| 1 | /* SPDX-License-Identifier: GPL-2.0 */ | 
|---|
| 2 | /* | 
|---|
| 3 | * Linux Socket Filter Data Structures | 
|---|
| 4 | */ | 
|---|
| 5 | #ifndef __LINUX_FILTER_H__ | 
|---|
| 6 | #define __LINUX_FILTER_H__ | 
|---|
| 7 |  | 
|---|
| 8 | #include <linux/atomic.h> | 
|---|
| 9 | #include <linux/bpf.h> | 
|---|
| 10 | #include <linux/refcount.h> | 
|---|
| 11 | #include <linux/compat.h> | 
|---|
| 12 | #include <linux/skbuff.h> | 
|---|
| 13 | #include <linux/linkage.h> | 
|---|
| 14 | #include <linux/printk.h> | 
|---|
| 15 | #include <linux/workqueue.h> | 
|---|
| 16 | #include <linux/sched.h> | 
|---|
| 17 | #include <linux/sched/clock.h> | 
|---|
| 18 | #include <linux/capability.h> | 
|---|
| 19 | #include <linux/set_memory.h> | 
|---|
| 20 | #include <linux/kallsyms.h> | 
|---|
| 21 | #include <linux/if_vlan.h> | 
|---|
| 22 | #include <linux/vmalloc.h> | 
|---|
| 23 | #include <linux/sockptr.h> | 
|---|
| 24 | #include <crypto/sha1.h> | 
|---|
| 25 | #include <linux/u64_stats_sync.h> | 
|---|
| 26 |  | 
|---|
| 27 | #include <net/sch_generic.h> | 
|---|
| 28 |  | 
|---|
| 29 | #include <asm/byteorder.h> | 
|---|
| 30 | #include <uapi/linux/filter.h> | 
|---|
| 31 |  | 
|---|
| 32 | struct sk_buff; | 
|---|
| 33 | struct sock; | 
|---|
| 34 | struct seccomp_data; | 
|---|
| 35 | struct bpf_prog_aux; | 
|---|
| 36 | struct xdp_rxq_info; | 
|---|
| 37 | struct xdp_buff; | 
|---|
| 38 | struct sock_reuseport; | 
|---|
| 39 | struct ctl_table; | 
|---|
| 40 | struct ; | 
|---|
| 41 |  | 
|---|
| 42 | /* ArgX, context and stack frame pointer register positions. Note, | 
|---|
| 43 | * Arg1, Arg2, Arg3, etc are used as argument mappings of function | 
|---|
| 44 | * calls in BPF_CALL instruction. | 
|---|
| 45 | */ | 
|---|
| 46 | #define BPF_REG_ARG1	BPF_REG_1 | 
|---|
| 47 | #define BPF_REG_ARG2	BPF_REG_2 | 
|---|
| 48 | #define BPF_REG_ARG3	BPF_REG_3 | 
|---|
| 49 | #define BPF_REG_ARG4	BPF_REG_4 | 
|---|
| 50 | #define BPF_REG_ARG5	BPF_REG_5 | 
|---|
| 51 | #define BPF_REG_CTX	BPF_REG_6 | 
|---|
| 52 | #define BPF_REG_FP	BPF_REG_10 | 
|---|
| 53 |  | 
|---|
| 54 | /* Additional register mappings for converted user programs. */ | 
|---|
| 55 | #define BPF_REG_A	BPF_REG_0 | 
|---|
| 56 | #define BPF_REG_X	BPF_REG_7 | 
|---|
| 57 | #define BPF_REG_TMP	BPF_REG_2	/* scratch reg */ | 
|---|
| 58 | #define BPF_REG_D	BPF_REG_8	/* data, callee-saved */ | 
|---|
| 59 | #define BPF_REG_H	BPF_REG_9	/* hlen, callee-saved */ | 
|---|
| 60 |  | 
|---|
| 61 | /* Kernel hidden auxiliary/helper register. */ | 
|---|
| 62 | #define BPF_REG_AX		MAX_BPF_REG | 
|---|
| 63 | #define MAX_BPF_EXT_REG		(MAX_BPF_REG + 1) | 
|---|
| 64 | #define MAX_BPF_JIT_REG		MAX_BPF_EXT_REG | 
|---|
| 65 |  | 
|---|
| 66 | /* unused opcode to mark special call to bpf_tail_call() helper */ | 
|---|
| 67 | #define BPF_TAIL_CALL	0xf0 | 
|---|
| 68 |  | 
|---|
| 69 | /* unused opcode to mark special load instruction. Same as BPF_ABS */ | 
|---|
| 70 | #define BPF_PROBE_MEM	0x20 | 
|---|
| 71 |  | 
|---|
| 72 | /* unused opcode to mark special ldsx instruction. Same as BPF_IND */ | 
|---|
| 73 | #define BPF_PROBE_MEMSX	0x40 | 
|---|
| 74 |  | 
|---|
| 75 | /* unused opcode to mark special load instruction. Same as BPF_MSH */ | 
|---|
| 76 | #define BPF_PROBE_MEM32	0xa0 | 
|---|
| 77 |  | 
|---|
| 78 | /* unused opcode to mark special atomic instruction */ | 
|---|
| 79 | #define BPF_PROBE_ATOMIC 0xe0 | 
|---|
| 80 |  | 
|---|
| 81 | /* unused opcode to mark special ldsx instruction. Same as BPF_NOSPEC */ | 
|---|
| 82 | #define BPF_PROBE_MEM32SX 0xc0 | 
|---|
| 83 |  | 
|---|
| 84 | /* unused opcode to mark call to interpreter with arguments */ | 
|---|
| 85 | #define BPF_CALL_ARGS	0xe0 | 
|---|
| 86 |  | 
|---|
| 87 | /* unused opcode to mark speculation barrier for mitigating | 
|---|
| 88 | * Spectre v1 and v4 | 
|---|
| 89 | */ | 
|---|
| 90 | #define BPF_NOSPEC	0xc0 | 
|---|
| 91 |  | 
|---|
| 92 | /* As per nm, we expose JITed images as text (code) section for | 
|---|
| 93 | * kallsyms. That way, tools like perf can find it to match | 
|---|
| 94 | * addresses. | 
|---|
| 95 | */ | 
|---|
| 96 | #define BPF_SYM_ELF_TYPE	't' | 
|---|
| 97 |  | 
|---|
| 98 | /* BPF program can access up to 512 bytes of stack space. */ | 
|---|
| 99 | #define MAX_BPF_STACK	512 | 
|---|
| 100 |  | 
|---|
| 101 | /* Helper macros for filter block array initializers. */ | 
|---|
| 102 |  | 
|---|
| 103 | /* ALU ops on registers, bpf_add|sub|...: dst_reg += src_reg */ | 
|---|
| 104 |  | 
|---|
| 105 | #define BPF_ALU64_REG_OFF(OP, DST, SRC, OFF)			\ | 
|---|
| 106 | ((struct bpf_insn) {					\ | 
|---|
| 107 | .code  = BPF_ALU64 | BPF_OP(OP) | BPF_X,	\ | 
|---|
| 108 | .dst_reg = DST,					\ | 
|---|
| 109 | .src_reg = SRC,					\ | 
|---|
| 110 | .off   = OFF,					\ | 
|---|
| 111 | .imm   = 0 }) | 
|---|
| 112 |  | 
|---|
| 113 | #define BPF_ALU64_REG(OP, DST, SRC)				\ | 
|---|
| 114 | BPF_ALU64_REG_OFF(OP, DST, SRC, 0) | 
|---|
| 115 |  | 
|---|
| 116 | #define BPF_ALU32_REG_OFF(OP, DST, SRC, OFF)			\ | 
|---|
| 117 | ((struct bpf_insn) {					\ | 
|---|
| 118 | .code  = BPF_ALU | BPF_OP(OP) | BPF_X,		\ | 
|---|
| 119 | .dst_reg = DST,					\ | 
|---|
| 120 | .src_reg = SRC,					\ | 
|---|
| 121 | .off   = OFF,					\ | 
|---|
| 122 | .imm   = 0 }) | 
|---|
| 123 |  | 
|---|
| 124 | #define BPF_ALU32_REG(OP, DST, SRC)				\ | 
|---|
| 125 | BPF_ALU32_REG_OFF(OP, DST, SRC, 0) | 
|---|
| 126 |  | 
|---|
| 127 | /* ALU ops on immediates, bpf_add|sub|...: dst_reg += imm32 */ | 
|---|
| 128 |  | 
|---|
| 129 | #define BPF_ALU64_IMM_OFF(OP, DST, IMM, OFF)			\ | 
|---|
| 130 | ((struct bpf_insn) {					\ | 
|---|
| 131 | .code  = BPF_ALU64 | BPF_OP(OP) | BPF_K,	\ | 
|---|
| 132 | .dst_reg = DST,					\ | 
|---|
| 133 | .src_reg = 0,					\ | 
|---|
| 134 | .off   = OFF,					\ | 
|---|
| 135 | .imm   = IMM }) | 
|---|
| 136 | #define BPF_ALU64_IMM(OP, DST, IMM)				\ | 
|---|
| 137 | BPF_ALU64_IMM_OFF(OP, DST, IMM, 0) | 
|---|
| 138 |  | 
|---|
| 139 | #define BPF_ALU32_IMM_OFF(OP, DST, IMM, OFF)			\ | 
|---|
| 140 | ((struct bpf_insn) {					\ | 
|---|
| 141 | .code  = BPF_ALU | BPF_OP(OP) | BPF_K,		\ | 
|---|
| 142 | .dst_reg = DST,					\ | 
|---|
| 143 | .src_reg = 0,					\ | 
|---|
| 144 | .off   = OFF,					\ | 
|---|
| 145 | .imm   = IMM }) | 
|---|
| 146 | #define BPF_ALU32_IMM(OP, DST, IMM)				\ | 
|---|
| 147 | BPF_ALU32_IMM_OFF(OP, DST, IMM, 0) | 
|---|
| 148 |  | 
|---|
| 149 | /* Endianess conversion, cpu_to_{l,b}e(), {l,b}e_to_cpu() */ | 
|---|
| 150 |  | 
|---|
| 151 | #define BPF_ENDIAN(TYPE, DST, LEN)				\ | 
|---|
| 152 | ((struct bpf_insn) {					\ | 
|---|
| 153 | .code  = BPF_ALU | BPF_END | BPF_SRC(TYPE),	\ | 
|---|
| 154 | .dst_reg = DST,					\ | 
|---|
| 155 | .src_reg = 0,					\ | 
|---|
| 156 | .off   = 0,					\ | 
|---|
| 157 | .imm   = LEN }) | 
|---|
| 158 |  | 
|---|
| 159 | /* Byte Swap, bswap16/32/64 */ | 
|---|
| 160 |  | 
|---|
| 161 | #define BPF_BSWAP(DST, LEN)					\ | 
|---|
| 162 | ((struct bpf_insn) {					\ | 
|---|
| 163 | .code  = BPF_ALU64 | BPF_END | BPF_SRC(BPF_TO_LE),	\ | 
|---|
| 164 | .dst_reg = DST,					\ | 
|---|
| 165 | .src_reg = 0,					\ | 
|---|
| 166 | .off   = 0,					\ | 
|---|
| 167 | .imm   = LEN }) | 
|---|
| 168 |  | 
|---|
| 169 | /* Short form of mov, dst_reg = src_reg */ | 
|---|
| 170 |  | 
|---|
| 171 | #define BPF_MOV64_REG(DST, SRC)					\ | 
|---|
| 172 | ((struct bpf_insn) {					\ | 
|---|
| 173 | .code  = BPF_ALU64 | BPF_MOV | BPF_X,		\ | 
|---|
| 174 | .dst_reg = DST,					\ | 
|---|
| 175 | .src_reg = SRC,					\ | 
|---|
| 176 | .off   = 0,					\ | 
|---|
| 177 | .imm   = 0 }) | 
|---|
| 178 |  | 
|---|
| 179 | #define BPF_MOV32_REG(DST, SRC)					\ | 
|---|
| 180 | ((struct bpf_insn) {					\ | 
|---|
| 181 | .code  = BPF_ALU | BPF_MOV | BPF_X,		\ | 
|---|
| 182 | .dst_reg = DST,					\ | 
|---|
| 183 | .src_reg = SRC,					\ | 
|---|
| 184 | .off   = 0,					\ | 
|---|
| 185 | .imm   = 0 }) | 
|---|
| 186 |  | 
|---|
| 187 | /* Special (internal-only) form of mov, used to resolve per-CPU addrs: | 
|---|
| 188 | * dst_reg = src_reg + <percpu_base_off> | 
|---|
| 189 | * BPF_ADDR_PERCPU is used as a special insn->off value. | 
|---|
| 190 | */ | 
|---|
| 191 | #define BPF_ADDR_PERCPU	(-1) | 
|---|
| 192 |  | 
|---|
| 193 | #define BPF_MOV64_PERCPU_REG(DST, SRC)				\ | 
|---|
| 194 | ((struct bpf_insn) {					\ | 
|---|
| 195 | .code  = BPF_ALU64 | BPF_MOV | BPF_X,		\ | 
|---|
| 196 | .dst_reg = DST,					\ | 
|---|
| 197 | .src_reg = SRC,					\ | 
|---|
| 198 | .off   = BPF_ADDR_PERCPU,			\ | 
|---|
| 199 | .imm   = 0 }) | 
|---|
| 200 |  | 
|---|
| 201 | static inline bool insn_is_mov_percpu_addr(const struct bpf_insn *insn) | 
|---|
| 202 | { | 
|---|
| 203 | return insn->code == (BPF_ALU64 | BPF_MOV | BPF_X) && insn->off == BPF_ADDR_PERCPU; | 
|---|
| 204 | } | 
|---|
| 205 |  | 
|---|
| 206 | /* Short form of mov, dst_reg = imm32 */ | 
|---|
| 207 |  | 
|---|
| 208 | #define BPF_MOV64_IMM(DST, IMM)					\ | 
|---|
| 209 | ((struct bpf_insn) {					\ | 
|---|
| 210 | .code  = BPF_ALU64 | BPF_MOV | BPF_K,		\ | 
|---|
| 211 | .dst_reg = DST,					\ | 
|---|
| 212 | .src_reg = 0,					\ | 
|---|
| 213 | .off   = 0,					\ | 
|---|
| 214 | .imm   = IMM }) | 
|---|
| 215 |  | 
|---|
| 216 | #define BPF_MOV32_IMM(DST, IMM)					\ | 
|---|
| 217 | ((struct bpf_insn) {					\ | 
|---|
| 218 | .code  = BPF_ALU | BPF_MOV | BPF_K,		\ | 
|---|
| 219 | .dst_reg = DST,					\ | 
|---|
| 220 | .src_reg = 0,					\ | 
|---|
| 221 | .off   = 0,					\ | 
|---|
| 222 | .imm   = IMM }) | 
|---|
| 223 |  | 
|---|
| 224 | /* Short form of movsx, dst_reg = (s8,s16,s32)src_reg */ | 
|---|
| 225 |  | 
|---|
| 226 | #define BPF_MOVSX64_REG(DST, SRC, OFF)				\ | 
|---|
| 227 | ((struct bpf_insn) {					\ | 
|---|
| 228 | .code  = BPF_ALU64 | BPF_MOV | BPF_X,		\ | 
|---|
| 229 | .dst_reg = DST,					\ | 
|---|
| 230 | .src_reg = SRC,					\ | 
|---|
| 231 | .off   = OFF,					\ | 
|---|
| 232 | .imm   = 0 }) | 
|---|
| 233 |  | 
|---|
| 234 | #define BPF_MOVSX32_REG(DST, SRC, OFF)				\ | 
|---|
| 235 | ((struct bpf_insn) {					\ | 
|---|
| 236 | .code  = BPF_ALU | BPF_MOV | BPF_X,		\ | 
|---|
| 237 | .dst_reg = DST,					\ | 
|---|
| 238 | .src_reg = SRC,					\ | 
|---|
| 239 | .off   = OFF,					\ | 
|---|
| 240 | .imm   = 0 }) | 
|---|
| 241 |  | 
|---|
| 242 | /* Special form of mov32, used for doing explicit zero extension on dst. */ | 
|---|
| 243 | #define BPF_ZEXT_REG(DST)					\ | 
|---|
| 244 | ((struct bpf_insn) {					\ | 
|---|
| 245 | .code  = BPF_ALU | BPF_MOV | BPF_X,		\ | 
|---|
| 246 | .dst_reg = DST,					\ | 
|---|
| 247 | .src_reg = DST,					\ | 
|---|
| 248 | .off   = 0,					\ | 
|---|
| 249 | .imm   = 1 }) | 
|---|
| 250 |  | 
|---|
| 251 | static inline bool insn_is_zext(const struct bpf_insn *insn) | 
|---|
| 252 | { | 
|---|
| 253 | return insn->code == (BPF_ALU | BPF_MOV | BPF_X) && insn->imm == 1; | 
|---|
| 254 | } | 
|---|
| 255 |  | 
|---|
| 256 | /* addr_space_cast from as(0) to as(1) is for converting bpf arena pointers | 
|---|
| 257 | * to pointers in user vma. | 
|---|
| 258 | */ | 
|---|
| 259 | static inline bool insn_is_cast_user(const struct bpf_insn *insn) | 
|---|
| 260 | { | 
|---|
| 261 | return insn->code == (BPF_ALU64 | BPF_MOV | BPF_X) && | 
|---|
| 262 | insn->off == BPF_ADDR_SPACE_CAST && | 
|---|
| 263 | insn->imm == 1U << 16; | 
|---|
| 264 | } | 
|---|
| 265 |  | 
|---|
| 266 | /* BPF_LD_IMM64 macro encodes single 'load 64-bit immediate' insn */ | 
|---|
| 267 | #define BPF_LD_IMM64(DST, IMM)					\ | 
|---|
| 268 | BPF_LD_IMM64_RAW(DST, 0, IMM) | 
|---|
| 269 |  | 
|---|
| 270 | #define BPF_LD_IMM64_RAW(DST, SRC, IMM)				\ | 
|---|
| 271 | ((struct bpf_insn) {					\ | 
|---|
| 272 | .code  = BPF_LD | BPF_DW | BPF_IMM,		\ | 
|---|
| 273 | .dst_reg = DST,					\ | 
|---|
| 274 | .src_reg = SRC,					\ | 
|---|
| 275 | .off   = 0,					\ | 
|---|
| 276 | .imm   = (__u32) (IMM) }),			\ | 
|---|
| 277 | ((struct bpf_insn) {					\ | 
|---|
| 278 | .code  = 0, /* zero is reserved opcode */	\ | 
|---|
| 279 | .dst_reg = 0,					\ | 
|---|
| 280 | .src_reg = 0,					\ | 
|---|
| 281 | .off   = 0,					\ | 
|---|
| 282 | .imm   = ((__u64) (IMM)) >> 32 }) | 
|---|
| 283 |  | 
|---|
| 284 | /* pseudo BPF_LD_IMM64 insn used to refer to process-local map_fd */ | 
|---|
| 285 | #define BPF_LD_MAP_FD(DST, MAP_FD)				\ | 
|---|
| 286 | BPF_LD_IMM64_RAW(DST, BPF_PSEUDO_MAP_FD, MAP_FD) | 
|---|
| 287 |  | 
|---|
| 288 | /* Short form of mov based on type, BPF_X: dst_reg = src_reg, BPF_K: dst_reg = imm32 */ | 
|---|
| 289 |  | 
|---|
| 290 | #define BPF_MOV64_RAW(TYPE, DST, SRC, IMM)			\ | 
|---|
| 291 | ((struct bpf_insn) {					\ | 
|---|
| 292 | .code  = BPF_ALU64 | BPF_MOV | BPF_SRC(TYPE),	\ | 
|---|
| 293 | .dst_reg = DST,					\ | 
|---|
| 294 | .src_reg = SRC,					\ | 
|---|
| 295 | .off   = 0,					\ | 
|---|
| 296 | .imm   = IMM }) | 
|---|
| 297 |  | 
|---|
| 298 | #define BPF_MOV32_RAW(TYPE, DST, SRC, IMM)			\ | 
|---|
| 299 | ((struct bpf_insn) {					\ | 
|---|
| 300 | .code  = BPF_ALU | BPF_MOV | BPF_SRC(TYPE),	\ | 
|---|
| 301 | .dst_reg = DST,					\ | 
|---|
| 302 | .src_reg = SRC,					\ | 
|---|
| 303 | .off   = 0,					\ | 
|---|
| 304 | .imm   = IMM }) | 
|---|
| 305 |  | 
|---|
| 306 | /* Direct packet access, R0 = *(uint *) (skb->data + imm32) */ | 
|---|
| 307 |  | 
|---|
| 308 | #define BPF_LD_ABS(SIZE, IMM)					\ | 
|---|
| 309 | ((struct bpf_insn) {					\ | 
|---|
| 310 | .code  = BPF_LD | BPF_SIZE(SIZE) | BPF_ABS,	\ | 
|---|
| 311 | .dst_reg = 0,					\ | 
|---|
| 312 | .src_reg = 0,					\ | 
|---|
| 313 | .off   = 0,					\ | 
|---|
| 314 | .imm   = IMM }) | 
|---|
| 315 |  | 
|---|
| 316 | /* Indirect packet access, R0 = *(uint *) (skb->data + src_reg + imm32) */ | 
|---|
| 317 |  | 
|---|
| 318 | #define BPF_LD_IND(SIZE, SRC, IMM)				\ | 
|---|
| 319 | ((struct bpf_insn) {					\ | 
|---|
| 320 | .code  = BPF_LD | BPF_SIZE(SIZE) | BPF_IND,	\ | 
|---|
| 321 | .dst_reg = 0,					\ | 
|---|
| 322 | .src_reg = SRC,					\ | 
|---|
| 323 | .off   = 0,					\ | 
|---|
| 324 | .imm   = IMM }) | 
|---|
| 325 |  | 
|---|
| 326 | /* Memory load, dst_reg = *(uint *) (src_reg + off16) */ | 
|---|
| 327 |  | 
|---|
| 328 | #define BPF_LDX_MEM(SIZE, DST, SRC, OFF)			\ | 
|---|
| 329 | ((struct bpf_insn) {					\ | 
|---|
| 330 | .code  = BPF_LDX | BPF_SIZE(SIZE) | BPF_MEM,	\ | 
|---|
| 331 | .dst_reg = DST,					\ | 
|---|
| 332 | .src_reg = SRC,					\ | 
|---|
| 333 | .off   = OFF,					\ | 
|---|
| 334 | .imm   = 0 }) | 
|---|
| 335 |  | 
|---|
| 336 | /* Memory load, dst_reg = *(signed size *) (src_reg + off16) */ | 
|---|
| 337 |  | 
|---|
| 338 | #define BPF_LDX_MEMSX(SIZE, DST, SRC, OFF)			\ | 
|---|
| 339 | ((struct bpf_insn) {					\ | 
|---|
| 340 | .code  = BPF_LDX | BPF_SIZE(SIZE) | BPF_MEMSX,	\ | 
|---|
| 341 | .dst_reg = DST,					\ | 
|---|
| 342 | .src_reg = SRC,					\ | 
|---|
| 343 | .off   = OFF,					\ | 
|---|
| 344 | .imm   = 0 }) | 
|---|
| 345 |  | 
|---|
| 346 | /* Memory store, *(uint *) (dst_reg + off16) = src_reg */ | 
|---|
| 347 |  | 
|---|
| 348 | #define BPF_STX_MEM(SIZE, DST, SRC, OFF)			\ | 
|---|
| 349 | ((struct bpf_insn) {					\ | 
|---|
| 350 | .code  = BPF_STX | BPF_SIZE(SIZE) | BPF_MEM,	\ | 
|---|
| 351 | .dst_reg = DST,					\ | 
|---|
| 352 | .src_reg = SRC,					\ | 
|---|
| 353 | .off   = OFF,					\ | 
|---|
| 354 | .imm   = 0 }) | 
|---|
| 355 |  | 
|---|
| 356 |  | 
|---|
| 357 | /* | 
|---|
| 358 | * Atomic operations: | 
|---|
| 359 | * | 
|---|
| 360 | *   BPF_ADD                  *(uint *) (dst_reg + off16) += src_reg | 
|---|
| 361 | *   BPF_AND                  *(uint *) (dst_reg + off16) &= src_reg | 
|---|
| 362 | *   BPF_OR                   *(uint *) (dst_reg + off16) |= src_reg | 
|---|
| 363 | *   BPF_XOR                  *(uint *) (dst_reg + off16) ^= src_reg | 
|---|
| 364 | *   BPF_ADD | BPF_FETCH      src_reg = atomic_fetch_add(dst_reg + off16, src_reg); | 
|---|
| 365 | *   BPF_AND | BPF_FETCH      src_reg = atomic_fetch_and(dst_reg + off16, src_reg); | 
|---|
| 366 | *   BPF_OR | BPF_FETCH       src_reg = atomic_fetch_or(dst_reg + off16, src_reg); | 
|---|
| 367 | *   BPF_XOR | BPF_FETCH      src_reg = atomic_fetch_xor(dst_reg + off16, src_reg); | 
|---|
| 368 | *   BPF_XCHG                 src_reg = atomic_xchg(dst_reg + off16, src_reg) | 
|---|
| 369 | *   BPF_CMPXCHG              r0 = atomic_cmpxchg(dst_reg + off16, r0, src_reg) | 
|---|
| 370 | *   BPF_LOAD_ACQ             dst_reg = smp_load_acquire(src_reg + off16) | 
|---|
| 371 | *   BPF_STORE_REL            smp_store_release(dst_reg + off16, src_reg) | 
|---|
| 372 | */ | 
|---|
| 373 |  | 
|---|
| 374 | #define BPF_ATOMIC_OP(SIZE, OP, DST, SRC, OFF)			\ | 
|---|
| 375 | ((struct bpf_insn) {					\ | 
|---|
| 376 | .code  = BPF_STX | BPF_SIZE(SIZE) | BPF_ATOMIC,	\ | 
|---|
| 377 | .dst_reg = DST,					\ | 
|---|
| 378 | .src_reg = SRC,					\ | 
|---|
| 379 | .off   = OFF,					\ | 
|---|
| 380 | .imm   = OP }) | 
|---|
| 381 |  | 
|---|
| 382 | /* Legacy alias */ | 
|---|
| 383 | #define BPF_STX_XADD(SIZE, DST, SRC, OFF) BPF_ATOMIC_OP(SIZE, BPF_ADD, DST, SRC, OFF) | 
|---|
| 384 |  | 
|---|
| 385 | /* Memory store, *(uint *) (dst_reg + off16) = imm32 */ | 
|---|
| 386 |  | 
|---|
| 387 | #define BPF_ST_MEM(SIZE, DST, OFF, IMM)				\ | 
|---|
| 388 | ((struct bpf_insn) {					\ | 
|---|
| 389 | .code  = BPF_ST | BPF_SIZE(SIZE) | BPF_MEM,	\ | 
|---|
| 390 | .dst_reg = DST,					\ | 
|---|
| 391 | .src_reg = 0,					\ | 
|---|
| 392 | .off   = OFF,					\ | 
|---|
| 393 | .imm   = IMM }) | 
|---|
| 394 |  | 
|---|
| 395 | /* Conditional jumps against registers, if (dst_reg 'op' src_reg) goto pc + off16 */ | 
|---|
| 396 |  | 
|---|
| 397 | #define BPF_JMP_REG(OP, DST, SRC, OFF)				\ | 
|---|
| 398 | ((struct bpf_insn) {					\ | 
|---|
| 399 | .code  = BPF_JMP | BPF_OP(OP) | BPF_X,		\ | 
|---|
| 400 | .dst_reg = DST,					\ | 
|---|
| 401 | .src_reg = SRC,					\ | 
|---|
| 402 | .off   = OFF,					\ | 
|---|
| 403 | .imm   = 0 }) | 
|---|
| 404 |  | 
|---|
| 405 | /* Conditional jumps against immediates, if (dst_reg 'op' imm32) goto pc + off16 */ | 
|---|
| 406 |  | 
|---|
| 407 | #define BPF_JMP_IMM(OP, DST, IMM, OFF)				\ | 
|---|
| 408 | ((struct bpf_insn) {					\ | 
|---|
| 409 | .code  = BPF_JMP | BPF_OP(OP) | BPF_K,		\ | 
|---|
| 410 | .dst_reg = DST,					\ | 
|---|
| 411 | .src_reg = 0,					\ | 
|---|
| 412 | .off   = OFF,					\ | 
|---|
| 413 | .imm   = IMM }) | 
|---|
| 414 |  | 
|---|
| 415 | /* Like BPF_JMP_REG, but with 32-bit wide operands for comparison. */ | 
|---|
| 416 |  | 
|---|
| 417 | #define BPF_JMP32_REG(OP, DST, SRC, OFF)			\ | 
|---|
| 418 | ((struct bpf_insn) {					\ | 
|---|
| 419 | .code  = BPF_JMP32 | BPF_OP(OP) | BPF_X,	\ | 
|---|
| 420 | .dst_reg = DST,					\ | 
|---|
| 421 | .src_reg = SRC,					\ | 
|---|
| 422 | .off   = OFF,					\ | 
|---|
| 423 | .imm   = 0 }) | 
|---|
| 424 |  | 
|---|
| 425 | /* Like BPF_JMP_IMM, but with 32-bit wide operands for comparison. */ | 
|---|
| 426 |  | 
|---|
| 427 | #define BPF_JMP32_IMM(OP, DST, IMM, OFF)			\ | 
|---|
| 428 | ((struct bpf_insn) {					\ | 
|---|
| 429 | .code  = BPF_JMP32 | BPF_OP(OP) | BPF_K,	\ | 
|---|
| 430 | .dst_reg = DST,					\ | 
|---|
| 431 | .src_reg = 0,					\ | 
|---|
| 432 | .off   = OFF,					\ | 
|---|
| 433 | .imm   = IMM }) | 
|---|
| 434 |  | 
|---|
| 435 | /* Unconditional jumps, goto pc + off16 */ | 
|---|
| 436 |  | 
|---|
| 437 | #define BPF_JMP_A(OFF)						\ | 
|---|
| 438 | ((struct bpf_insn) {					\ | 
|---|
| 439 | .code  = BPF_JMP | BPF_JA,			\ | 
|---|
| 440 | .dst_reg = 0,					\ | 
|---|
| 441 | .src_reg = 0,					\ | 
|---|
| 442 | .off   = OFF,					\ | 
|---|
| 443 | .imm   = 0 }) | 
|---|
| 444 |  | 
|---|
| 445 | /* Unconditional jumps, gotol pc + imm32 */ | 
|---|
| 446 |  | 
|---|
| 447 | #define BPF_JMP32_A(IMM)					\ | 
|---|
| 448 | ((struct bpf_insn) {					\ | 
|---|
| 449 | .code  = BPF_JMP32 | BPF_JA,			\ | 
|---|
| 450 | .dst_reg = 0,					\ | 
|---|
| 451 | .src_reg = 0,					\ | 
|---|
| 452 | .off   = 0,					\ | 
|---|
| 453 | .imm   = IMM }) | 
|---|
| 454 |  | 
|---|
| 455 | /* Relative call */ | 
|---|
| 456 |  | 
|---|
| 457 | #define BPF_CALL_REL(TGT)					\ | 
|---|
| 458 | ((struct bpf_insn) {					\ | 
|---|
| 459 | .code  = BPF_JMP | BPF_CALL,			\ | 
|---|
| 460 | .dst_reg = 0,					\ | 
|---|
| 461 | .src_reg = BPF_PSEUDO_CALL,			\ | 
|---|
| 462 | .off   = 0,					\ | 
|---|
| 463 | .imm   = TGT }) | 
|---|
| 464 |  | 
|---|
| 465 | /* Convert function address to BPF immediate */ | 
|---|
| 466 |  | 
|---|
| 467 | #define BPF_CALL_IMM(x)	((void *)(x) - (void *)__bpf_call_base) | 
|---|
| 468 |  | 
|---|
| 469 | #define BPF_EMIT_CALL(FUNC)					\ | 
|---|
| 470 | ((struct bpf_insn) {					\ | 
|---|
| 471 | .code  = BPF_JMP | BPF_CALL,			\ | 
|---|
| 472 | .dst_reg = 0,					\ | 
|---|
| 473 | .src_reg = 0,					\ | 
|---|
| 474 | .off   = 0,					\ | 
|---|
| 475 | .imm   = BPF_CALL_IMM(FUNC) }) | 
|---|
| 476 |  | 
|---|
| 477 | /* Kfunc call */ | 
|---|
| 478 |  | 
|---|
| 479 | #define BPF_CALL_KFUNC(OFF, IMM)				\ | 
|---|
| 480 | ((struct bpf_insn) {					\ | 
|---|
| 481 | .code  = BPF_JMP | BPF_CALL,			\ | 
|---|
| 482 | .dst_reg = 0,					\ | 
|---|
| 483 | .src_reg = BPF_PSEUDO_KFUNC_CALL,		\ | 
|---|
| 484 | .off   = OFF,					\ | 
|---|
| 485 | .imm   = IMM }) | 
|---|
| 486 |  | 
|---|
| 487 | /* Raw code statement block */ | 
|---|
| 488 |  | 
|---|
| 489 | #define BPF_RAW_INSN(CODE, DST, SRC, OFF, IMM)			\ | 
|---|
| 490 | ((struct bpf_insn) {					\ | 
|---|
| 491 | .code  = CODE,					\ | 
|---|
| 492 | .dst_reg = DST,					\ | 
|---|
| 493 | .src_reg = SRC,					\ | 
|---|
| 494 | .off   = OFF,					\ | 
|---|
| 495 | .imm   = IMM }) | 
|---|
| 496 |  | 
|---|
| 497 | /* Program exit */ | 
|---|
| 498 |  | 
|---|
| 499 | #define BPF_EXIT_INSN()						\ | 
|---|
| 500 | ((struct bpf_insn) {					\ | 
|---|
| 501 | .code  = BPF_JMP | BPF_EXIT,			\ | 
|---|
| 502 | .dst_reg = 0,					\ | 
|---|
| 503 | .src_reg = 0,					\ | 
|---|
| 504 | .off   = 0,					\ | 
|---|
| 505 | .imm   = 0 }) | 
|---|
| 506 |  | 
|---|
| 507 | /* Speculation barrier */ | 
|---|
| 508 |  | 
|---|
| 509 | #define BPF_ST_NOSPEC()						\ | 
|---|
| 510 | ((struct bpf_insn) {					\ | 
|---|
| 511 | .code  = BPF_ST | BPF_NOSPEC,			\ | 
|---|
| 512 | .dst_reg = 0,					\ | 
|---|
| 513 | .src_reg = 0,					\ | 
|---|
| 514 | .off   = 0,					\ | 
|---|
| 515 | .imm   = 0 }) | 
|---|
| 516 |  | 
|---|
| 517 | /* Internal classic blocks for direct assignment */ | 
|---|
| 518 |  | 
|---|
| 519 | #define __BPF_STMT(CODE, K)					\ | 
|---|
| 520 | ((struct sock_filter) BPF_STMT(CODE, K)) | 
|---|
| 521 |  | 
|---|
| 522 | #define __BPF_JUMP(CODE, K, JT, JF)				\ | 
|---|
| 523 | ((struct sock_filter) BPF_JUMP(CODE, K, JT, JF)) | 
|---|
| 524 |  | 
|---|
| 525 | #define bytes_to_bpf_size(bytes)				\ | 
|---|
| 526 | ({								\ | 
|---|
| 527 | int bpf_size = -EINVAL;					\ | 
|---|
| 528 | \ | 
|---|
| 529 | if (bytes == sizeof(u8))				\ | 
|---|
| 530 | bpf_size = BPF_B;				\ | 
|---|
| 531 | else if (bytes == sizeof(u16))				\ | 
|---|
| 532 | bpf_size = BPF_H;				\ | 
|---|
| 533 | else if (bytes == sizeof(u32))				\ | 
|---|
| 534 | bpf_size = BPF_W;				\ | 
|---|
| 535 | else if (bytes == sizeof(u64))				\ | 
|---|
| 536 | bpf_size = BPF_DW;				\ | 
|---|
| 537 | \ | 
|---|
| 538 | bpf_size;						\ | 
|---|
| 539 | }) | 
|---|
| 540 |  | 
|---|
| 541 | #define bpf_size_to_bytes(bpf_size)				\ | 
|---|
| 542 | ({								\ | 
|---|
| 543 | int bytes = -EINVAL;					\ | 
|---|
| 544 | \ | 
|---|
| 545 | if (bpf_size == BPF_B)					\ | 
|---|
| 546 | bytes = sizeof(u8);				\ | 
|---|
| 547 | else if (bpf_size == BPF_H)				\ | 
|---|
| 548 | bytes = sizeof(u16);				\ | 
|---|
| 549 | else if (bpf_size == BPF_W)				\ | 
|---|
| 550 | bytes = sizeof(u32);				\ | 
|---|
| 551 | else if (bpf_size == BPF_DW)				\ | 
|---|
| 552 | bytes = sizeof(u64);				\ | 
|---|
| 553 | \ | 
|---|
| 554 | bytes;							\ | 
|---|
| 555 | }) | 
|---|
| 556 |  | 
|---|
| 557 | #define BPF_SIZEOF(type)					\ | 
|---|
| 558 | ({							\ | 
|---|
| 559 | const int __size = bytes_to_bpf_size(sizeof(type)); \ | 
|---|
| 560 | BUILD_BUG_ON(__size < 0);			\ | 
|---|
| 561 | __size;						\ | 
|---|
| 562 | }) | 
|---|
| 563 |  | 
|---|
| 564 | #define BPF_FIELD_SIZEOF(type, field)				\ | 
|---|
| 565 | ({							\ | 
|---|
| 566 | const int __size = bytes_to_bpf_size(sizeof_field(type, field)); \ | 
|---|
| 567 | BUILD_BUG_ON(__size < 0);			\ | 
|---|
| 568 | __size;						\ | 
|---|
| 569 | }) | 
|---|
| 570 |  | 
|---|
| 571 | #define BPF_LDST_BYTES(insn)					\ | 
|---|
| 572 | ({							\ | 
|---|
| 573 | const int __size = bpf_size_to_bytes(BPF_SIZE((insn)->code)); \ | 
|---|
| 574 | WARN_ON(__size < 0);				\ | 
|---|
| 575 | __size;						\ | 
|---|
| 576 | }) | 
|---|
| 577 |  | 
|---|
| 578 | #define __BPF_MAP_0(m, v, ...) v | 
|---|
| 579 | #define __BPF_MAP_1(m, v, t, a, ...) m(t, a) | 
|---|
| 580 | #define __BPF_MAP_2(m, v, t, a, ...) m(t, a), __BPF_MAP_1(m, v, __VA_ARGS__) | 
|---|
| 581 | #define __BPF_MAP_3(m, v, t, a, ...) m(t, a), __BPF_MAP_2(m, v, __VA_ARGS__) | 
|---|
| 582 | #define __BPF_MAP_4(m, v, t, a, ...) m(t, a), __BPF_MAP_3(m, v, __VA_ARGS__) | 
|---|
| 583 | #define __BPF_MAP_5(m, v, t, a, ...) m(t, a), __BPF_MAP_4(m, v, __VA_ARGS__) | 
|---|
| 584 |  | 
|---|
| 585 | #define __BPF_REG_0(...) __BPF_PAD(5) | 
|---|
| 586 | #define __BPF_REG_1(...) __BPF_MAP(1, __VA_ARGS__), __BPF_PAD(4) | 
|---|
| 587 | #define __BPF_REG_2(...) __BPF_MAP(2, __VA_ARGS__), __BPF_PAD(3) | 
|---|
| 588 | #define __BPF_REG_3(...) __BPF_MAP(3, __VA_ARGS__), __BPF_PAD(2) | 
|---|
| 589 | #define __BPF_REG_4(...) __BPF_MAP(4, __VA_ARGS__), __BPF_PAD(1) | 
|---|
| 590 | #define __BPF_REG_5(...) __BPF_MAP(5, __VA_ARGS__) | 
|---|
| 591 |  | 
|---|
| 592 | #define __BPF_MAP(n, ...) __BPF_MAP_##n(__VA_ARGS__) | 
|---|
| 593 | #define __BPF_REG(n, ...) __BPF_REG_##n(__VA_ARGS__) | 
|---|
| 594 |  | 
|---|
| 595 | #define __BPF_CAST(t, a)						       \ | 
|---|
| 596 | (__force t)							       \ | 
|---|
| 597 | (__force							       \ | 
|---|
| 598 | typeof(__builtin_choose_expr(sizeof(t) == sizeof(unsigned long),      \ | 
|---|
| 599 | (unsigned long)0, (t)0))) a | 
|---|
| 600 | #define __BPF_V void | 
|---|
| 601 | #define __BPF_N | 
|---|
| 602 |  | 
|---|
| 603 | #define __BPF_DECL_ARGS(t, a) t   a | 
|---|
| 604 | #define __BPF_DECL_REGS(t, a) u64 a | 
|---|
| 605 |  | 
|---|
| 606 | #define __BPF_PAD(n)							       \ | 
|---|
| 607 | __BPF_MAP(n, __BPF_DECL_ARGS, __BPF_N, u64, __ur_1, u64, __ur_2,       \ | 
|---|
| 608 | u64, __ur_3, u64, __ur_4, u64, __ur_5) | 
|---|
| 609 |  | 
|---|
| 610 | #define BPF_CALL_x(x, attr, name, ...)					       \ | 
|---|
| 611 | static __always_inline						       \ | 
|---|
| 612 | u64 ____##name(__BPF_MAP(x, __BPF_DECL_ARGS, __BPF_V, __VA_ARGS__));   \ | 
|---|
| 613 | typedef u64 (*btf_##name)(__BPF_MAP(x, __BPF_DECL_ARGS, __BPF_V, __VA_ARGS__)); \ | 
|---|
| 614 | attr u64 name(__BPF_REG(x, __BPF_DECL_REGS, __BPF_N, __VA_ARGS__));    \ | 
|---|
| 615 | attr u64 name(__BPF_REG(x, __BPF_DECL_REGS, __BPF_N, __VA_ARGS__))     \ | 
|---|
| 616 | {								       \ | 
|---|
| 617 | return ((btf_##name)____##name)(__BPF_MAP(x,__BPF_CAST,__BPF_N,__VA_ARGS__));\ | 
|---|
| 618 | }								       \ | 
|---|
| 619 | static __always_inline						       \ | 
|---|
| 620 | u64 ____##name(__BPF_MAP(x, __BPF_DECL_ARGS, __BPF_V, __VA_ARGS__)) | 
|---|
| 621 |  | 
|---|
| 622 | #define __NOATTR | 
|---|
| 623 | #define BPF_CALL_0(name, ...)	BPF_CALL_x(0, __NOATTR, name, __VA_ARGS__) | 
|---|
| 624 | #define BPF_CALL_1(name, ...)	BPF_CALL_x(1, __NOATTR, name, __VA_ARGS__) | 
|---|
| 625 | #define BPF_CALL_2(name, ...)	BPF_CALL_x(2, __NOATTR, name, __VA_ARGS__) | 
|---|
| 626 | #define BPF_CALL_3(name, ...)	BPF_CALL_x(3, __NOATTR, name, __VA_ARGS__) | 
|---|
| 627 | #define BPF_CALL_4(name, ...)	BPF_CALL_x(4, __NOATTR, name, __VA_ARGS__) | 
|---|
| 628 | #define BPF_CALL_5(name, ...)	BPF_CALL_x(5, __NOATTR, name, __VA_ARGS__) | 
|---|
| 629 |  | 
|---|
| 630 | #define NOTRACE_BPF_CALL_1(name, ...)	BPF_CALL_x(1, notrace, name, __VA_ARGS__) | 
|---|
| 631 |  | 
|---|
| 632 | #define bpf_ctx_range(TYPE, MEMBER)						\ | 
|---|
| 633 | offsetof(TYPE, MEMBER) ... offsetofend(TYPE, MEMBER) - 1 | 
|---|
| 634 | #define bpf_ctx_range_till(TYPE, MEMBER1, MEMBER2)				\ | 
|---|
| 635 | offsetof(TYPE, MEMBER1) ... offsetofend(TYPE, MEMBER2) - 1 | 
|---|
| 636 | #if BITS_PER_LONG == 64 | 
|---|
| 637 | # define bpf_ctx_range_ptr(TYPE, MEMBER)					\ | 
|---|
| 638 | offsetof(TYPE, MEMBER) ... offsetofend(TYPE, MEMBER) - 1 | 
|---|
| 639 | #else | 
|---|
| 640 | # define bpf_ctx_range_ptr(TYPE, MEMBER)					\ | 
|---|
| 641 | offsetof(TYPE, MEMBER) ... offsetof(TYPE, MEMBER) + 8 - 1 | 
|---|
| 642 | #endif /* BITS_PER_LONG == 64 */ | 
|---|
| 643 |  | 
|---|
| 644 | #define bpf_target_off(TYPE, MEMBER, SIZE, PTR_SIZE)				\ | 
|---|
| 645 | ({									\ | 
|---|
| 646 | BUILD_BUG_ON(sizeof_field(TYPE, MEMBER) != (SIZE));		\ | 
|---|
| 647 | *(PTR_SIZE) = (SIZE);						\ | 
|---|
| 648 | offsetof(TYPE, MEMBER);						\ | 
|---|
| 649 | }) | 
|---|
| 650 |  | 
|---|
| 651 | /* A struct sock_filter is architecture independent. */ | 
|---|
| 652 | struct compat_sock_fprog { | 
|---|
| 653 | u16		len; | 
|---|
| 654 | compat_uptr_t	filter;	/* struct sock_filter * */ | 
|---|
| 655 | }; | 
|---|
| 656 |  | 
|---|
| 657 | struct sock_fprog_kern { | 
|---|
| 658 | u16			len; | 
|---|
| 659 | struct sock_filter	*filter; | 
|---|
| 660 | }; | 
|---|
| 661 |  | 
|---|
| 662 | /* Some arches need doubleword alignment for their instructions and/or data */ | 
|---|
| 663 | #define BPF_IMAGE_ALIGNMENT 8 | 
|---|
| 664 |  | 
|---|
| 665 | struct  { | 
|---|
| 666 | u32 ; | 
|---|
| 667 | u8 [] __aligned(BPF_IMAGE_ALIGNMENT); | 
|---|
| 668 | }; | 
|---|
| 669 |  | 
|---|
| 670 | struct bpf_prog_stats { | 
|---|
| 671 | u64_stats_t cnt; | 
|---|
| 672 | u64_stats_t nsecs; | 
|---|
| 673 | u64_stats_t misses; | 
|---|
| 674 | struct u64_stats_sync syncp; | 
|---|
| 675 | } __aligned(2 * sizeof(u64)); | 
|---|
| 676 |  | 
|---|
| 677 | struct bpf_timed_may_goto { | 
|---|
| 678 | u64 count; | 
|---|
| 679 | u64 timestamp; | 
|---|
| 680 | }; | 
|---|
| 681 |  | 
|---|
| 682 | struct sk_filter { | 
|---|
| 683 | refcount_t	refcnt; | 
|---|
| 684 | struct rcu_head	rcu; | 
|---|
| 685 | struct bpf_prog	*prog; | 
|---|
| 686 | }; | 
|---|
| 687 |  | 
|---|
| 688 | DECLARE_STATIC_KEY_FALSE(bpf_stats_enabled_key); | 
|---|
| 689 |  | 
|---|
| 690 | extern struct mutex nf_conn_btf_access_lock; | 
|---|
| 691 | extern int (*nfct_btf_struct_access)(struct bpf_verifier_log *log, | 
|---|
| 692 | const struct bpf_reg_state *reg, | 
|---|
| 693 | int off, int size); | 
|---|
| 694 |  | 
|---|
| 695 | typedef unsigned int (*bpf_dispatcher_fn)(const void *ctx, | 
|---|
| 696 | const struct bpf_insn *insnsi, | 
|---|
| 697 | unsigned int (*bpf_func)(const void *, | 
|---|
| 698 | const struct bpf_insn *)); | 
|---|
| 699 |  | 
|---|
| 700 | static __always_inline u32 __bpf_prog_run(const struct bpf_prog *prog, | 
|---|
| 701 | const void *ctx, | 
|---|
| 702 | bpf_dispatcher_fn dfunc) | 
|---|
| 703 | { | 
|---|
| 704 | u32 ret; | 
|---|
| 705 |  | 
|---|
| 706 | cant_migrate(); | 
|---|
| 707 | if (static_branch_unlikely(&bpf_stats_enabled_key)) { | 
|---|
| 708 | struct bpf_prog_stats *stats; | 
|---|
| 709 | u64 duration, start = sched_clock(); | 
|---|
| 710 | unsigned long flags; | 
|---|
| 711 |  | 
|---|
| 712 | ret = dfunc(ctx, prog->insnsi, prog->bpf_func); | 
|---|
| 713 |  | 
|---|
| 714 | duration = sched_clock() - start; | 
|---|
| 715 | stats = this_cpu_ptr(prog->stats); | 
|---|
| 716 | flags = u64_stats_update_begin_irqsave(syncp: &stats->syncp); | 
|---|
| 717 | u64_stats_inc(p: &stats->cnt); | 
|---|
| 718 | u64_stats_add(p: &stats->nsecs, val: duration); | 
|---|
| 719 | u64_stats_update_end_irqrestore(syncp: &stats->syncp, flags); | 
|---|
| 720 | } else { | 
|---|
| 721 | ret = dfunc(ctx, prog->insnsi, prog->bpf_func); | 
|---|
| 722 | } | 
|---|
| 723 | return ret; | 
|---|
| 724 | } | 
|---|
| 725 |  | 
|---|
| 726 | static __always_inline u32 bpf_prog_run(const struct bpf_prog *prog, const void *ctx) | 
|---|
| 727 | { | 
|---|
| 728 | return __bpf_prog_run(prog, ctx, dfunc: bpf_dispatcher_nop_func); | 
|---|
| 729 | } | 
|---|
| 730 |  | 
|---|
| 731 | /* | 
|---|
| 732 | * Use in preemptible and therefore migratable context to make sure that | 
|---|
| 733 | * the execution of the BPF program runs on one CPU. | 
|---|
| 734 | * | 
|---|
| 735 | * This uses migrate_disable/enable() explicitly to document that the | 
|---|
| 736 | * invocation of a BPF program does not require reentrancy protection | 
|---|
| 737 | * against a BPF program which is invoked from a preempting task. | 
|---|
| 738 | */ | 
|---|
| 739 | static inline u32 bpf_prog_run_pin_on_cpu(const struct bpf_prog *prog, | 
|---|
| 740 | const void *ctx) | 
|---|
| 741 | { | 
|---|
| 742 | u32 ret; | 
|---|
| 743 |  | 
|---|
| 744 | migrate_disable(); | 
|---|
| 745 | ret = bpf_prog_run(prog, ctx); | 
|---|
| 746 | migrate_enable(); | 
|---|
| 747 | return ret; | 
|---|
| 748 | } | 
|---|
| 749 |  | 
|---|
| 750 | #define BPF_SKB_CB_LEN QDISC_CB_PRIV_LEN | 
|---|
| 751 |  | 
|---|
| 752 | struct bpf_skb_data_end { | 
|---|
| 753 | struct qdisc_skb_cb qdisc_cb; | 
|---|
| 754 | void *data_meta; | 
|---|
| 755 | void *data_end; | 
|---|
| 756 | }; | 
|---|
| 757 |  | 
|---|
| 758 | struct bpf_nh_params { | 
|---|
| 759 | u32 nh_family; | 
|---|
| 760 | union { | 
|---|
| 761 | u32 ipv4_nh; | 
|---|
| 762 | struct in6_addr ipv6_nh; | 
|---|
| 763 | }; | 
|---|
| 764 | }; | 
|---|
| 765 |  | 
|---|
| 766 | /* flags for bpf_redirect_info kern_flags */ | 
|---|
| 767 | #define BPF_RI_F_RF_NO_DIRECT	BIT(0)	/* no napi_direct on return_frame */ | 
|---|
| 768 | #define BPF_RI_F_RI_INIT	BIT(1) | 
|---|
| 769 | #define BPF_RI_F_CPU_MAP_INIT	BIT(2) | 
|---|
| 770 | #define BPF_RI_F_DEV_MAP_INIT	BIT(3) | 
|---|
| 771 | #define BPF_RI_F_XSK_MAP_INIT	BIT(4) | 
|---|
| 772 |  | 
|---|
| 773 | struct bpf_redirect_info { | 
|---|
| 774 | u64 tgt_index; | 
|---|
| 775 | void *tgt_value; | 
|---|
| 776 | struct bpf_map *map; | 
|---|
| 777 | u32 flags; | 
|---|
| 778 | u32 map_id; | 
|---|
| 779 | enum bpf_map_type map_type; | 
|---|
| 780 | struct bpf_nh_params nh; | 
|---|
| 781 | u32 kern_flags; | 
|---|
| 782 | }; | 
|---|
| 783 |  | 
|---|
| 784 | struct bpf_net_context { | 
|---|
| 785 | struct bpf_redirect_info ri; | 
|---|
| 786 | struct list_head cpu_map_flush_list; | 
|---|
| 787 | struct list_head dev_map_flush_list; | 
|---|
| 788 | struct list_head xskmap_map_flush_list; | 
|---|
| 789 | }; | 
|---|
| 790 |  | 
|---|
| 791 | static inline struct bpf_net_context *bpf_net_ctx_set(struct bpf_net_context *bpf_net_ctx) | 
|---|
| 792 | { | 
|---|
| 793 | struct task_struct *tsk = current; | 
|---|
| 794 |  | 
|---|
| 795 | if (tsk->bpf_net_context != NULL) | 
|---|
| 796 | return NULL; | 
|---|
| 797 | bpf_net_ctx->ri.kern_flags = 0; | 
|---|
| 798 |  | 
|---|
| 799 | tsk->bpf_net_context = bpf_net_ctx; | 
|---|
| 800 | return bpf_net_ctx; | 
|---|
| 801 | } | 
|---|
| 802 |  | 
|---|
| 803 | static inline void bpf_net_ctx_clear(struct bpf_net_context *bpf_net_ctx) | 
|---|
| 804 | { | 
|---|
| 805 | if (bpf_net_ctx) | 
|---|
| 806 | current->bpf_net_context = NULL; | 
|---|
| 807 | } | 
|---|
| 808 |  | 
|---|
| 809 | static inline struct bpf_net_context *bpf_net_ctx_get(void) | 
|---|
| 810 | { | 
|---|
| 811 | return current->bpf_net_context; | 
|---|
| 812 | } | 
|---|
| 813 |  | 
|---|
| 814 | static inline struct bpf_redirect_info *bpf_net_ctx_get_ri(void) | 
|---|
| 815 | { | 
|---|
| 816 | struct bpf_net_context *bpf_net_ctx = bpf_net_ctx_get(); | 
|---|
| 817 |  | 
|---|
| 818 | if (!(bpf_net_ctx->ri.kern_flags & BPF_RI_F_RI_INIT)) { | 
|---|
| 819 | memset(s: &bpf_net_ctx->ri, c: 0, offsetof(struct bpf_net_context, ri.nh)); | 
|---|
| 820 | bpf_net_ctx->ri.kern_flags |= BPF_RI_F_RI_INIT; | 
|---|
| 821 | } | 
|---|
| 822 |  | 
|---|
| 823 | return &bpf_net_ctx->ri; | 
|---|
| 824 | } | 
|---|
| 825 |  | 
|---|
| 826 | static inline struct list_head *bpf_net_ctx_get_cpu_map_flush_list(void) | 
|---|
| 827 | { | 
|---|
| 828 | struct bpf_net_context *bpf_net_ctx = bpf_net_ctx_get(); | 
|---|
| 829 |  | 
|---|
| 830 | if (!(bpf_net_ctx->ri.kern_flags & BPF_RI_F_CPU_MAP_INIT)) { | 
|---|
| 831 | INIT_LIST_HEAD(list: &bpf_net_ctx->cpu_map_flush_list); | 
|---|
| 832 | bpf_net_ctx->ri.kern_flags |= BPF_RI_F_CPU_MAP_INIT; | 
|---|
| 833 | } | 
|---|
| 834 |  | 
|---|
| 835 | return &bpf_net_ctx->cpu_map_flush_list; | 
|---|
| 836 | } | 
|---|
| 837 |  | 
|---|
| 838 | static inline struct list_head *bpf_net_ctx_get_dev_flush_list(void) | 
|---|
| 839 | { | 
|---|
| 840 | struct bpf_net_context *bpf_net_ctx = bpf_net_ctx_get(); | 
|---|
| 841 |  | 
|---|
| 842 | if (!(bpf_net_ctx->ri.kern_flags & BPF_RI_F_DEV_MAP_INIT)) { | 
|---|
| 843 | INIT_LIST_HEAD(list: &bpf_net_ctx->dev_map_flush_list); | 
|---|
| 844 | bpf_net_ctx->ri.kern_flags |= BPF_RI_F_DEV_MAP_INIT; | 
|---|
| 845 | } | 
|---|
| 846 |  | 
|---|
| 847 | return &bpf_net_ctx->dev_map_flush_list; | 
|---|
| 848 | } | 
|---|
| 849 |  | 
|---|
| 850 | static inline struct list_head *bpf_net_ctx_get_xskmap_flush_list(void) | 
|---|
| 851 | { | 
|---|
| 852 | struct bpf_net_context *bpf_net_ctx = bpf_net_ctx_get(); | 
|---|
| 853 |  | 
|---|
| 854 | if (!(bpf_net_ctx->ri.kern_flags & BPF_RI_F_XSK_MAP_INIT)) { | 
|---|
| 855 | INIT_LIST_HEAD(list: &bpf_net_ctx->xskmap_map_flush_list); | 
|---|
| 856 | bpf_net_ctx->ri.kern_flags |= BPF_RI_F_XSK_MAP_INIT; | 
|---|
| 857 | } | 
|---|
| 858 |  | 
|---|
| 859 | return &bpf_net_ctx->xskmap_map_flush_list; | 
|---|
| 860 | } | 
|---|
| 861 |  | 
|---|
| 862 | static inline void bpf_net_ctx_get_all_used_flush_lists(struct list_head **lh_map, | 
|---|
| 863 | struct list_head **lh_dev, | 
|---|
| 864 | struct list_head **lh_xsk) | 
|---|
| 865 | { | 
|---|
| 866 | struct bpf_net_context *bpf_net_ctx = bpf_net_ctx_get(); | 
|---|
| 867 | u32 kern_flags = bpf_net_ctx->ri.kern_flags; | 
|---|
| 868 | struct list_head *lh; | 
|---|
| 869 |  | 
|---|
| 870 | *lh_map = *lh_dev = *lh_xsk = NULL; | 
|---|
| 871 |  | 
|---|
| 872 | if (!IS_ENABLED(CONFIG_BPF_SYSCALL)) | 
|---|
| 873 | return; | 
|---|
| 874 |  | 
|---|
| 875 | lh = &bpf_net_ctx->dev_map_flush_list; | 
|---|
| 876 | if (kern_flags & BPF_RI_F_DEV_MAP_INIT && !list_empty(head: lh)) | 
|---|
| 877 | *lh_dev = lh; | 
|---|
| 878 |  | 
|---|
| 879 | lh = &bpf_net_ctx->cpu_map_flush_list; | 
|---|
| 880 | if (kern_flags & BPF_RI_F_CPU_MAP_INIT && !list_empty(head: lh)) | 
|---|
| 881 | *lh_map = lh; | 
|---|
| 882 |  | 
|---|
| 883 | lh = &bpf_net_ctx->xskmap_map_flush_list; | 
|---|
| 884 | if (IS_ENABLED(CONFIG_XDP_SOCKETS) && | 
|---|
| 885 | kern_flags & BPF_RI_F_XSK_MAP_INIT && !list_empty(head: lh)) | 
|---|
| 886 | *lh_xsk = lh; | 
|---|
| 887 | } | 
|---|
| 888 |  | 
|---|
| 889 | /* Compute the linear packet data range [data, data_end) which | 
|---|
| 890 | * will be accessed by various program types (cls_bpf, act_bpf, | 
|---|
| 891 | * lwt, ...). Subsystems allowing direct data access must (!) | 
|---|
| 892 | * ensure that cb[] area can be written to when BPF program is | 
|---|
| 893 | * invoked (otherwise cb[] save/restore is necessary). | 
|---|
| 894 | */ | 
|---|
| 895 | static inline void bpf_compute_data_pointers(struct sk_buff *skb) | 
|---|
| 896 | { | 
|---|
| 897 | struct bpf_skb_data_end *cb = (struct bpf_skb_data_end *)skb->cb; | 
|---|
| 898 |  | 
|---|
| 899 | BUILD_BUG_ON(sizeof(*cb) > sizeof_field(struct sk_buff, cb)); | 
|---|
| 900 | cb->data_meta = skb->data - skb_metadata_len(skb); | 
|---|
| 901 | cb->data_end  = skb->data + skb_headlen(skb); | 
|---|
| 902 | } | 
|---|
| 903 |  | 
|---|
| 904 | /* Similar to bpf_compute_data_pointers(), except that save orginal | 
|---|
| 905 | * data in cb->data and cb->meta_data for restore. | 
|---|
| 906 | */ | 
|---|
| 907 | static inline void bpf_compute_and_save_data_end( | 
|---|
| 908 | struct sk_buff *skb, void **saved_data_end) | 
|---|
| 909 | { | 
|---|
| 910 | struct bpf_skb_data_end *cb = (struct bpf_skb_data_end *)skb->cb; | 
|---|
| 911 |  | 
|---|
| 912 | *saved_data_end = cb->data_end; | 
|---|
| 913 | cb->data_end  = skb->data + skb_headlen(skb); | 
|---|
| 914 | } | 
|---|
| 915 |  | 
|---|
| 916 | /* Restore data saved by bpf_compute_and_save_data_end(). */ | 
|---|
| 917 | static inline void bpf_restore_data_end( | 
|---|
| 918 | struct sk_buff *skb, void *saved_data_end) | 
|---|
| 919 | { | 
|---|
| 920 | struct bpf_skb_data_end *cb = (struct bpf_skb_data_end *)skb->cb; | 
|---|
| 921 |  | 
|---|
| 922 | cb->data_end = saved_data_end; | 
|---|
| 923 | } | 
|---|
| 924 |  | 
|---|
| 925 | static inline u8 *bpf_skb_cb(const struct sk_buff *skb) | 
|---|
| 926 | { | 
|---|
| 927 | /* eBPF programs may read/write skb->cb[] area to transfer meta | 
|---|
| 928 | * data between tail calls. Since this also needs to work with | 
|---|
| 929 | * tc, that scratch memory is mapped to qdisc_skb_cb's data area. | 
|---|
| 930 | * | 
|---|
| 931 | * In some socket filter cases, the cb unfortunately needs to be | 
|---|
| 932 | * saved/restored so that protocol specific skb->cb[] data won't | 
|---|
| 933 | * be lost. In any case, due to unpriviledged eBPF programs | 
|---|
| 934 | * attached to sockets, we need to clear the bpf_skb_cb() area | 
|---|
| 935 | * to not leak previous contents to user space. | 
|---|
| 936 | */ | 
|---|
| 937 | BUILD_BUG_ON(sizeof_field(struct __sk_buff, cb) != BPF_SKB_CB_LEN); | 
|---|
| 938 | BUILD_BUG_ON(sizeof_field(struct __sk_buff, cb) != | 
|---|
| 939 | sizeof_field(struct qdisc_skb_cb, data)); | 
|---|
| 940 |  | 
|---|
| 941 | return qdisc_skb_cb(skb)->data; | 
|---|
| 942 | } | 
|---|
| 943 |  | 
|---|
| 944 | /* Must be invoked with migration disabled */ | 
|---|
| 945 | static inline u32 __bpf_prog_run_save_cb(const struct bpf_prog *prog, | 
|---|
| 946 | const void *ctx) | 
|---|
| 947 | { | 
|---|
| 948 | const struct sk_buff *skb = ctx; | 
|---|
| 949 | u8 *cb_data = bpf_skb_cb(skb); | 
|---|
| 950 | u8 cb_saved[BPF_SKB_CB_LEN]; | 
|---|
| 951 | u32 res; | 
|---|
| 952 |  | 
|---|
| 953 | if (unlikely(prog->cb_access)) { | 
|---|
| 954 | memcpy(to: cb_saved, from: cb_data, len: sizeof(cb_saved)); | 
|---|
| 955 | memset(s: cb_data, c: 0, n: sizeof(cb_saved)); | 
|---|
| 956 | } | 
|---|
| 957 |  | 
|---|
| 958 | res = bpf_prog_run(prog, ctx: skb); | 
|---|
| 959 |  | 
|---|
| 960 | if (unlikely(prog->cb_access)) | 
|---|
| 961 | memcpy(to: cb_data, from: cb_saved, len: sizeof(cb_saved)); | 
|---|
| 962 |  | 
|---|
| 963 | return res; | 
|---|
| 964 | } | 
|---|
| 965 |  | 
|---|
| 966 | static inline u32 bpf_prog_run_save_cb(const struct bpf_prog *prog, | 
|---|
| 967 | struct sk_buff *skb) | 
|---|
| 968 | { | 
|---|
| 969 | u32 res; | 
|---|
| 970 |  | 
|---|
| 971 | migrate_disable(); | 
|---|
| 972 | res = __bpf_prog_run_save_cb(prog, ctx: skb); | 
|---|
| 973 | migrate_enable(); | 
|---|
| 974 | return res; | 
|---|
| 975 | } | 
|---|
| 976 |  | 
|---|
| 977 | static inline u32 bpf_prog_run_clear_cb(const struct bpf_prog *prog, | 
|---|
| 978 | struct sk_buff *skb) | 
|---|
| 979 | { | 
|---|
| 980 | u8 *cb_data = bpf_skb_cb(skb); | 
|---|
| 981 | u32 res; | 
|---|
| 982 |  | 
|---|
| 983 | if (unlikely(prog->cb_access)) | 
|---|
| 984 | memset(s: cb_data, c: 0, BPF_SKB_CB_LEN); | 
|---|
| 985 |  | 
|---|
| 986 | res = bpf_prog_run_pin_on_cpu(prog, ctx: skb); | 
|---|
| 987 | return res; | 
|---|
| 988 | } | 
|---|
| 989 |  | 
|---|
| 990 | DECLARE_BPF_DISPATCHER(xdp) | 
|---|
| 991 |  | 
|---|
| 992 | DECLARE_STATIC_KEY_FALSE(bpf_master_redirect_enabled_key); | 
|---|
| 993 |  | 
|---|
| 994 | u32 xdp_master_redirect(struct xdp_buff *xdp); | 
|---|
| 995 |  | 
|---|
| 996 | void bpf_prog_change_xdp(struct bpf_prog *prev_prog, struct bpf_prog *prog); | 
|---|
| 997 |  | 
|---|
| 998 | static inline u32 bpf_prog_insn_size(const struct bpf_prog *prog) | 
|---|
| 999 | { | 
|---|
| 1000 | return prog->len * sizeof(struct bpf_insn); | 
|---|
| 1001 | } | 
|---|
| 1002 |  | 
|---|
| 1003 | static inline unsigned int bpf_prog_size(unsigned int proglen) | 
|---|
| 1004 | { | 
|---|
| 1005 | return max(sizeof(struct bpf_prog), | 
|---|
| 1006 | offsetof(struct bpf_prog, insns[proglen])); | 
|---|
| 1007 | } | 
|---|
| 1008 |  | 
|---|
| 1009 | static inline bool bpf_prog_was_classic(const struct bpf_prog *prog) | 
|---|
| 1010 | { | 
|---|
| 1011 | /* When classic BPF programs have been loaded and the arch | 
|---|
| 1012 | * does not have a classic BPF JIT (anymore), they have been | 
|---|
| 1013 | * converted via bpf_migrate_filter() to eBPF and thus always | 
|---|
| 1014 | * have an unspec program type. | 
|---|
| 1015 | */ | 
|---|
| 1016 | return prog->type == BPF_PROG_TYPE_UNSPEC; | 
|---|
| 1017 | } | 
|---|
| 1018 |  | 
|---|
| 1019 | static inline u32 bpf_ctx_off_adjust_machine(u32 size) | 
|---|
| 1020 | { | 
|---|
| 1021 | const u32 size_machine = sizeof(unsigned long); | 
|---|
| 1022 |  | 
|---|
| 1023 | if (size > size_machine && size % size_machine == 0) | 
|---|
| 1024 | size = size_machine; | 
|---|
| 1025 |  | 
|---|
| 1026 | return size; | 
|---|
| 1027 | } | 
|---|
| 1028 |  | 
|---|
| 1029 | static inline bool | 
|---|
| 1030 | bpf_ctx_narrow_access_ok(u32 off, u32 size, u32 size_default) | 
|---|
| 1031 | { | 
|---|
| 1032 | return size <= size_default && (size & (size - 1)) == 0; | 
|---|
| 1033 | } | 
|---|
| 1034 |  | 
|---|
| 1035 | static inline u8 | 
|---|
| 1036 | bpf_ctx_narrow_access_offset(u32 off, u32 size, u32 size_default) | 
|---|
| 1037 | { | 
|---|
| 1038 | u8 access_off = off & (size_default - 1); | 
|---|
| 1039 |  | 
|---|
| 1040 | #ifdef __LITTLE_ENDIAN | 
|---|
| 1041 | return access_off; | 
|---|
| 1042 | #else | 
|---|
| 1043 | return size_default - (access_off + size); | 
|---|
| 1044 | #endif | 
|---|
| 1045 | } | 
|---|
| 1046 |  | 
|---|
| 1047 | #define bpf_ctx_wide_access_ok(off, size, type, field)			\ | 
|---|
| 1048 | (size == sizeof(__u64) &&					\ | 
|---|
| 1049 | off >= offsetof(type, field) &&					\ | 
|---|
| 1050 | off + sizeof(__u64) <= offsetofend(type, field) &&		\ | 
|---|
| 1051 | off % sizeof(__u64) == 0) | 
|---|
| 1052 |  | 
|---|
| 1053 | #define bpf_classic_proglen(fprog) (fprog->len * sizeof(fprog->filter[0])) | 
|---|
| 1054 |  | 
|---|
| 1055 | static inline int __must_check bpf_prog_lock_ro(struct bpf_prog *fp) | 
|---|
| 1056 | { | 
|---|
| 1057 | #ifndef CONFIG_BPF_JIT_ALWAYS_ON | 
|---|
| 1058 | if (!fp->jited) { | 
|---|
| 1059 | set_vm_flush_reset_perms(fp); | 
|---|
| 1060 | return set_memory_ro(addr: (unsigned long)fp, numpages: fp->pages); | 
|---|
| 1061 | } | 
|---|
| 1062 | #endif | 
|---|
| 1063 | return 0; | 
|---|
| 1064 | } | 
|---|
| 1065 |  | 
|---|
| 1066 | static inline int __must_check | 
|---|
| 1067 | bpf_jit_binary_lock_ro(struct bpf_binary_header *hdr) | 
|---|
| 1068 | { | 
|---|
| 1069 | set_vm_flush_reset_perms(hdr); | 
|---|
| 1070 | return set_memory_rox(addr: (unsigned long)hdr, numpages: hdr->size >> PAGE_SHIFT); | 
|---|
| 1071 | } | 
|---|
| 1072 |  | 
|---|
| 1073 | int sk_filter_trim_cap(struct sock *sk, struct sk_buff *skb, unsigned int cap, | 
|---|
| 1074 | enum skb_drop_reason *reason); | 
|---|
| 1075 |  | 
|---|
| 1076 | static inline int sk_filter(struct sock *sk, struct sk_buff *skb) | 
|---|
| 1077 | { | 
|---|
| 1078 | enum skb_drop_reason ignore_reason; | 
|---|
| 1079 |  | 
|---|
| 1080 | return sk_filter_trim_cap(sk, skb, cap: 1, reason: &ignore_reason); | 
|---|
| 1081 | } | 
|---|
| 1082 |  | 
|---|
| 1083 | static inline int sk_filter_reason(struct sock *sk, struct sk_buff *skb, | 
|---|
| 1084 | enum skb_drop_reason *reason) | 
|---|
| 1085 | { | 
|---|
| 1086 | return sk_filter_trim_cap(sk, skb, cap: 1, reason); | 
|---|
| 1087 | } | 
|---|
| 1088 |  | 
|---|
| 1089 | struct bpf_prog *bpf_prog_select_runtime(struct bpf_prog *fp, int *err); | 
|---|
| 1090 | void bpf_prog_free(struct bpf_prog *fp); | 
|---|
| 1091 |  | 
|---|
| 1092 | bool bpf_opcode_in_insntable(u8 code); | 
|---|
| 1093 |  | 
|---|
| 1094 | void bpf_prog_fill_jited_linfo(struct bpf_prog *prog, | 
|---|
| 1095 | const u32 *insn_to_jit_off); | 
|---|
| 1096 | int bpf_prog_alloc_jited_linfo(struct bpf_prog *prog); | 
|---|
| 1097 | void bpf_prog_jit_attempt_done(struct bpf_prog *prog); | 
|---|
| 1098 |  | 
|---|
| 1099 | struct bpf_prog *bpf_prog_alloc(unsigned int size, gfp_t ); | 
|---|
| 1100 | struct bpf_prog *bpf_prog_alloc_no_stats(unsigned int size, gfp_t ); | 
|---|
| 1101 | struct bpf_prog *bpf_prog_realloc(struct bpf_prog *fp_old, unsigned int size, | 
|---|
| 1102 | gfp_t ); | 
|---|
| 1103 | void __bpf_prog_free(struct bpf_prog *fp); | 
|---|
| 1104 |  | 
|---|
| 1105 | static inline void bpf_prog_unlock_free(struct bpf_prog *fp) | 
|---|
| 1106 | { | 
|---|
| 1107 | __bpf_prog_free(fp); | 
|---|
| 1108 | } | 
|---|
| 1109 |  | 
|---|
| 1110 | typedef int (*bpf_aux_classic_check_t)(struct sock_filter *filter, | 
|---|
| 1111 | unsigned int flen); | 
|---|
| 1112 |  | 
|---|
| 1113 | int bpf_prog_create(struct bpf_prog **pfp, struct sock_fprog_kern *fprog); | 
|---|
| 1114 | int bpf_prog_create_from_user(struct bpf_prog **pfp, struct sock_fprog *fprog, | 
|---|
| 1115 | bpf_aux_classic_check_t trans, bool save_orig); | 
|---|
| 1116 | void bpf_prog_destroy(struct bpf_prog *fp); | 
|---|
| 1117 |  | 
|---|
| 1118 | int sk_attach_filter(struct sock_fprog *fprog, struct sock *sk); | 
|---|
| 1119 | int sk_attach_bpf(u32 ufd, struct sock *sk); | 
|---|
| 1120 | int sk_reuseport_attach_filter(struct sock_fprog *fprog, struct sock *sk); | 
|---|
| 1121 | int sk_reuseport_attach_bpf(u32 ufd, struct sock *sk); | 
|---|
| 1122 | void sk_reuseport_prog_free(struct bpf_prog *prog); | 
|---|
| 1123 | int sk_detach_filter(struct sock *sk); | 
|---|
| 1124 | int sk_get_filter(struct sock *sk, sockptr_t optval, unsigned int len); | 
|---|
| 1125 |  | 
|---|
| 1126 | bool sk_filter_charge(struct sock *sk, struct sk_filter *fp); | 
|---|
| 1127 | void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp); | 
|---|
| 1128 |  | 
|---|
| 1129 | u64 __bpf_call_base(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); | 
|---|
| 1130 | #define __bpf_call_base_args \ | 
|---|
| 1131 | ((u64 (*)(u64, u64, u64, u64, u64, const struct bpf_insn *)) \ | 
|---|
| 1132 | (void *)__bpf_call_base) | 
|---|
| 1133 |  | 
|---|
| 1134 | struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog); | 
|---|
| 1135 | void bpf_jit_compile(struct bpf_prog *prog); | 
|---|
| 1136 | bool bpf_jit_needs_zext(void); | 
|---|
| 1137 | bool bpf_jit_inlines_helper_call(s32 imm); | 
|---|
| 1138 | bool bpf_jit_supports_subprog_tailcalls(void); | 
|---|
| 1139 | bool bpf_jit_supports_percpu_insn(void); | 
|---|
| 1140 | bool bpf_jit_supports_kfunc_call(void); | 
|---|
| 1141 | bool bpf_jit_supports_far_kfunc_call(void); | 
|---|
| 1142 | bool bpf_jit_supports_exceptions(void); | 
|---|
| 1143 | bool bpf_jit_supports_ptr_xchg(void); | 
|---|
| 1144 | bool bpf_jit_supports_arena(void); | 
|---|
| 1145 | bool bpf_jit_supports_insn(struct bpf_insn *insn, bool in_arena); | 
|---|
| 1146 | bool bpf_jit_supports_private_stack(void); | 
|---|
| 1147 | bool bpf_jit_supports_timed_may_goto(void); | 
|---|
| 1148 | u64 bpf_arch_uaddress_limit(void); | 
|---|
| 1149 | void arch_bpf_stack_walk(bool (*consume_fn)(void *cookie, u64 ip, u64 sp, u64 bp), void *cookie); | 
|---|
| 1150 | u64 arch_bpf_timed_may_goto(void); | 
|---|
| 1151 | u64 bpf_check_timed_may_goto(struct bpf_timed_may_goto *); | 
|---|
| 1152 | bool bpf_helper_changes_pkt_data(enum bpf_func_id func_id); | 
|---|
| 1153 |  | 
|---|
| 1154 | static inline bool bpf_dump_raw_ok(const struct cred *cred) | 
|---|
| 1155 | { | 
|---|
| 1156 | /* Reconstruction of call-sites is dependent on kallsyms, | 
|---|
| 1157 | * thus make dump the same restriction. | 
|---|
| 1158 | */ | 
|---|
| 1159 | return kallsyms_show_value(cred); | 
|---|
| 1160 | } | 
|---|
| 1161 |  | 
|---|
| 1162 | struct bpf_prog *bpf_patch_insn_single(struct bpf_prog *prog, u32 off, | 
|---|
| 1163 | const struct bpf_insn *patch, u32 len); | 
|---|
| 1164 | int bpf_remove_insns(struct bpf_prog *prog, u32 off, u32 cnt); | 
|---|
| 1165 |  | 
|---|
| 1166 | static inline bool xdp_return_frame_no_direct(void) | 
|---|
| 1167 | { | 
|---|
| 1168 | struct bpf_redirect_info *ri = bpf_net_ctx_get_ri(); | 
|---|
| 1169 |  | 
|---|
| 1170 | return ri->kern_flags & BPF_RI_F_RF_NO_DIRECT; | 
|---|
| 1171 | } | 
|---|
| 1172 |  | 
|---|
| 1173 | static inline void xdp_set_return_frame_no_direct(void) | 
|---|
| 1174 | { | 
|---|
| 1175 | struct bpf_redirect_info *ri = bpf_net_ctx_get_ri(); | 
|---|
| 1176 |  | 
|---|
| 1177 | ri->kern_flags |= BPF_RI_F_RF_NO_DIRECT; | 
|---|
| 1178 | } | 
|---|
| 1179 |  | 
|---|
| 1180 | static inline void xdp_clear_return_frame_no_direct(void) | 
|---|
| 1181 | { | 
|---|
| 1182 | struct bpf_redirect_info *ri = bpf_net_ctx_get_ri(); | 
|---|
| 1183 |  | 
|---|
| 1184 | ri->kern_flags &= ~BPF_RI_F_RF_NO_DIRECT; | 
|---|
| 1185 | } | 
|---|
| 1186 |  | 
|---|
| 1187 | static inline int xdp_ok_fwd_dev(const struct net_device *fwd, | 
|---|
| 1188 | unsigned int pktlen) | 
|---|
| 1189 | { | 
|---|
| 1190 | unsigned int len; | 
|---|
| 1191 |  | 
|---|
| 1192 | if (unlikely(!(fwd->flags & IFF_UP))) | 
|---|
| 1193 | return -ENETDOWN; | 
|---|
| 1194 |  | 
|---|
| 1195 | len = fwd->mtu + fwd->hard_header_len + VLAN_HLEN; | 
|---|
| 1196 | if (pktlen > len) | 
|---|
| 1197 | return -EMSGSIZE; | 
|---|
| 1198 |  | 
|---|
| 1199 | return 0; | 
|---|
| 1200 | } | 
|---|
| 1201 |  | 
|---|
| 1202 | /* The pair of xdp_do_redirect and xdp_do_flush MUST be called in the | 
|---|
| 1203 | * same cpu context. Further for best results no more than a single map | 
|---|
| 1204 | * for the do_redirect/do_flush pair should be used. This limitation is | 
|---|
| 1205 | * because we only track one map and force a flush when the map changes. | 
|---|
| 1206 | * This does not appear to be a real limitation for existing software. | 
|---|
| 1207 | */ | 
|---|
| 1208 | int xdp_do_generic_redirect(struct net_device *dev, struct sk_buff *skb, | 
|---|
| 1209 | struct xdp_buff *xdp, const struct bpf_prog *prog); | 
|---|
| 1210 | int xdp_do_redirect(struct net_device *dev, | 
|---|
| 1211 | struct xdp_buff *xdp, | 
|---|
| 1212 | const struct bpf_prog *prog); | 
|---|
| 1213 | int xdp_do_redirect_frame(struct net_device *dev, | 
|---|
| 1214 | struct xdp_buff *xdp, | 
|---|
| 1215 | struct xdp_frame *xdpf, | 
|---|
| 1216 | const struct bpf_prog *prog); | 
|---|
| 1217 | void xdp_do_flush(void); | 
|---|
| 1218 |  | 
|---|
| 1219 | void bpf_warn_invalid_xdp_action(const struct net_device *dev, | 
|---|
| 1220 | const struct bpf_prog *prog, u32 act); | 
|---|
| 1221 |  | 
|---|
| 1222 | #ifdef CONFIG_INET | 
|---|
| 1223 | struct sock *bpf_run_sk_reuseport(struct sock_reuseport *reuse, struct sock *sk, | 
|---|
| 1224 | struct bpf_prog *prog, struct sk_buff *skb, | 
|---|
| 1225 | struct sock *migrating_sk, | 
|---|
| 1226 | u32 hash); | 
|---|
| 1227 | #else | 
|---|
| 1228 | static inline struct sock * | 
|---|
| 1229 | bpf_run_sk_reuseport(struct sock_reuseport *reuse, struct sock *sk, | 
|---|
| 1230 | struct bpf_prog *prog, struct sk_buff *skb, | 
|---|
| 1231 | struct sock *migrating_sk, | 
|---|
| 1232 | u32 hash) | 
|---|
| 1233 | { | 
|---|
| 1234 | return NULL; | 
|---|
| 1235 | } | 
|---|
| 1236 | #endif | 
|---|
| 1237 |  | 
|---|
| 1238 | #ifdef CONFIG_BPF_JIT | 
|---|
| 1239 | extern int bpf_jit_enable; | 
|---|
| 1240 | extern int bpf_jit_harden; | 
|---|
| 1241 | extern int bpf_jit_kallsyms; | 
|---|
| 1242 | extern long bpf_jit_limit; | 
|---|
| 1243 | extern long bpf_jit_limit_max; | 
|---|
| 1244 |  | 
|---|
| 1245 | typedef void (*bpf_jit_fill_hole_t)(void *area, unsigned int size); | 
|---|
| 1246 |  | 
|---|
| 1247 | void bpf_jit_fill_hole_with_zero(void *area, unsigned int size); | 
|---|
| 1248 |  | 
|---|
| 1249 | struct bpf_binary_header * | 
|---|
| 1250 | bpf_jit_binary_alloc(unsigned int proglen, u8 **image_ptr, | 
|---|
| 1251 | unsigned int alignment, | 
|---|
| 1252 | bpf_jit_fill_hole_t bpf_fill_ill_insns); | 
|---|
| 1253 | void bpf_jit_binary_free(struct bpf_binary_header *hdr); | 
|---|
| 1254 | u64 bpf_jit_alloc_exec_limit(void); | 
|---|
| 1255 | void *bpf_jit_alloc_exec(unsigned long size); | 
|---|
| 1256 | void bpf_jit_free_exec(void *addr); | 
|---|
| 1257 | void bpf_jit_free(struct bpf_prog *fp); | 
|---|
| 1258 | struct bpf_binary_header * | 
|---|
| 1259 | bpf_jit_binary_pack_hdr(const struct bpf_prog *fp); | 
|---|
| 1260 |  | 
|---|
| 1261 | void *bpf_prog_pack_alloc(u32 size, bpf_jit_fill_hole_t bpf_fill_ill_insns); | 
|---|
| 1262 | void bpf_prog_pack_free(void *ptr, u32 size); | 
|---|
| 1263 |  | 
|---|
| 1264 | static inline bool bpf_prog_kallsyms_verify_off(const struct bpf_prog *fp) | 
|---|
| 1265 | { | 
|---|
| 1266 | return list_empty(&fp->aux->ksym.lnode) || | 
|---|
| 1267 | fp->aux->ksym.lnode.prev == LIST_POISON2; | 
|---|
| 1268 | } | 
|---|
| 1269 |  | 
|---|
| 1270 | struct bpf_binary_header * | 
|---|
| 1271 | bpf_jit_binary_pack_alloc(unsigned int proglen, u8 **ro_image, | 
|---|
| 1272 | unsigned int alignment, | 
|---|
| 1273 | struct bpf_binary_header **rw_hdr, | 
|---|
| 1274 | u8 **rw_image, | 
|---|
| 1275 | bpf_jit_fill_hole_t bpf_fill_ill_insns); | 
|---|
| 1276 | int bpf_jit_binary_pack_finalize(struct bpf_binary_header *ro_header, | 
|---|
| 1277 | struct bpf_binary_header *rw_header); | 
|---|
| 1278 | void bpf_jit_binary_pack_free(struct bpf_binary_header *ro_header, | 
|---|
| 1279 | struct bpf_binary_header *rw_header); | 
|---|
| 1280 |  | 
|---|
| 1281 | int bpf_jit_add_poke_descriptor(struct bpf_prog *prog, | 
|---|
| 1282 | struct bpf_jit_poke_descriptor *poke); | 
|---|
| 1283 |  | 
|---|
| 1284 | int bpf_jit_get_func_addr(const struct bpf_prog *prog, | 
|---|
| 1285 | const struct bpf_insn *insn, bool extra_pass, | 
|---|
| 1286 | u64 *func_addr, bool *func_addr_fixed); | 
|---|
| 1287 |  | 
|---|
| 1288 | const char *bpf_jit_get_prog_name(struct bpf_prog *prog); | 
|---|
| 1289 |  | 
|---|
| 1290 | struct bpf_prog *bpf_jit_blind_constants(struct bpf_prog *fp); | 
|---|
| 1291 | void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other); | 
|---|
| 1292 |  | 
|---|
| 1293 | static inline void bpf_jit_dump(unsigned int flen, unsigned int proglen, | 
|---|
| 1294 | u32 pass, void *image) | 
|---|
| 1295 | { | 
|---|
| 1296 | pr_err( "flen=%u proglen=%u pass=%u image=%p from=%s pid=%d\n", flen, | 
|---|
| 1297 | proglen, pass, image, current->comm, task_pid_nr(current)); | 
|---|
| 1298 |  | 
|---|
| 1299 | if (image) | 
|---|
| 1300 | print_hex_dump(KERN_ERR, "JIT code: ", DUMP_PREFIX_OFFSET, | 
|---|
| 1301 | 16, 1, image, proglen, false); | 
|---|
| 1302 | } | 
|---|
| 1303 |  | 
|---|
| 1304 | static inline bool bpf_jit_is_ebpf(void) | 
|---|
| 1305 | { | 
|---|
| 1306 | # ifdef CONFIG_HAVE_EBPF_JIT | 
|---|
| 1307 | return true; | 
|---|
| 1308 | # else | 
|---|
| 1309 | return false; | 
|---|
| 1310 | # endif | 
|---|
| 1311 | } | 
|---|
| 1312 |  | 
|---|
| 1313 | static inline bool ebpf_jit_enabled(void) | 
|---|
| 1314 | { | 
|---|
| 1315 | return bpf_jit_enable && bpf_jit_is_ebpf(); | 
|---|
| 1316 | } | 
|---|
| 1317 |  | 
|---|
| 1318 | static inline bool bpf_prog_ebpf_jited(const struct bpf_prog *fp) | 
|---|
| 1319 | { | 
|---|
| 1320 | return fp->jited && bpf_jit_is_ebpf(); | 
|---|
| 1321 | } | 
|---|
| 1322 |  | 
|---|
| 1323 | static inline bool bpf_jit_blinding_enabled(struct bpf_prog *prog) | 
|---|
| 1324 | { | 
|---|
| 1325 | /* These are the prerequisites, should someone ever have the | 
|---|
| 1326 | * idea to call blinding outside of them, we make sure to | 
|---|
| 1327 | * bail out. | 
|---|
| 1328 | */ | 
|---|
| 1329 | if (!bpf_jit_is_ebpf()) | 
|---|
| 1330 | return false; | 
|---|
| 1331 | if (!prog->jit_requested) | 
|---|
| 1332 | return false; | 
|---|
| 1333 | if (!bpf_jit_harden) | 
|---|
| 1334 | return false; | 
|---|
| 1335 | if (bpf_jit_harden == 1 && bpf_token_capable(prog->aux->token, CAP_BPF)) | 
|---|
| 1336 | return false; | 
|---|
| 1337 |  | 
|---|
| 1338 | return true; | 
|---|
| 1339 | } | 
|---|
| 1340 |  | 
|---|
| 1341 | static inline bool bpf_jit_kallsyms_enabled(void) | 
|---|
| 1342 | { | 
|---|
| 1343 | /* There are a couple of corner cases where kallsyms should | 
|---|
| 1344 | * not be enabled f.e. on hardening. | 
|---|
| 1345 | */ | 
|---|
| 1346 | if (bpf_jit_harden) | 
|---|
| 1347 | return false; | 
|---|
| 1348 | if (!bpf_jit_kallsyms) | 
|---|
| 1349 | return false; | 
|---|
| 1350 | if (bpf_jit_kallsyms == 1) | 
|---|
| 1351 | return true; | 
|---|
| 1352 |  | 
|---|
| 1353 | return false; | 
|---|
| 1354 | } | 
|---|
| 1355 |  | 
|---|
| 1356 | int __bpf_address_lookup(unsigned long addr, unsigned long *size, | 
|---|
| 1357 | unsigned long *off, char *sym); | 
|---|
| 1358 | bool is_bpf_text_address(unsigned long addr); | 
|---|
| 1359 | int bpf_get_kallsym(unsigned int symnum, unsigned long *value, char *type, | 
|---|
| 1360 | char *sym); | 
|---|
| 1361 | struct bpf_prog *bpf_prog_ksym_find(unsigned long addr); | 
|---|
| 1362 |  | 
|---|
| 1363 | static inline int | 
|---|
| 1364 | bpf_address_lookup(unsigned long addr, unsigned long *size, | 
|---|
| 1365 | unsigned long *off, char **modname, char *sym) | 
|---|
| 1366 | { | 
|---|
| 1367 | int ret = __bpf_address_lookup(addr, size, off, sym); | 
|---|
| 1368 |  | 
|---|
| 1369 | if (ret && modname) | 
|---|
| 1370 | *modname = NULL; | 
|---|
| 1371 | return ret; | 
|---|
| 1372 | } | 
|---|
| 1373 |  | 
|---|
| 1374 | void bpf_prog_kallsyms_add(struct bpf_prog *fp); | 
|---|
| 1375 | void bpf_prog_kallsyms_del(struct bpf_prog *fp); | 
|---|
| 1376 |  | 
|---|
| 1377 | #else /* CONFIG_BPF_JIT */ | 
|---|
| 1378 |  | 
|---|
| 1379 | static inline bool ebpf_jit_enabled(void) | 
|---|
| 1380 | { | 
|---|
| 1381 | return false; | 
|---|
| 1382 | } | 
|---|
| 1383 |  | 
|---|
| 1384 | static inline bool bpf_jit_blinding_enabled(struct bpf_prog *prog) | 
|---|
| 1385 | { | 
|---|
| 1386 | return false; | 
|---|
| 1387 | } | 
|---|
| 1388 |  | 
|---|
| 1389 | static inline bool bpf_prog_ebpf_jited(const struct bpf_prog *fp) | 
|---|
| 1390 | { | 
|---|
| 1391 | return false; | 
|---|
| 1392 | } | 
|---|
| 1393 |  | 
|---|
| 1394 | static inline int | 
|---|
| 1395 | bpf_jit_add_poke_descriptor(struct bpf_prog *prog, | 
|---|
| 1396 | struct bpf_jit_poke_descriptor *poke) | 
|---|
| 1397 | { | 
|---|
| 1398 | return -ENOTSUPP; | 
|---|
| 1399 | } | 
|---|
| 1400 |  | 
|---|
| 1401 | static inline void bpf_jit_free(struct bpf_prog *fp) | 
|---|
| 1402 | { | 
|---|
| 1403 | bpf_prog_unlock_free(fp); | 
|---|
| 1404 | } | 
|---|
| 1405 |  | 
|---|
| 1406 | static inline bool bpf_jit_kallsyms_enabled(void) | 
|---|
| 1407 | { | 
|---|
| 1408 | return false; | 
|---|
| 1409 | } | 
|---|
| 1410 |  | 
|---|
| 1411 | static inline int | 
|---|
| 1412 | __bpf_address_lookup(unsigned long addr, unsigned long *size, | 
|---|
| 1413 | unsigned long *off, char *sym) | 
|---|
| 1414 | { | 
|---|
| 1415 | return 0; | 
|---|
| 1416 | } | 
|---|
| 1417 |  | 
|---|
| 1418 | static inline bool is_bpf_text_address(unsigned long addr) | 
|---|
| 1419 | { | 
|---|
| 1420 | return false; | 
|---|
| 1421 | } | 
|---|
| 1422 |  | 
|---|
| 1423 | static inline int bpf_get_kallsym(unsigned int symnum, unsigned long *value, | 
|---|
| 1424 | char *type, char *sym) | 
|---|
| 1425 | { | 
|---|
| 1426 | return -ERANGE; | 
|---|
| 1427 | } | 
|---|
| 1428 |  | 
|---|
| 1429 | static inline struct bpf_prog *bpf_prog_ksym_find(unsigned long addr) | 
|---|
| 1430 | { | 
|---|
| 1431 | return NULL; | 
|---|
| 1432 | } | 
|---|
| 1433 |  | 
|---|
| 1434 | static inline int | 
|---|
| 1435 | bpf_address_lookup(unsigned long addr, unsigned long *size, | 
|---|
| 1436 | unsigned long *off, char **modname, char *sym) | 
|---|
| 1437 | { | 
|---|
| 1438 | return 0; | 
|---|
| 1439 | } | 
|---|
| 1440 |  | 
|---|
| 1441 | static inline void bpf_prog_kallsyms_add(struct bpf_prog *fp) | 
|---|
| 1442 | { | 
|---|
| 1443 | } | 
|---|
| 1444 |  | 
|---|
| 1445 | static inline void bpf_prog_kallsyms_del(struct bpf_prog *fp) | 
|---|
| 1446 | { | 
|---|
| 1447 | } | 
|---|
| 1448 |  | 
|---|
| 1449 | #endif /* CONFIG_BPF_JIT */ | 
|---|
| 1450 |  | 
|---|
| 1451 | void bpf_prog_kallsyms_del_all(struct bpf_prog *fp); | 
|---|
| 1452 |  | 
|---|
| 1453 | #define BPF_ANC		BIT(15) | 
|---|
| 1454 |  | 
|---|
| 1455 | static inline bool bpf_needs_clear_a(const struct sock_filter *first) | 
|---|
| 1456 | { | 
|---|
| 1457 | switch (first->code) { | 
|---|
| 1458 | case BPF_RET | BPF_K: | 
|---|
| 1459 | case BPF_LD | BPF_W | BPF_LEN: | 
|---|
| 1460 | return false; | 
|---|
| 1461 |  | 
|---|
| 1462 | case BPF_LD | BPF_W | BPF_ABS: | 
|---|
| 1463 | case BPF_LD | BPF_H | BPF_ABS: | 
|---|
| 1464 | case BPF_LD | BPF_B | BPF_ABS: | 
|---|
| 1465 | if (first->k == SKF_AD_OFF + SKF_AD_ALU_XOR_X) | 
|---|
| 1466 | return true; | 
|---|
| 1467 | return false; | 
|---|
| 1468 |  | 
|---|
| 1469 | default: | 
|---|
| 1470 | return true; | 
|---|
| 1471 | } | 
|---|
| 1472 | } | 
|---|
| 1473 |  | 
|---|
| 1474 | static inline u16 bpf_anc_helper(const struct sock_filter *ftest) | 
|---|
| 1475 | { | 
|---|
| 1476 | BUG_ON(ftest->code & BPF_ANC); | 
|---|
| 1477 |  | 
|---|
| 1478 | switch (ftest->code) { | 
|---|
| 1479 | case BPF_LD | BPF_W | BPF_ABS: | 
|---|
| 1480 | case BPF_LD | BPF_H | BPF_ABS: | 
|---|
| 1481 | case BPF_LD | BPF_B | BPF_ABS: | 
|---|
| 1482 | #define BPF_ANCILLARY(CODE)	case SKF_AD_OFF + SKF_AD_##CODE:	\ | 
|---|
| 1483 | return BPF_ANC | SKF_AD_##CODE | 
|---|
| 1484 | switch (ftest->k) { | 
|---|
| 1485 | BPF_ANCILLARY(PROTOCOL); | 
|---|
| 1486 | BPF_ANCILLARY(PKTTYPE); | 
|---|
| 1487 | BPF_ANCILLARY(IFINDEX); | 
|---|
| 1488 | BPF_ANCILLARY(NLATTR); | 
|---|
| 1489 | BPF_ANCILLARY(NLATTR_NEST); | 
|---|
| 1490 | BPF_ANCILLARY(MARK); | 
|---|
| 1491 | BPF_ANCILLARY(QUEUE); | 
|---|
| 1492 | BPF_ANCILLARY(HATYPE); | 
|---|
| 1493 | BPF_ANCILLARY(RXHASH); | 
|---|
| 1494 | BPF_ANCILLARY(CPU); | 
|---|
| 1495 | BPF_ANCILLARY(ALU_XOR_X); | 
|---|
| 1496 | BPF_ANCILLARY(VLAN_TAG); | 
|---|
| 1497 | BPF_ANCILLARY(VLAN_TAG_PRESENT); | 
|---|
| 1498 | BPF_ANCILLARY(PAY_OFFSET); | 
|---|
| 1499 | BPF_ANCILLARY(RANDOM); | 
|---|
| 1500 | BPF_ANCILLARY(VLAN_TPID); | 
|---|
| 1501 | } | 
|---|
| 1502 | fallthrough; | 
|---|
| 1503 | default: | 
|---|
| 1504 | return ftest->code; | 
|---|
| 1505 | } | 
|---|
| 1506 | } | 
|---|
| 1507 |  | 
|---|
| 1508 | void *bpf_internal_load_pointer_neg_helper(const struct sk_buff *skb, | 
|---|
| 1509 | int k, unsigned int size); | 
|---|
| 1510 |  | 
|---|
| 1511 | static inline int bpf_tell_extensions(void) | 
|---|
| 1512 | { | 
|---|
| 1513 | return SKF_AD_MAX; | 
|---|
| 1514 | } | 
|---|
| 1515 |  | 
|---|
| 1516 | struct bpf_sock_addr_kern { | 
|---|
| 1517 | struct sock *sk; | 
|---|
| 1518 | struct sockaddr *uaddr; | 
|---|
| 1519 | /* Temporary "register" to make indirect stores to nested structures | 
|---|
| 1520 | * defined above. We need three registers to make such a store, but | 
|---|
| 1521 | * only two (src and dst) are available at convert_ctx_access time | 
|---|
| 1522 | */ | 
|---|
| 1523 | u64 tmp_reg; | 
|---|
| 1524 | void *t_ctx;	/* Attach type specific context. */ | 
|---|
| 1525 | u32 uaddrlen; | 
|---|
| 1526 | }; | 
|---|
| 1527 |  | 
|---|
| 1528 | struct bpf_sock_ops_kern { | 
|---|
| 1529 | struct	sock *sk; | 
|---|
| 1530 | union { | 
|---|
| 1531 | u32 args[4]; | 
|---|
| 1532 | u32 reply; | 
|---|
| 1533 | u32 replylong[4]; | 
|---|
| 1534 | }; | 
|---|
| 1535 | struct sk_buff	*syn_skb; | 
|---|
| 1536 | struct sk_buff	*skb; | 
|---|
| 1537 | void	*skb_data_end; | 
|---|
| 1538 | u8	op; | 
|---|
| 1539 | u8	is_fullsock; | 
|---|
| 1540 | u8	is_locked_tcp_sock; | 
|---|
| 1541 | u8	remaining_opt_len; | 
|---|
| 1542 | u64	temp;			/* temp and everything after is not | 
|---|
| 1543 | * initialized to 0 before calling | 
|---|
| 1544 | * the BPF program. New fields that | 
|---|
| 1545 | * should be initialized to 0 should | 
|---|
| 1546 | * be inserted before temp. | 
|---|
| 1547 | * temp is scratch storage used by | 
|---|
| 1548 | * sock_ops_convert_ctx_access | 
|---|
| 1549 | * as temporary storage of a register. | 
|---|
| 1550 | */ | 
|---|
| 1551 | }; | 
|---|
| 1552 |  | 
|---|
| 1553 | struct bpf_sysctl_kern { | 
|---|
| 1554 | struct ctl_table_header *head; | 
|---|
| 1555 | const struct ctl_table *table; | 
|---|
| 1556 | void *cur_val; | 
|---|
| 1557 | size_t cur_len; | 
|---|
| 1558 | void *new_val; | 
|---|
| 1559 | size_t new_len; | 
|---|
| 1560 | int new_updated; | 
|---|
| 1561 | int write; | 
|---|
| 1562 | loff_t *ppos; | 
|---|
| 1563 | /* Temporary "register" for indirect stores to ppos. */ | 
|---|
| 1564 | u64 tmp_reg; | 
|---|
| 1565 | }; | 
|---|
| 1566 |  | 
|---|
| 1567 | #define BPF_SOCKOPT_KERN_BUF_SIZE	32 | 
|---|
| 1568 | struct bpf_sockopt_buf { | 
|---|
| 1569 | u8		data[BPF_SOCKOPT_KERN_BUF_SIZE]; | 
|---|
| 1570 | }; | 
|---|
| 1571 |  | 
|---|
| 1572 | struct bpf_sockopt_kern { | 
|---|
| 1573 | struct sock	*sk; | 
|---|
| 1574 | u8		*optval; | 
|---|
| 1575 | u8		*optval_end; | 
|---|
| 1576 | s32		level; | 
|---|
| 1577 | s32		optname; | 
|---|
| 1578 | s32		optlen; | 
|---|
| 1579 | /* for retval in struct bpf_cg_run_ctx */ | 
|---|
| 1580 | struct task_struct *current_task; | 
|---|
| 1581 | /* Temporary "register" for indirect stores to ppos. */ | 
|---|
| 1582 | u64		tmp_reg; | 
|---|
| 1583 | }; | 
|---|
| 1584 |  | 
|---|
| 1585 | int copy_bpf_fprog_from_user(struct sock_fprog *dst, sockptr_t src, int len); | 
|---|
| 1586 |  | 
|---|
| 1587 | struct bpf_sk_lookup_kern { | 
|---|
| 1588 | u16		family; | 
|---|
| 1589 | u16		protocol; | 
|---|
| 1590 | __be16		sport; | 
|---|
| 1591 | u16		dport; | 
|---|
| 1592 | struct { | 
|---|
| 1593 | __be32 saddr; | 
|---|
| 1594 | __be32 daddr; | 
|---|
| 1595 | } v4; | 
|---|
| 1596 | struct { | 
|---|
| 1597 | const struct in6_addr *saddr; | 
|---|
| 1598 | const struct in6_addr *daddr; | 
|---|
| 1599 | } v6; | 
|---|
| 1600 | struct sock	*selected_sk; | 
|---|
| 1601 | u32		ingress_ifindex; | 
|---|
| 1602 | bool		no_reuseport; | 
|---|
| 1603 | }; | 
|---|
| 1604 |  | 
|---|
| 1605 | extern struct static_key_false bpf_sk_lookup_enabled; | 
|---|
| 1606 |  | 
|---|
| 1607 | /* Runners for BPF_SK_LOOKUP programs to invoke on socket lookup. | 
|---|
| 1608 | * | 
|---|
| 1609 | * Allowed return values for a BPF SK_LOOKUP program are SK_PASS and | 
|---|
| 1610 | * SK_DROP. Their meaning is as follows: | 
|---|
| 1611 | * | 
|---|
| 1612 | *  SK_PASS && ctx.selected_sk != NULL: use selected_sk as lookup result | 
|---|
| 1613 | *  SK_PASS && ctx.selected_sk == NULL: continue to htable-based socket lookup | 
|---|
| 1614 | *  SK_DROP                           : terminate lookup with -ECONNREFUSED | 
|---|
| 1615 | * | 
|---|
| 1616 | * This macro aggregates return values and selected sockets from | 
|---|
| 1617 | * multiple BPF programs according to following rules in order: | 
|---|
| 1618 | * | 
|---|
| 1619 | *  1. If any program returned SK_PASS and a non-NULL ctx.selected_sk, | 
|---|
| 1620 | *     macro result is SK_PASS and last ctx.selected_sk is used. | 
|---|
| 1621 | *  2. If any program returned SK_DROP return value, | 
|---|
| 1622 | *     macro result is SK_DROP. | 
|---|
| 1623 | *  3. Otherwise result is SK_PASS and ctx.selected_sk is NULL. | 
|---|
| 1624 | * | 
|---|
| 1625 | * Caller must ensure that the prog array is non-NULL, and that the | 
|---|
| 1626 | * array as well as the programs it contains remain valid. | 
|---|
| 1627 | */ | 
|---|
| 1628 | #define BPF_PROG_SK_LOOKUP_RUN_ARRAY(array, ctx, func)			\ | 
|---|
| 1629 | ({								\ | 
|---|
| 1630 | struct bpf_sk_lookup_kern *_ctx = &(ctx);		\ | 
|---|
| 1631 | struct bpf_prog_array_item *_item;			\ | 
|---|
| 1632 | struct sock *_selected_sk = NULL;			\ | 
|---|
| 1633 | bool _no_reuseport = false;				\ | 
|---|
| 1634 | struct bpf_prog *_prog;					\ | 
|---|
| 1635 | bool _all_pass = true;					\ | 
|---|
| 1636 | u32 _ret;						\ | 
|---|
| 1637 | \ | 
|---|
| 1638 | migrate_disable();					\ | 
|---|
| 1639 | _item = &(array)->items[0];				\ | 
|---|
| 1640 | while ((_prog = READ_ONCE(_item->prog))) {		\ | 
|---|
| 1641 | /* restore most recent selection */		\ | 
|---|
| 1642 | _ctx->selected_sk = _selected_sk;		\ | 
|---|
| 1643 | _ctx->no_reuseport = _no_reuseport;		\ | 
|---|
| 1644 | \ | 
|---|
| 1645 | _ret = func(_prog, _ctx);			\ | 
|---|
| 1646 | if (_ret == SK_PASS && _ctx->selected_sk) {	\ | 
|---|
| 1647 | /* remember last non-NULL socket */	\ | 
|---|
| 1648 | _selected_sk = _ctx->selected_sk;	\ | 
|---|
| 1649 | _no_reuseport = _ctx->no_reuseport;	\ | 
|---|
| 1650 | } else if (_ret == SK_DROP && _all_pass) {	\ | 
|---|
| 1651 | _all_pass = false;			\ | 
|---|
| 1652 | }						\ | 
|---|
| 1653 | _item++;					\ | 
|---|
| 1654 | }							\ | 
|---|
| 1655 | _ctx->selected_sk = _selected_sk;			\ | 
|---|
| 1656 | _ctx->no_reuseport = _no_reuseport;			\ | 
|---|
| 1657 | migrate_enable();					\ | 
|---|
| 1658 | _all_pass || _selected_sk ? SK_PASS : SK_DROP;		\ | 
|---|
| 1659 | }) | 
|---|
| 1660 |  | 
|---|
| 1661 | static inline bool bpf_sk_lookup_run_v4(const struct net *net, int protocol, | 
|---|
| 1662 | const __be32 saddr, const __be16 sport, | 
|---|
| 1663 | const __be32 daddr, const u16 dport, | 
|---|
| 1664 | const int ifindex, struct sock **psk) | 
|---|
| 1665 | { | 
|---|
| 1666 | struct bpf_prog_array *run_array; | 
|---|
| 1667 | struct sock *selected_sk = NULL; | 
|---|
| 1668 | bool no_reuseport = false; | 
|---|
| 1669 |  | 
|---|
| 1670 | rcu_read_lock(); | 
|---|
| 1671 | run_array = rcu_dereference(net->bpf.run_array[NETNS_BPF_SK_LOOKUP]); | 
|---|
| 1672 | if (run_array) { | 
|---|
| 1673 | struct bpf_sk_lookup_kern ctx = { | 
|---|
| 1674 | .family		= AF_INET, | 
|---|
| 1675 | .protocol	= protocol, | 
|---|
| 1676 | .v4.saddr	= saddr, | 
|---|
| 1677 | .v4.daddr	= daddr, | 
|---|
| 1678 | .sport		= sport, | 
|---|
| 1679 | .dport		= dport, | 
|---|
| 1680 | .ingress_ifindex	= ifindex, | 
|---|
| 1681 | }; | 
|---|
| 1682 | u32 act; | 
|---|
| 1683 |  | 
|---|
| 1684 | act = BPF_PROG_SK_LOOKUP_RUN_ARRAY(run_array, ctx, bpf_prog_run); | 
|---|
| 1685 | if (act == SK_PASS) { | 
|---|
| 1686 | selected_sk = ctx.selected_sk; | 
|---|
| 1687 | no_reuseport = ctx.no_reuseport; | 
|---|
| 1688 | } else { | 
|---|
| 1689 | selected_sk = ERR_PTR(error: -ECONNREFUSED); | 
|---|
| 1690 | } | 
|---|
| 1691 | } | 
|---|
| 1692 | rcu_read_unlock(); | 
|---|
| 1693 | *psk = selected_sk; | 
|---|
| 1694 | return no_reuseport; | 
|---|
| 1695 | } | 
|---|
| 1696 |  | 
|---|
| 1697 | #if IS_ENABLED(CONFIG_IPV6) | 
|---|
| 1698 | static inline bool bpf_sk_lookup_run_v6(const struct net *net, int protocol, | 
|---|
| 1699 | const struct in6_addr *saddr, | 
|---|
| 1700 | const __be16 sport, | 
|---|
| 1701 | const struct in6_addr *daddr, | 
|---|
| 1702 | const u16 dport, | 
|---|
| 1703 | const int ifindex, struct sock **psk) | 
|---|
| 1704 | { | 
|---|
| 1705 | struct bpf_prog_array *run_array; | 
|---|
| 1706 | struct sock *selected_sk = NULL; | 
|---|
| 1707 | bool no_reuseport = false; | 
|---|
| 1708 |  | 
|---|
| 1709 | rcu_read_lock(); | 
|---|
| 1710 | run_array = rcu_dereference(net->bpf.run_array[NETNS_BPF_SK_LOOKUP]); | 
|---|
| 1711 | if (run_array) { | 
|---|
| 1712 | struct bpf_sk_lookup_kern ctx = { | 
|---|
| 1713 | .family		= AF_INET6, | 
|---|
| 1714 | .protocol	= protocol, | 
|---|
| 1715 | .v6.saddr	= saddr, | 
|---|
| 1716 | .v6.daddr	= daddr, | 
|---|
| 1717 | .sport		= sport, | 
|---|
| 1718 | .dport		= dport, | 
|---|
| 1719 | .ingress_ifindex	= ifindex, | 
|---|
| 1720 | }; | 
|---|
| 1721 | u32 act; | 
|---|
| 1722 |  | 
|---|
| 1723 | act = BPF_PROG_SK_LOOKUP_RUN_ARRAY(run_array, ctx, bpf_prog_run); | 
|---|
| 1724 | if (act == SK_PASS) { | 
|---|
| 1725 | selected_sk = ctx.selected_sk; | 
|---|
| 1726 | no_reuseport = ctx.no_reuseport; | 
|---|
| 1727 | } else { | 
|---|
| 1728 | selected_sk = ERR_PTR(error: -ECONNREFUSED); | 
|---|
| 1729 | } | 
|---|
| 1730 | } | 
|---|
| 1731 | rcu_read_unlock(); | 
|---|
| 1732 | *psk = selected_sk; | 
|---|
| 1733 | return no_reuseport; | 
|---|
| 1734 | } | 
|---|
| 1735 | #endif /* IS_ENABLED(CONFIG_IPV6) */ | 
|---|
| 1736 |  | 
|---|
| 1737 | static __always_inline long __bpf_xdp_redirect_map(struct bpf_map *map, u64 index, | 
|---|
| 1738 | u64 flags, const u64 flag_mask, | 
|---|
| 1739 | void *lookup_elem(struct bpf_map *map, u32 key)) | 
|---|
| 1740 | { | 
|---|
| 1741 | struct bpf_redirect_info *ri = bpf_net_ctx_get_ri(); | 
|---|
| 1742 | const u64 action_mask = XDP_ABORTED | XDP_DROP | XDP_PASS | XDP_TX; | 
|---|
| 1743 |  | 
|---|
| 1744 | /* Lower bits of the flags are used as return code on lookup failure */ | 
|---|
| 1745 | if (unlikely(flags & ~(action_mask | flag_mask))) | 
|---|
| 1746 | return XDP_ABORTED; | 
|---|
| 1747 |  | 
|---|
| 1748 | ri->tgt_value = lookup_elem(map, index); | 
|---|
| 1749 | if (unlikely(!ri->tgt_value) && !(flags & BPF_F_BROADCAST)) { | 
|---|
| 1750 | /* If the lookup fails we want to clear out the state in the | 
|---|
| 1751 | * redirect_info struct completely, so that if an eBPF program | 
|---|
| 1752 | * performs multiple lookups, the last one always takes | 
|---|
| 1753 | * precedence. | 
|---|
| 1754 | */ | 
|---|
| 1755 | ri->map_id = INT_MAX; /* Valid map id idr range: [1,INT_MAX[ */ | 
|---|
| 1756 | ri->map_type = BPF_MAP_TYPE_UNSPEC; | 
|---|
| 1757 | return flags & action_mask; | 
|---|
| 1758 | } | 
|---|
| 1759 |  | 
|---|
| 1760 | ri->tgt_index = index; | 
|---|
| 1761 | ri->map_id = map->id; | 
|---|
| 1762 | ri->map_type = map->map_type; | 
|---|
| 1763 |  | 
|---|
| 1764 | if (flags & BPF_F_BROADCAST) { | 
|---|
| 1765 | WRITE_ONCE(ri->map, map); | 
|---|
| 1766 | ri->flags = flags; | 
|---|
| 1767 | } else { | 
|---|
| 1768 | WRITE_ONCE(ri->map, NULL); | 
|---|
| 1769 | ri->flags = 0; | 
|---|
| 1770 | } | 
|---|
| 1771 |  | 
|---|
| 1772 | return XDP_REDIRECT; | 
|---|
| 1773 | } | 
|---|
| 1774 |  | 
|---|
| 1775 | #ifdef CONFIG_NET | 
|---|
| 1776 | int __bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len); | 
|---|
| 1777 | int __bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, | 
|---|
| 1778 | u32 len, u64 flags); | 
|---|
| 1779 | int __bpf_xdp_load_bytes(struct xdp_buff *xdp, u32 offset, void *buf, u32 len); | 
|---|
| 1780 | int __bpf_xdp_store_bytes(struct xdp_buff *xdp, u32 offset, void *buf, u32 len); | 
|---|
| 1781 | void *bpf_xdp_pointer(struct xdp_buff *xdp, u32 offset, u32 len); | 
|---|
| 1782 | void bpf_xdp_copy_buf(struct xdp_buff *xdp, unsigned long off, | 
|---|
| 1783 | void *buf, unsigned long len, bool flush); | 
|---|
| 1784 | void *bpf_skb_meta_pointer(struct sk_buff *skb, u32 offset); | 
|---|
| 1785 | #else /* CONFIG_NET */ | 
|---|
| 1786 | static inline int __bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, | 
|---|
| 1787 | void *to, u32 len) | 
|---|
| 1788 | { | 
|---|
| 1789 | return -EOPNOTSUPP; | 
|---|
| 1790 | } | 
|---|
| 1791 |  | 
|---|
| 1792 | static inline int __bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, | 
|---|
| 1793 | const void *from, u32 len, u64 flags) | 
|---|
| 1794 | { | 
|---|
| 1795 | return -EOPNOTSUPP; | 
|---|
| 1796 | } | 
|---|
| 1797 |  | 
|---|
| 1798 | static inline int __bpf_xdp_load_bytes(struct xdp_buff *xdp, u32 offset, | 
|---|
| 1799 | void *buf, u32 len) | 
|---|
| 1800 | { | 
|---|
| 1801 | return -EOPNOTSUPP; | 
|---|
| 1802 | } | 
|---|
| 1803 |  | 
|---|
| 1804 | static inline int __bpf_xdp_store_bytes(struct xdp_buff *xdp, u32 offset, | 
|---|
| 1805 | void *buf, u32 len) | 
|---|
| 1806 | { | 
|---|
| 1807 | return -EOPNOTSUPP; | 
|---|
| 1808 | } | 
|---|
| 1809 |  | 
|---|
| 1810 | static inline void *bpf_xdp_pointer(struct xdp_buff *xdp, u32 offset, u32 len) | 
|---|
| 1811 | { | 
|---|
| 1812 | return NULL; | 
|---|
| 1813 | } | 
|---|
| 1814 |  | 
|---|
| 1815 | static inline void bpf_xdp_copy_buf(struct xdp_buff *xdp, unsigned long off, void *buf, | 
|---|
| 1816 | unsigned long len, bool flush) | 
|---|
| 1817 | { | 
|---|
| 1818 | } | 
|---|
| 1819 |  | 
|---|
| 1820 | static inline void *bpf_skb_meta_pointer(struct sk_buff *skb, u32 offset) | 
|---|
| 1821 | { | 
|---|
| 1822 | return ERR_PTR(-EOPNOTSUPP); | 
|---|
| 1823 | } | 
|---|
| 1824 | #endif /* CONFIG_NET */ | 
|---|
| 1825 |  | 
|---|
| 1826 | #endif /* __LINUX_FILTER_H__ */ | 
|---|
| 1827 |  | 
|---|