb2c | 826 | int bpf_prog_create_from_user(struct bpf_prog **pfp, struct sock_fprog *fprog, |
bab18991 | 827 | bpf_aux_classic_check_t trans, bool save_orig); |
7ae457c1 | 828 | void bpf_prog_destroy(struct bpf_prog *fp); |
a3ea269b | 829 | |
fbc907f0 | 830 | int sk_attach_filter(struct sock_fprog *fprog, struct sock *sk); |
89aa0758 | 831 | int sk_attach_bpf(u32 ufd, struct sock *sk); |
538950a1 CG |
832 | int sk_reuseport_attach_filter(struct sock_fprog *fprog, struct sock *sk); |
833 | int sk_reuseport_attach_bpf(u32 ufd, struct sock *sk); | |
8217ca65 | 834 | void sk_reuseport_prog_free(struct bpf_prog *prog); |
fbc907f0 | 835 | int sk_detach_filter(struct sock *sk); |
fbc907f0 DB |
836 | int sk_get_filter(struct sock *sk, struct sock_filter __user *filter, |
837 | unsigned int len); | |
fbc907f0 | 838 | |
278571ba | 839 | bool sk_filter_charge(struct sock *sk, struct sk_filter *fp); |
fbc907f0 | 840 | void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp); |
0a14842f | 841 | |
62258278 | 842 | u64 __bpf_call_base(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); |
1ea47e01 AS |
843 | #define __bpf_call_base_args \ |
844 | ((u64 (*)(u64, u64, u64, u64, u64, const struct bpf_insn *)) \ | |
845 | __bpf_call_base) | |
d1c55ab5 DB |
846 | |
847 | struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog); | |
9383191d | 848 | void bpf_jit_compile(struct bpf_prog *prog); |
a4b1d3c1 | 849 | bool bpf_jit_needs_zext(void); |
17bedab2 | 850 | bool bpf_helper_changes_pkt_data(void *func); |
62258278 | 851 | |
7105e828 DB |
852 | static inline bool bpf_dump_raw_ok(void) |
853 | { | |
854 | /* Reconstruction of call-sites is dependent on kallsyms, | |
855 | * thus make dump the same restriction. | |
856 | */ | |
857 | return kallsyms_show_value() == 1; | |
858 | } | |
859 | ||
c237ee5e DB |
860 | struct bpf_prog *bpf_patch_insn_single(struct bpf_prog *prog, u32 off, |
861 | const struct bpf_insn *patch, u32 len); | |
52875a04 | 862 | int bpf_remove_insns(struct bpf_prog *prog, u32 off, u32 cnt); |
814abfab | 863 | |
f6069b9a DB |
864 | void bpf_clear_redirect_map(struct bpf_map *map); |
865 | ||
2539650f TM |
866 | static inline bool xdp_return_frame_no_direct(void) |
867 | { | |
868 | struct bpf_redirect_info *ri = this_cpu_ptr(&bpf_redirect_info); | |
869 | ||
870 | return ri->kern_flags & BPF_RI_F_RF_NO_DIRECT; | |
871 | } | |
872 | ||
873 | static inline void xdp_set_return_frame_no_direct(void) | |
874 | { | |
875 | struct bpf_redirect_info *ri = this_cpu_ptr(&bpf_redirect_info); | |
876 | ||
877 | ri->kern_flags |= BPF_RI_F_RF_NO_DIRECT; | |
878 | } | |
879 | ||
880 | static inline void xdp_clear_return_frame_no_direct(void) | |
881 | { | |
882 | struct bpf_redirect_info *ri = this_cpu_ptr(&bpf_redirect_info); | |
883 | ||
884 | ri->kern_flags &= ~BPF_RI_F_RF_NO_DIRECT; | |
885 | } | |
886 | ||
d8d7218a TM |
887 | static inline int xdp_ok_fwd_dev(const struct net_device *fwd, |
888 | unsigned int pktlen) | |
6d5fc195 TM |
889 | { |
890 | unsigned int len; | |
891 | ||
892 | if (unlikely(!(fwd->flags & IFF_UP))) | |
893 | return -ENETDOWN; | |
894 | ||
895 | len = fwd->mtu + fwd->hard_header_len + VLAN_HLEN; | |
d8d7218a | 896 | if (pktlen > len) |
6d5fc195 TM |
897 | return -EMSGSIZE; |
898 | ||
899 | return 0; | |
900 | } | |
901 | ||
11393cc9 JF |
902 | /* The pair of xdp_do_redirect and xdp_do_flush_map MUST be called in the |
903 | * same cpu context. Further for best results no more than a single map | |
904 | * for the do_redirect/do_flush pair should be used. This limitation is | |
905 | * because we only track one map and force a flush when the map changes. | |
2ddf71e2 | 906 | * This does not appear to be a real limitation for existing software. |
11393cc9 | 907 | */ |
2facaad6 | 908 | int xdp_do_generic_redirect(struct net_device *dev, struct sk_buff *skb, |
02671e23 | 909 | struct xdp_buff *xdp, struct bpf_prog *prog); |
5acaee0a JF |
910 | int xdp_do_redirect(struct net_device *dev, |
911 | struct xdp_buff *xdp, | |
912 | struct bpf_prog *prog); | |
11393cc9 | 913 | void xdp_do_flush_map(void); |
814abfab | 914 | |
6a773a15 | 915 | void bpf_warn_invalid_xdp_action(u32 act); |
c237ee5e | 916 | |
2dbb9b9e MKL |
917 | #ifdef CONFIG_INET |
918 | struct sock *bpf_run_sk_reuseport(struct sock_reuseport *reuse, struct sock *sk, | |
919 | struct bpf_prog *prog, struct sk_buff *skb, | |
920 | u32 hash); | |
921 | #else | |
922 | static inline struct sock * | |
923 | bpf_run_sk_reuseport(struct sock_reuseport *reuse, struct sock *sk, | |
924 | struct bpf_prog *prog, struct sk_buff *skb, | |
925 | u32 hash) | |
926 | { | |
927 | return NULL; | |
928 | } | |
929 | #endif | |
930 | ||
b954d834 | 931 | #ifdef CONFIG_BPF_JIT |
c94987e4 | 932 | extern int bpf_jit_enable; |
4f3446bb | 933 | extern int bpf_jit_harden; |
74451e66 | 934 | extern int bpf_jit_kallsyms; |
fdadd049 | 935 | extern long bpf_jit_limit; |
c94987e4 | 936 | |
b954d834 DB |
937 | typedef void (*bpf_jit_fill_hole_t)(void *area, unsigned int size); |
938 | ||
939 | struct bpf_binary_header * | |
940 | bpf_jit_binary_alloc(unsigned int proglen, u8 **image_ptr, | |
941 | unsigned int alignment, | |
942 | bpf_jit_fill_hole_t bpf_fill_ill_insns); | |
943 | void bpf_jit_binary_free(struct bpf_binary_header *hdr); | |
116bfa96 VK |
944 | u64 bpf_jit_alloc_exec_limit(void); |
945 | void *bpf_jit_alloc_exec(unsigned long size); | |
946 | void bpf_jit_free_exec(void *addr); | |
b954d834 DB |
947 | void bpf_jit_free(struct bpf_prog *fp); |
948 | ||
e2c95a61 DB |
949 | int bpf_jit_get_func_addr(const struct bpf_prog *prog, |
950 | const struct bpf_insn *insn, bool extra_pass, | |
951 | u64 *func_addr, bool *func_addr_fixed); | |
952 | ||
4f3446bb DB |
953 | struct bpf_prog *bpf_jit_blind_constants(struct bpf_prog *fp); |
954 | void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other); | |
955 | ||
b954d834 DB |
956 | static inline void bpf_jit_dump(unsigned int flen, unsigned int proglen, |
957 | u32 pass, void *image) | |
958 | { | |
b13138ef DB |
959 | pr_err("flen=%u proglen=%u pass=%u image=%pK from=%s pid=%d\n", flen, |
960 | proglen, pass, image, current->comm, task_pid_nr(current)); | |
961 | ||
b954d834 DB |
962 | if (image) |
963 | print_hex_dump(KERN_ERR, "JIT code: ", DUMP_PREFIX_OFFSET, | |
964 | 16, 1, image, proglen, false); | |
965 | } | |
4f3446bb DB |
966 | |
967 | static inline bool bpf_jit_is_ebpf(void) | |
968 | { | |
969 | # ifdef CONFIG_HAVE_EBPF_JIT | |
970 | return true; | |
971 | # else | |
972 | return false; | |
973 | # endif | |
974 | } | |
975 | ||
81ed18ab AS |
976 | static inline bool ebpf_jit_enabled(void) |
977 | { | |
978 | return bpf_jit_enable && bpf_jit_is_ebpf(); | |
979 | } | |
980 | ||
74451e66 DB |
981 | static inline bool bpf_prog_ebpf_jited(const struct bpf_prog *fp) |
982 | { | |
983 | return fp->jited && bpf_jit_is_ebpf(); | |
984 | } | |
985 | ||
60b58afc | 986 | static inline bool bpf_jit_blinding_enabled(struct bpf_prog *prog) |
4f3446bb DB |
987 | { |
988 | /* These are the prerequisites, should someone ever have the | |
989 | * idea to call blinding outside of them, we make sure to | |
990 | * bail out. | |
991 | */ | |
992 | if (!bpf_jit_is_ebpf()) | |
993 | return false; | |
60b58afc | 994 | if (!prog->jit_requested) |
4f3446bb DB |
995 | return false; |
996 | if (!bpf_jit_harden) | |
997 | return false; | |
998 | if (bpf_jit_harden == 1 && capable(CAP_SYS_ADMIN)) | |
999 | return false; | |
1000 | ||
1001 | return true; | |
1002 | } | |
74451e66 DB |
1003 | |
1004 | static inline bool bpf_jit_kallsyms_enabled(void) | |
1005 | { | |
1006 | /* There are a couple of corner cases where kallsyms should | |
1007 | * not be enabled f.e. on hardening. | |
1008 | */ | |
1009 | if (bpf_jit_harden) | |
1010 | return false; | |
1011 | if (!bpf_jit_kallsyms) | |
1012 | return false; | |
1013 | if (bpf_jit_kallsyms == 1) | |
1014 | return true; | |
1015 | ||
1016 | return false; | |
1017 | } | |
1018 | ||
1019 | const char *__bpf_address_lookup(unsigned long addr, unsigned long *size, | |
1020 | unsigned long *off, char *sym); | |
1021 | bool is_bpf_text_address(unsigned long addr); | |
1022 | int bpf_get_kallsym(unsigned int symnum, unsigned long *value, char *type, | |
1023 | char *sym); | |
1024 | ||
1025 | static inline const char * | |
1026 | bpf_address_lookup(unsigned long addr, unsigned long *size, | |
1027 | unsigned long *off, char **modname, char *sym) | |
1028 | { | |
1029 | const char *ret = __bpf_address_lookup(addr, size, off, sym); | |
1030 | ||
1031 | if (ret && modname) | |
1032 | *modname = NULL; | |
1033 | return ret; | |
1034 | } | |
1035 | ||
1036 | void bpf_prog_kallsyms_add(struct bpf_prog *fp); | |
1037 | void bpf_prog_kallsyms_del(struct bpf_prog *fp); | |
6ee52e2a | 1038 | void bpf_get_prog_name(const struct bpf_prog *prog, char *sym); |
74451e66 DB |
1039 | |
1040 | #else /* CONFIG_BPF_JIT */ | |
1041 | ||
81ed18ab AS |
1042 | static inline bool ebpf_jit_enabled(void) |
1043 | { | |
1044 | return false; | |
1045 | } | |
1046 | ||
74451e66 DB |
1047 | static inline bool bpf_prog_ebpf_jited(const struct bpf_prog *fp) |
1048 | { | |
1049 | return false; | |
1050 | } | |
1051 | ||
b954d834 DB |
1052 | static inline void bpf_jit_free(struct bpf_prog *fp) |
1053 | { | |
1054 | bpf_prog_unlock_free(fp); | |
1055 | } | |
74451e66 DB |
1056 | |
1057 | static inline bool bpf_jit_kallsyms_enabled(void) | |
1058 | { | |
1059 | return false; | |
1060 | } | |
1061 | ||
1062 | static inline const char * | |
1063 | __bpf_address_lookup(unsigned long addr, unsigned long *size, | |
1064 | unsigned long *off, char *sym) | |
1065 | { | |
1066 | return NULL; | |
1067 | } | |
1068 | ||
1069 | static inline bool is_bpf_text_address(unsigned long addr) | |
1070 | { | |
1071 | return false; | |
1072 | } | |
1073 | ||
1074 | static inline int bpf_get_kallsym(unsigned int symnum, unsigned long *value, | |
1075 | char *type, char *sym) | |
1076 | { | |
1077 | return -ERANGE; | |
1078 | } | |
1079 | ||
1080 | static inline const char * | |
1081 | bpf_address_lookup(unsigned long addr, unsigned long *size, | |
1082 | unsigned long *off, char **modname, char *sym) | |
1083 | { | |
1084 | return NULL; | |
1085 | } | |
1086 | ||
1087 | static inline void bpf_prog_kallsyms_add(struct bpf_prog *fp) | |
1088 | { | |
1089 | } | |
1090 | ||
1091 | static inline void bpf_prog_kallsyms_del(struct bpf_prog *fp) | |
1092 | { | |
1093 | } | |
6ee52e2a SL |
1094 | |
1095 | static inline void bpf_get_prog_name(const struct bpf_prog *prog, char *sym) | |
1096 | { | |
1097 | sym[0] = '\0'; | |
1098 | } | |
1099 | ||
b954d834 DB |
1100 | #endif /* CONFIG_BPF_JIT */ |
1101 | ||
7d1982b4 DB |
1102 | void bpf_prog_kallsyms_del_subprogs(struct bpf_prog *fp); |
1103 | void bpf_prog_kallsyms_del_all(struct bpf_prog *fp); | |
1104 | ||
34805931 DB |
1105 | #define BPF_ANC BIT(15) |
1106 | ||
55795ef5 RV |
1107 | static inline bool bpf_needs_clear_a(const struct sock_filter *first) |
1108 | { | |
1109 | switch (first->code) { | |
1110 | case BPF_RET | BPF_K: | |
1111 | case BPF_LD | BPF_W | BPF_LEN: | |
1112 | return false; | |
1113 | ||
1114 | case BPF_LD | BPF_W | BPF_ABS: | |
1115 | case BPF_LD | BPF_H | BPF_ABS: | |
1116 | case BPF_LD | BPF_B | BPF_ABS: | |
1117 | if (first->k == SKF_AD_OFF + SKF_AD_ALU_XOR_X) | |
1118 | return true; | |
1119 | return false; | |
1120 | ||
1121 | default: | |
1122 | return true; | |
1123 | } | |
1124 | } | |
1125 | ||
34805931 DB |
1126 | static inline u16 bpf_anc_helper(const struct sock_filter *ftest) |
1127 | { | |
1128 | BUG_ON(ftest->code & BPF_ANC); | |
1129 | ||
1130 | switch (ftest->code) { | |
1131 | case BPF_LD | BPF_W | BPF_ABS: | |
1132 | case BPF_LD | BPF_H | BPF_ABS: | |
1133 | case BPF_LD | BPF_B | BPF_ABS: | |
1134 | #define BPF_ANCILLARY(CODE) case SKF_AD_OFF + SKF_AD_##CODE: \ | |
1135 | return BPF_ANC | SKF_AD_##CODE | |
1136 | switch (ftest->k) { | |
1137 | BPF_ANCILLARY(PROTOCOL); | |
1138 | BPF_ANCILLARY(PKTTYPE); | |
1139 | BPF_ANCILLARY(IFINDEX); | |
1140 | BPF_ANCILLARY(NLATTR); | |
1141 | BPF_ANCILLARY(NLATTR_NEST); | |
1142 | BPF_ANCILLARY(MARK); | |
1143 | BPF_ANCILLARY(QUEUE); | |
1144 | BPF_ANCILLARY(HATYPE); | |
1145 | BPF_ANCILLARY(RXHASH); | |
1146 | BPF_ANCILLARY(CPU); | |
1147 | BPF_ANCILLARY(ALU_XOR_X); | |
1148 | BPF_ANCILLARY(VLAN_TAG); | |
1149 | BPF_ANCILLARY(VLAN_TAG_PRESENT); | |
1150 | BPF_ANCILLARY(PAY_OFFSET); | |
1151 | BPF_ANCILLARY(RANDOM); | |
27cd5452 | 1152 | BPF_ANCILLARY(VLAN_TPID); |
34805931 DB |
1153 | } |
1154 | /* Fallthrough. */ | |
1155 | default: | |
1156 | return ftest->code; | |
1157 | } | |
1158 | } | |
1159 | ||
9f12fbe6 ZSL |
1160 | void *bpf_internal_load_pointer_neg_helper(const struct sk_buff *skb, |
1161 | int k, unsigned int size); | |
1162 | ||
1163 | static inline void *bpf_load_pointer(const struct sk_buff *skb, int k, | |
1164 | unsigned int size, void *buffer) | |
1165 | { | |
1166 | if (k >= 0) | |
1167 | return skb_header_pointer(skb, k, size, buffer); | |
1168 | ||
1169 | return bpf_internal_load_pointer_neg_helper(skb, k, size); | |
1170 | } | |
1171 | ||
ea02f941 MS |
1172 | static inline int bpf_tell_extensions(void) |
1173 | { | |
37692299 | 1174 | return SKF_AD_MAX; |
ea02f941 MS |
1175 | } |
1176 | ||
4fbac77d AI |
1177 | struct bpf_sock_addr_kern { |
1178 | struct sock *sk; | |
1179 | struct sockaddr *uaddr; | |
1180 | /* Temporary "register" to make indirect stores to nested structures | |
1181 | * defined above. We need three registers to make such a store, but | |
1182 | * only two (src and dst) are available at convert_ctx_access time | |
1183 | */ | |
1184 | u64 tmp_reg; | |
1cedee13 | 1185 | void *t_ctx; /* Attach type specific context. */ |
4fbac77d AI |
1186 | }; |
1187 | ||
40304b2a LB |
1188 | struct bpf_sock_ops_kern { |
1189 | struct sock *sk; | |
1190 | u32 op; | |
1191 | union { | |
de525be2 | 1192 | u32 args[4]; |
40304b2a LB |
1193 | u32 reply; |
1194 | u32 replylong[4]; | |
1195 | }; | |
f19397a5 | 1196 | u32 is_fullsock; |
b73042b8 LB |
1197 | u64 temp; /* temp and everything after is not |
1198 | * initialized to 0 before calling | |
1199 | * the BPF program. New fields that | |
1200 | * should be initialized to 0 should | |
1201 | * be inserted before temp. | |
1202 | * temp is scratch storage used by | |
1203 | * sock_ops_convert_ctx_access | |
1204 | * as temporary storage of a register. | |
1205 | */ | |
40304b2a LB |
1206 | }; |
1207 | ||
7b146ceb AI |
1208 | struct bpf_sysctl_kern { |
1209 | struct ctl_table_header *head; | |
1210 | struct ctl_table *table; | |
1d11b301 AI |
1211 | void *cur_val; |
1212 | size_t cur_len; | |
4e63acdf AI |
1213 | void *new_val; |
1214 | size_t new_len; | |
1215 | int new_updated; | |
7b146ceb | 1216 | int write; |
e1550bfe AI |
1217 | loff_t *ppos; |
1218 | /* Temporary "register" for indirect stores to ppos. */ | |
1219 | u64 tmp_reg; | |
7b146ceb AI |
1220 | }; |
1221 | ||
0d01da6a SF |
1222 | struct bpf_sockopt_kern { |
1223 | struct sock *sk; | |
1224 | u8 *optval; | |
1225 | u8 *optval_end; | |
1226 | s32 level; | |
1227 | s32 optname; | |
1228 | s32 optlen; | |
1229 | s32 retval; | |
1230 | }; | |
1231 | ||
1da177e4 | 1232 | #endif /* __LINUX_FILTER_H__ */ |