1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * BPF JIT compiler 4 * 5 * Copyright (C) 2011-2013 Eric Dumazet (eric.dumazet@gmail.com) 6 * Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 7 */ 8 #include <linux/netdevice.h> 9 #include <linux/filter.h> 10 #include <linux/if_vlan.h> 11 #include <linux/bpf.h> 12 #include <linux/memory.h> 13 #include <linux/sort.h> 14 #include <asm/extable.h> 15 #include <asm/set_memory.h> 16 #include <asm/nospec-branch.h> 17 #include <asm/text-patching.h> 18 19 static u8 *emit_code(u8 *ptr, u32 bytes, unsigned int len) 20 { 21 if (len == 1) 22 *ptr = bytes; 23 else if (len == 2) 24 *(u16 *)ptr = bytes; 25 else { 26 *(u32 *)ptr = bytes; 27 barrier(); 28 } 29 return ptr + len; 30 } 31 32 #define EMIT(bytes, len) \ 33 do { prog = emit_code(prog, bytes, len); } while (0) 34 35 #define EMIT1(b1) EMIT(b1, 1) 36 #define EMIT2(b1, b2) EMIT((b1) + ((b2) << 8), 2) 37 #define EMIT3(b1, b2, b3) EMIT((b1) + ((b2) << 8) + ((b3) << 16), 3) 38 #define EMIT4(b1, b2, b3, b4) EMIT((b1) + ((b2) << 8) + ((b3) << 16) + ((b4) << 24), 4) 39 40 #define EMIT1_off32(b1, off) \ 41 do { EMIT1(b1); EMIT(off, 4); } while (0) 42 #define EMIT2_off32(b1, b2, off) \ 43 do { EMIT2(b1, b2); EMIT(off, 4); } while (0) 44 #define EMIT3_off32(b1, b2, b3, off) \ 45 do { EMIT3(b1, b2, b3); EMIT(off, 4); } while (0) 46 #define EMIT4_off32(b1, b2, b3, b4, off) \ 47 do { EMIT4(b1, b2, b3, b4); EMIT(off, 4); } while (0) 48 49 #ifdef CONFIG_X86_KERNEL_IBT 50 #define EMIT_ENDBR() EMIT(gen_endbr(), 4) 51 #else 52 #define EMIT_ENDBR() 53 #endif 54 55 static bool is_imm8(int value) 56 { 57 return value <= 127 && value >= -128; 58 } 59 60 static bool is_simm32(s64 value) 61 { 62 return value == (s64)(s32)value; 63 } 64 65 static bool is_uimm32(u64 value) 66 { 67 return value == (u64)(u32)value; 68 } 69 70 /* mov dst, src */ 71 #define EMIT_mov(DST, SRC) \ 72 do { \ 73 if (DST != SRC) \ 74 EMIT3(add_2mod(0x48, DST, SRC), 0x89, add_2reg(0xC0, DST, SRC)); \ 75 } while (0) 76 77 static int bpf_size_to_x86_bytes(int bpf_size) 78 { 79 if (bpf_size == BPF_W) 80 return 4; 81 else if (bpf_size == BPF_H) 82 return 2; 83 else if (bpf_size == BPF_B) 84 return 1; 85 else if (bpf_size == BPF_DW) 86 return 4; /* imm32 */ 87 else 88 return 0; 89 } 90 91 /* 92 * List of x86 cond jumps opcodes (. + s8) 93 * Add 0x10 (and an extra 0x0f) to generate far jumps (. + s32) 94 */ 95 #define X86_JB 0x72 96 #define X86_JAE 0x73 97 #define X86_JE 0x74 98 #define X86_JNE 0x75 99 #define X86_JBE 0x76 100 #define X86_JA 0x77 101 #define X86_JL 0x7C 102 #define X86_JGE 0x7D 103 #define X86_JLE 0x7E 104 #define X86_JG 0x7F 105 106 /* Pick a register outside of BPF range for JIT internal work */ 107 #define AUX_REG (MAX_BPF_JIT_REG + 1) 108 #define X86_REG_R9 (MAX_BPF_JIT_REG + 2) 109 110 /* 111 * The following table maps BPF registers to x86-64 registers. 112 * 113 * x86-64 register R12 is unused, since if used as base address 114 * register in load/store instructions, it always needs an 115 * extra byte of encoding and is callee saved. 116 * 117 * x86-64 register R9 is not used by BPF programs, but can be used by BPF 118 * trampoline. x86-64 register R10 is used for blinding (if enabled). 119 */ 120 static const int reg2hex[] = { 121 [BPF_REG_0] = 0, /* RAX */ 122 [BPF_REG_1] = 7, /* RDI */ 123 [BPF_REG_2] = 6, /* RSI */ 124 [BPF_REG_3] = 2, /* RDX */ 125 [BPF_REG_4] = 1, /* RCX */ 126 [BPF_REG_5] = 0, /* R8 */ 127 [BPF_REG_6] = 3, /* RBX callee saved */ 128 [BPF_REG_7] = 5, /* R13 callee saved */ 129 [BPF_REG_8] = 6, /* R14 callee saved */ 130 [BPF_REG_9] = 7, /* R15 callee saved */ 131 [BPF_REG_FP] = 5, /* RBP readonly */ 132 [BPF_REG_AX] = 2, /* R10 temp register */ 133 [AUX_REG] = 3, /* R11 temp register */ 134 [X86_REG_R9] = 1, /* R9 register, 6th function argument */ 135 }; 136 137 static const int reg2pt_regs[] = { 138 [BPF_REG_0] = offsetof(struct pt_regs, ax), 139 [BPF_REG_1] = offsetof(struct pt_regs, di), 140 [BPF_REG_2] = offsetof(struct pt_regs, si), 141 [BPF_REG_3] = offsetof(struct pt_regs, dx), 142 [BPF_REG_4] = offsetof(struct pt_regs, cx), 143 [BPF_REG_5] = offsetof(struct pt_regs, r8), 144 [BPF_REG_6] = offsetof(struct pt_regs, bx), 145 [BPF_REG_7] = offsetof(struct pt_regs, r13), 146 [BPF_REG_8] = offsetof(struct pt_regs, r14), 147 [BPF_REG_9] = offsetof(struct pt_regs, r15), 148 }; 149 150 /* 151 * is_ereg() == true if BPF register 'reg' maps to x86-64 r8..r15 152 * which need extra byte of encoding. 153 * rax,rcx,...,rbp have simpler encoding 154 */ 155 static bool is_ereg(u32 reg) 156 { 157 return (1 << reg) & (BIT(BPF_REG_5) | 158 BIT(AUX_REG) | 159 BIT(BPF_REG_7) | 160 BIT(BPF_REG_8) | 161 BIT(BPF_REG_9) | 162 BIT(X86_REG_R9) | 163 BIT(BPF_REG_AX)); 164 } 165 166 /* 167 * is_ereg_8l() == true if BPF register 'reg' is mapped to access x86-64 168 * lower 8-bit registers dil,sil,bpl,spl,r8b..r15b, which need extra byte 169 * of encoding. al,cl,dl,bl have simpler encoding. 170 */ 171 static bool is_ereg_8l(u32 reg) 172 { 173 return is_ereg(reg) || 174 (1 << reg) & (BIT(BPF_REG_1) | 175 BIT(BPF_REG_2) | 176 BIT(BPF_REG_FP)); 177 } 178 179 static bool is_axreg(u32 reg) 180 { 181 return reg == BPF_REG_0; 182 } 183 184 /* Add modifiers if 'reg' maps to x86-64 registers R8..R15 */ 185 static u8 add_1mod(u8 byte, u32 reg) 186 { 187 if (is_ereg(reg)) 188 byte |= 1; 189 return byte; 190 } 191 192 static u8 add_2mod(u8 byte, u32 r1, u32 r2) 193 { 194 if (is_ereg(r1)) 195 byte |= 1; 196 if (is_ereg(r2)) 197 byte |= 4; 198 return byte; 199 } 200 201 /* Encode 'dst_reg' register into x86-64 opcode 'byte' */ 202 static u8 add_1reg(u8 byte, u32 dst_reg) 203 { 204 return byte + reg2hex[dst_reg]; 205 } 206 207 /* Encode 'dst_reg' and 'src_reg' registers into x86-64 opcode 'byte' */ 208 static u8 add_2reg(u8 byte, u32 dst_reg, u32 src_reg) 209 { 210 return byte + reg2hex[dst_reg] + (reg2hex[src_reg] << 3); 211 } 212 213 /* Some 1-byte opcodes for binary ALU operations */ 214 static u8 simple_alu_opcodes[] = { 215 [BPF_ADD] = 0x01, 216 [BPF_SUB] = 0x29, 217 [BPF_AND] = 0x21, 218 [BPF_OR] = 0x09, 219 [BPF_XOR] = 0x31, 220 [BPF_LSH] = 0xE0, 221 [BPF_RSH] = 0xE8, 222 [BPF_ARSH] = 0xF8, 223 }; 224 225 static void jit_fill_hole(void *area, unsigned int size) 226 { 227 /* Fill whole space with INT3 instructions */ 228 memset(area, 0xcc, size); 229 } 230 231 int bpf_arch_text_invalidate(void *dst, size_t len) 232 { 233 return IS_ERR_OR_NULL(text_poke_set(dst, 0xcc, len)); 234 } 235 236 struct jit_context { 237 int cleanup_addr; /* Epilogue code offset */ 238 239 /* 240 * Program specific offsets of labels in the code; these rely on the 241 * JIT doing at least 2 passes, recording the position on the first 242 * pass, only to generate the correct offset on the second pass. 243 */ 244 int tail_call_direct_label; 245 int tail_call_indirect_label; 246 }; 247 248 /* Maximum number of bytes emitted while JITing one eBPF insn */ 249 #define BPF_MAX_INSN_SIZE 128 250 #define BPF_INSN_SAFETY 64 251 252 /* Number of bytes emit_patch() needs to generate instructions */ 253 #define X86_PATCH_SIZE 5 254 /* Number of bytes that will be skipped on tailcall */ 255 #define X86_TAIL_CALL_OFFSET (11 + ENDBR_INSN_SIZE) 256 257 static void push_callee_regs(u8 **pprog, bool *callee_regs_used) 258 { 259 u8 *prog = *pprog; 260 261 if (callee_regs_used[0]) 262 EMIT1(0x53); /* push rbx */ 263 if (callee_regs_used[1]) 264 EMIT2(0x41, 0x55); /* push r13 */ 265 if (callee_regs_used[2]) 266 EMIT2(0x41, 0x56); /* push r14 */ 267 if (callee_regs_used[3]) 268 EMIT2(0x41, 0x57); /* push r15 */ 269 *pprog = prog; 270 } 271 272 static void pop_callee_regs(u8 **pprog, bool *callee_regs_used) 273 { 274 u8 *prog = *pprog; 275 276 if (callee_regs_used[3]) 277 EMIT2(0x41, 0x5F); /* pop r15 */ 278 if (callee_regs_used[2]) 279 EMIT2(0x41, 0x5E); /* pop r14 */ 280 if (callee_regs_used[1]) 281 EMIT2(0x41, 0x5D); /* pop r13 */ 282 if (callee_regs_used[0]) 283 EMIT1(0x5B); /* pop rbx */ 284 *pprog = prog; 285 } 286 287 /* 288 * Emit x86-64 prologue code for BPF program. 289 * bpf_tail_call helper will skip the first X86_TAIL_CALL_OFFSET bytes 290 * while jumping to another program 291 */ 292 static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf, 293 bool tail_call_reachable, bool is_subprog) 294 { 295 u8 *prog = *pprog; 296 297 /* BPF trampoline can be made to work without these nops, 298 * but let's waste 5 bytes for now and optimize later 299 */ 300 EMIT_ENDBR(); 301 memcpy(prog, x86_nops[5], X86_PATCH_SIZE); 302 prog += X86_PATCH_SIZE; 303 if (!ebpf_from_cbpf) { 304 if (tail_call_reachable && !is_subprog) 305 EMIT2(0x31, 0xC0); /* xor eax, eax */ 306 else 307 EMIT2(0x66, 0x90); /* nop2 */ 308 } 309 EMIT1(0x55); /* push rbp */ 310 EMIT3(0x48, 0x89, 0xE5); /* mov rbp, rsp */ 311 312 /* X86_TAIL_CALL_OFFSET is here */ 313 EMIT_ENDBR(); 314 315 /* sub rsp, rounded_stack_depth */ 316 if (stack_depth) 317 EMIT3_off32(0x48, 0x81, 0xEC, round_up(stack_depth, 8)); 318 if (tail_call_reachable) 319 EMIT1(0x50); /* push rax */ 320 *pprog = prog; 321 } 322 323 static int emit_patch(u8 **pprog, void *func, void *ip, u8 opcode) 324 { 325 u8 *prog = *pprog; 326 s64 offset; 327 328 offset = func - (ip + X86_PATCH_SIZE); 329 if (!is_simm32(offset)) { 330 pr_err("Target call %p is out of range\n", func); 331 return -ERANGE; 332 } 333 EMIT1_off32(opcode, offset); 334 *pprog = prog; 335 return 0; 336 } 337 338 static int emit_call(u8 **pprog, void *func, void *ip) 339 { 340 return emit_patch(pprog, func, ip, 0xE8); 341 } 342 343 static int emit_jump(u8 **pprog, void *func, void *ip) 344 { 345 return emit_patch(pprog, func, ip, 0xE9); 346 } 347 348 static int __bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 349 void *old_addr, void *new_addr) 350 { 351 const u8 *nop_insn = x86_nops[5]; 352 u8 old_insn[X86_PATCH_SIZE]; 353 u8 new_insn[X86_PATCH_SIZE]; 354 u8 *prog; 355 int ret; 356 357 memcpy(old_insn, nop_insn, X86_PATCH_SIZE); 358 if (old_addr) { 359 prog = old_insn; 360 ret = t == BPF_MOD_CALL ? 361 emit_call(&prog, old_addr, ip) : 362 emit_jump(&prog, old_addr, ip); 363 if (ret) 364 return ret; 365 } 366 367 memcpy(new_insn, nop_insn, X86_PATCH_SIZE); 368 if (new_addr) { 369 prog = new_insn; 370 ret = t == BPF_MOD_CALL ? 371 emit_call(&prog, new_addr, ip) : 372 emit_jump(&prog, new_addr, ip); 373 if (ret) 374 return ret; 375 } 376 377 ret = -EBUSY; 378 mutex_lock(&text_mutex); 379 if (memcmp(ip, old_insn, X86_PATCH_SIZE)) 380 goto out; 381 ret = 1; 382 if (memcmp(ip, new_insn, X86_PATCH_SIZE)) { 383 text_poke_bp(ip, new_insn, X86_PATCH_SIZE, NULL); 384 ret = 0; 385 } 386 out: 387 mutex_unlock(&text_mutex); 388 return ret; 389 } 390 391 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 392 void *old_addr, void *new_addr) 393 { 394 if (!is_kernel_text((long)ip) && 395 !is_bpf_text_address((long)ip)) 396 /* BPF poking in modules is not supported */ 397 return -EINVAL; 398 399 /* 400 * See emit_prologue(), for IBT builds the trampoline hook is preceded 401 * with an ENDBR instruction. 402 */ 403 if (is_endbr(*(u32 *)ip)) 404 ip += ENDBR_INSN_SIZE; 405 406 return __bpf_arch_text_poke(ip, t, old_addr, new_addr); 407 } 408 409 #define EMIT_LFENCE() EMIT3(0x0F, 0xAE, 0xE8) 410 411 static void emit_indirect_jump(u8 **pprog, int reg, u8 *ip) 412 { 413 u8 *prog = *pprog; 414 415 #ifdef CONFIG_RETPOLINE 416 if (cpu_feature_enabled(X86_FEATURE_RETPOLINE_LFENCE)) { 417 EMIT_LFENCE(); 418 EMIT2(0xFF, 0xE0 + reg); 419 } else if (cpu_feature_enabled(X86_FEATURE_RETPOLINE)) { 420 OPTIMIZER_HIDE_VAR(reg); 421 emit_jump(&prog, &__x86_indirect_thunk_array[reg], ip); 422 } else 423 #endif 424 EMIT2(0xFF, 0xE0 + reg); 425 426 *pprog = prog; 427 } 428 429 /* 430 * Generate the following code: 431 * 432 * ... bpf_tail_call(void *ctx, struct bpf_array *array, u64 index) ... 433 * if (index >= array->map.max_entries) 434 * goto out; 435 * if (tail_call_cnt++ >= MAX_TAIL_CALL_CNT) 436 * goto out; 437 * prog = array->ptrs[index]; 438 * if (prog == NULL) 439 * goto out; 440 * goto *(prog->bpf_func + prologue_size); 441 * out: 442 */ 443 static void emit_bpf_tail_call_indirect(u8 **pprog, bool *callee_regs_used, 444 u32 stack_depth, u8 *ip, 445 struct jit_context *ctx) 446 { 447 int tcc_off = -4 - round_up(stack_depth, 8); 448 u8 *prog = *pprog, *start = *pprog; 449 int offset; 450 451 /* 452 * rdi - pointer to ctx 453 * rsi - pointer to bpf_array 454 * rdx - index in bpf_array 455 */ 456 457 /* 458 * if (index >= array->map.max_entries) 459 * goto out; 460 */ 461 EMIT2(0x89, 0xD2); /* mov edx, edx */ 462 EMIT3(0x39, 0x56, /* cmp dword ptr [rsi + 16], edx */ 463 offsetof(struct bpf_array, map.max_entries)); 464 465 offset = ctx->tail_call_indirect_label - (prog + 2 - start); 466 EMIT2(X86_JBE, offset); /* jbe out */ 467 468 /* 469 * if (tail_call_cnt++ >= MAX_TAIL_CALL_CNT) 470 * goto out; 471 */ 472 EMIT2_off32(0x8B, 0x85, tcc_off); /* mov eax, dword ptr [rbp - tcc_off] */ 473 EMIT3(0x83, 0xF8, MAX_TAIL_CALL_CNT); /* cmp eax, MAX_TAIL_CALL_CNT */ 474 475 offset = ctx->tail_call_indirect_label - (prog + 2 - start); 476 EMIT2(X86_JAE, offset); /* jae out */ 477 EMIT3(0x83, 0xC0, 0x01); /* add eax, 1 */ 478 EMIT2_off32(0x89, 0x85, tcc_off); /* mov dword ptr [rbp - tcc_off], eax */ 479 480 /* prog = array->ptrs[index]; */ 481 EMIT4_off32(0x48, 0x8B, 0x8C, 0xD6, /* mov rcx, [rsi + rdx * 8 + offsetof(...)] */ 482 offsetof(struct bpf_array, ptrs)); 483 484 /* 485 * if (prog == NULL) 486 * goto out; 487 */ 488 EMIT3(0x48, 0x85, 0xC9); /* test rcx,rcx */ 489 490 offset = ctx->tail_call_indirect_label - (prog + 2 - start); 491 EMIT2(X86_JE, offset); /* je out */ 492 493 pop_callee_regs(&prog, callee_regs_used); 494 495 EMIT1(0x58); /* pop rax */ 496 if (stack_depth) 497 EMIT3_off32(0x48, 0x81, 0xC4, /* add rsp, sd */ 498 round_up(stack_depth, 8)); 499 500 /* goto *(prog->bpf_func + X86_TAIL_CALL_OFFSET); */ 501 EMIT4(0x48, 0x8B, 0x49, /* mov rcx, qword ptr [rcx + 32] */ 502 offsetof(struct bpf_prog, bpf_func)); 503 EMIT4(0x48, 0x83, 0xC1, /* add rcx, X86_TAIL_CALL_OFFSET */ 504 X86_TAIL_CALL_OFFSET); 505 /* 506 * Now we're ready to jump into next BPF program 507 * rdi == ctx (1st arg) 508 * rcx == prog->bpf_func + X86_TAIL_CALL_OFFSET 509 */ 510 emit_indirect_jump(&prog, 1 /* rcx */, ip + (prog - start)); 511 512 /* out: */ 513 ctx->tail_call_indirect_label = prog - start; 514 *pprog = prog; 515 } 516 517 static void emit_bpf_tail_call_direct(struct bpf_jit_poke_descriptor *poke, 518 u8 **pprog, u8 *ip, 519 bool *callee_regs_used, u32 stack_depth, 520 struct jit_context *ctx) 521 { 522 int tcc_off = -4 - round_up(stack_depth, 8); 523 u8 *prog = *pprog, *start = *pprog; 524 int offset; 525 526 /* 527 * if (tail_call_cnt++ >= MAX_TAIL_CALL_CNT) 528 * goto out; 529 */ 530 EMIT2_off32(0x8B, 0x85, tcc_off); /* mov eax, dword ptr [rbp - tcc_off] */ 531 EMIT3(0x83, 0xF8, MAX_TAIL_CALL_CNT); /* cmp eax, MAX_TAIL_CALL_CNT */ 532 533 offset = ctx->tail_call_direct_label - (prog + 2 - start); 534 EMIT2(X86_JAE, offset); /* jae out */ 535 EMIT3(0x83, 0xC0, 0x01); /* add eax, 1 */ 536 EMIT2_off32(0x89, 0x85, tcc_off); /* mov dword ptr [rbp - tcc_off], eax */ 537 538 poke->tailcall_bypass = ip + (prog - start); 539 poke->adj_off = X86_TAIL_CALL_OFFSET; 540 poke->tailcall_target = ip + ctx->tail_call_direct_label - X86_PATCH_SIZE; 541 poke->bypass_addr = (u8 *)poke->tailcall_target + X86_PATCH_SIZE; 542 543 emit_jump(&prog, (u8 *)poke->tailcall_target + X86_PATCH_SIZE, 544 poke->tailcall_bypass); 545 546 pop_callee_regs(&prog, callee_regs_used); 547 EMIT1(0x58); /* pop rax */ 548 if (stack_depth) 549 EMIT3_off32(0x48, 0x81, 0xC4, round_up(stack_depth, 8)); 550 551 memcpy(prog, x86_nops[5], X86_PATCH_SIZE); 552 prog += X86_PATCH_SIZE; 553 554 /* out: */ 555 ctx->tail_call_direct_label = prog - start; 556 557 *pprog = prog; 558 } 559 560 static void bpf_tail_call_direct_fixup(struct bpf_prog *prog) 561 { 562 struct bpf_jit_poke_descriptor *poke; 563 struct bpf_array *array; 564 struct bpf_prog *target; 565 int i, ret; 566 567 for (i = 0; i < prog->aux->size_poke_tab; i++) { 568 poke = &prog->aux->poke_tab[i]; 569 if (poke->aux && poke->aux != prog->aux) 570 continue; 571 572 WARN_ON_ONCE(READ_ONCE(poke->tailcall_target_stable)); 573 574 if (poke->reason != BPF_POKE_REASON_TAIL_CALL) 575 continue; 576 577 array = container_of(poke->tail_call.map, struct bpf_array, map); 578 mutex_lock(&array->aux->poke_mutex); 579 target = array->ptrs[poke->tail_call.key]; 580 if (target) { 581 ret = __bpf_arch_text_poke(poke->tailcall_target, 582 BPF_MOD_JUMP, NULL, 583 (u8 *)target->bpf_func + 584 poke->adj_off); 585 BUG_ON(ret < 0); 586 ret = __bpf_arch_text_poke(poke->tailcall_bypass, 587 BPF_MOD_JUMP, 588 (u8 *)poke->tailcall_target + 589 X86_PATCH_SIZE, NULL); 590 BUG_ON(ret < 0); 591 } 592 WRITE_ONCE(poke->tailcall_target_stable, true); 593 mutex_unlock(&array->aux->poke_mutex); 594 } 595 } 596 597 static void emit_mov_imm32(u8 **pprog, bool sign_propagate, 598 u32 dst_reg, const u32 imm32) 599 { 600 u8 *prog = *pprog; 601 u8 b1, b2, b3; 602 603 /* 604 * Optimization: if imm32 is positive, use 'mov %eax, imm32' 605 * (which zero-extends imm32) to save 2 bytes. 606 */ 607 if (sign_propagate && (s32)imm32 < 0) { 608 /* 'mov %rax, imm32' sign extends imm32 */ 609 b1 = add_1mod(0x48, dst_reg); 610 b2 = 0xC7; 611 b3 = 0xC0; 612 EMIT3_off32(b1, b2, add_1reg(b3, dst_reg), imm32); 613 goto done; 614 } 615 616 /* 617 * Optimization: if imm32 is zero, use 'xor %eax, %eax' 618 * to save 3 bytes. 619 */ 620 if (imm32 == 0) { 621 if (is_ereg(dst_reg)) 622 EMIT1(add_2mod(0x40, dst_reg, dst_reg)); 623 b2 = 0x31; /* xor */ 624 b3 = 0xC0; 625 EMIT2(b2, add_2reg(b3, dst_reg, dst_reg)); 626 goto done; 627 } 628 629 /* mov %eax, imm32 */ 630 if (is_ereg(dst_reg)) 631 EMIT1(add_1mod(0x40, dst_reg)); 632 EMIT1_off32(add_1reg(0xB8, dst_reg), imm32); 633 done: 634 *pprog = prog; 635 } 636 637 static void emit_mov_imm64(u8 **pprog, u32 dst_reg, 638 const u32 imm32_hi, const u32 imm32_lo) 639 { 640 u8 *prog = *pprog; 641 642 if (is_uimm32(((u64)imm32_hi << 32) | (u32)imm32_lo)) { 643 /* 644 * For emitting plain u32, where sign bit must not be 645 * propagated LLVM tends to load imm64 over mov32 646 * directly, so save couple of bytes by just doing 647 * 'mov %eax, imm32' instead. 648 */ 649 emit_mov_imm32(&prog, false, dst_reg, imm32_lo); 650 } else { 651 /* movabsq %rax, imm64 */ 652 EMIT2(add_1mod(0x48, dst_reg), add_1reg(0xB8, dst_reg)); 653 EMIT(imm32_lo, 4); 654 EMIT(imm32_hi, 4); 655 } 656 657 *pprog = prog; 658 } 659 660 static void emit_mov_reg(u8 **pprog, bool is64, u32 dst_reg, u32 src_reg) 661 { 662 u8 *prog = *pprog; 663 664 if (is64) { 665 /* mov dst, src */ 666 EMIT_mov(dst_reg, src_reg); 667 } else { 668 /* mov32 dst, src */ 669 if (is_ereg(dst_reg) || is_ereg(src_reg)) 670 EMIT1(add_2mod(0x40, dst_reg, src_reg)); 671 EMIT2(0x89, add_2reg(0xC0, dst_reg, src_reg)); 672 } 673 674 *pprog = prog; 675 } 676 677 /* Emit the suffix (ModR/M etc) for addressing *(ptr_reg + off) and val_reg */ 678 static void emit_insn_suffix(u8 **pprog, u32 ptr_reg, u32 val_reg, int off) 679 { 680 u8 *prog = *pprog; 681 682 if (is_imm8(off)) { 683 /* 1-byte signed displacement. 684 * 685 * If off == 0 we could skip this and save one extra byte, but 686 * special case of x86 R13 which always needs an offset is not 687 * worth the hassle 688 */ 689 EMIT2(add_2reg(0x40, ptr_reg, val_reg), off); 690 } else { 691 /* 4-byte signed displacement */ 692 EMIT1_off32(add_2reg(0x80, ptr_reg, val_reg), off); 693 } 694 *pprog = prog; 695 } 696 697 /* 698 * Emit a REX byte if it will be necessary to address these registers 699 */ 700 static void maybe_emit_mod(u8 **pprog, u32 dst_reg, u32 src_reg, bool is64) 701 { 702 u8 *prog = *pprog; 703 704 if (is64) 705 EMIT1(add_2mod(0x48, dst_reg, src_reg)); 706 else if (is_ereg(dst_reg) || is_ereg(src_reg)) 707 EMIT1(add_2mod(0x40, dst_reg, src_reg)); 708 *pprog = prog; 709 } 710 711 /* 712 * Similar version of maybe_emit_mod() for a single register 713 */ 714 static void maybe_emit_1mod(u8 **pprog, u32 reg, bool is64) 715 { 716 u8 *prog = *pprog; 717 718 if (is64) 719 EMIT1(add_1mod(0x48, reg)); 720 else if (is_ereg(reg)) 721 EMIT1(add_1mod(0x40, reg)); 722 *pprog = prog; 723 } 724 725 /* LDX: dst_reg = *(u8*)(src_reg + off) */ 726 static void emit_ldx(u8 **pprog, u32 size, u32 dst_reg, u32 src_reg, int off) 727 { 728 u8 *prog = *pprog; 729 730 switch (size) { 731 case BPF_B: 732 /* Emit 'movzx rax, byte ptr [rax + off]' */ 733 EMIT3(add_2mod(0x48, src_reg, dst_reg), 0x0F, 0xB6); 734 break; 735 case BPF_H: 736 /* Emit 'movzx rax, word ptr [rax + off]' */ 737 EMIT3(add_2mod(0x48, src_reg, dst_reg), 0x0F, 0xB7); 738 break; 739 case BPF_W: 740 /* Emit 'mov eax, dword ptr [rax+0x14]' */ 741 if (is_ereg(dst_reg) || is_ereg(src_reg)) 742 EMIT2(add_2mod(0x40, src_reg, dst_reg), 0x8B); 743 else 744 EMIT1(0x8B); 745 break; 746 case BPF_DW: 747 /* Emit 'mov rax, qword ptr [rax+0x14]' */ 748 EMIT2(add_2mod(0x48, src_reg, dst_reg), 0x8B); 749 break; 750 } 751 emit_insn_suffix(&prog, src_reg, dst_reg, off); 752 *pprog = prog; 753 } 754 755 /* STX: *(u8*)(dst_reg + off) = src_reg */ 756 static void emit_stx(u8 **pprog, u32 size, u32 dst_reg, u32 src_reg, int off) 757 { 758 u8 *prog = *pprog; 759 760 switch (size) { 761 case BPF_B: 762 /* Emit 'mov byte ptr [rax + off], al' */ 763 if (is_ereg(dst_reg) || is_ereg_8l(src_reg)) 764 /* Add extra byte for eregs or SIL,DIL,BPL in src_reg */ 765 EMIT2(add_2mod(0x40, dst_reg, src_reg), 0x88); 766 else 767 EMIT1(0x88); 768 break; 769 case BPF_H: 770 if (is_ereg(dst_reg) || is_ereg(src_reg)) 771 EMIT3(0x66, add_2mod(0x40, dst_reg, src_reg), 0x89); 772 else 773 EMIT2(0x66, 0x89); 774 break; 775 case BPF_W: 776 if (is_ereg(dst_reg) || is_ereg(src_reg)) 777 EMIT2(add_2mod(0x40, dst_reg, src_reg), 0x89); 778 else 779 EMIT1(0x89); 780 break; 781 case BPF_DW: 782 EMIT2(add_2mod(0x48, dst_reg, src_reg), 0x89); 783 break; 784 } 785 emit_insn_suffix(&prog, dst_reg, src_reg, off); 786 *pprog = prog; 787 } 788 789 static int emit_atomic(u8 **pprog, u8 atomic_op, 790 u32 dst_reg, u32 src_reg, s16 off, u8 bpf_size) 791 { 792 u8 *prog = *pprog; 793 794 EMIT1(0xF0); /* lock prefix */ 795 796 maybe_emit_mod(&prog, dst_reg, src_reg, bpf_size == BPF_DW); 797 798 /* emit opcode */ 799 switch (atomic_op) { 800 case BPF_ADD: 801 case BPF_AND: 802 case BPF_OR: 803 case BPF_XOR: 804 /* lock *(u32/u64*)(dst_reg + off) <op>= src_reg */ 805 EMIT1(simple_alu_opcodes[atomic_op]); 806 break; 807 case BPF_ADD | BPF_FETCH: 808 /* src_reg = atomic_fetch_add(dst_reg + off, src_reg); */ 809 EMIT2(0x0F, 0xC1); 810 break; 811 case BPF_XCHG: 812 /* src_reg = atomic_xchg(dst_reg + off, src_reg); */ 813 EMIT1(0x87); 814 break; 815 case BPF_CMPXCHG: 816 /* r0 = atomic_cmpxchg(dst_reg + off, r0, src_reg); */ 817 EMIT2(0x0F, 0xB1); 818 break; 819 default: 820 pr_err("bpf_jit: unknown atomic opcode %02x\n", atomic_op); 821 return -EFAULT; 822 } 823 824 emit_insn_suffix(&prog, dst_reg, src_reg, off); 825 826 *pprog = prog; 827 return 0; 828 } 829 830 bool ex_handler_bpf(const struct exception_table_entry *x, struct pt_regs *regs) 831 { 832 u32 reg = x->fixup >> 8; 833 834 /* jump over faulting load and clear dest register */ 835 *(unsigned long *)((void *)regs + reg) = 0; 836 regs->ip += x->fixup & 0xff; 837 return true; 838 } 839 840 static void detect_reg_usage(struct bpf_insn *insn, int insn_cnt, 841 bool *regs_used, bool *tail_call_seen) 842 { 843 int i; 844 845 for (i = 1; i <= insn_cnt; i++, insn++) { 846 if (insn->code == (BPF_JMP | BPF_TAIL_CALL)) 847 *tail_call_seen = true; 848 if (insn->dst_reg == BPF_REG_6 || insn->src_reg == BPF_REG_6) 849 regs_used[0] = true; 850 if (insn->dst_reg == BPF_REG_7 || insn->src_reg == BPF_REG_7) 851 regs_used[1] = true; 852 if (insn->dst_reg == BPF_REG_8 || insn->src_reg == BPF_REG_8) 853 regs_used[2] = true; 854 if (insn->dst_reg == BPF_REG_9 || insn->src_reg == BPF_REG_9) 855 regs_used[3] = true; 856 } 857 } 858 859 static void emit_nops(u8 **pprog, int len) 860 { 861 u8 *prog = *pprog; 862 int i, noplen; 863 864 while (len > 0) { 865 noplen = len; 866 867 if (noplen > ASM_NOP_MAX) 868 noplen = ASM_NOP_MAX; 869 870 for (i = 0; i < noplen; i++) 871 EMIT1(x86_nops[noplen][i]); 872 len -= noplen; 873 } 874 875 *pprog = prog; 876 } 877 878 #define INSN_SZ_DIFF (((addrs[i] - addrs[i - 1]) - (prog - temp))) 879 880 static int do_jit(struct bpf_prog *bpf_prog, int *addrs, u8 *image, u8 *rw_image, 881 int oldproglen, struct jit_context *ctx, bool jmp_padding) 882 { 883 bool tail_call_reachable = bpf_prog->aux->tail_call_reachable; 884 struct bpf_insn *insn = bpf_prog->insnsi; 885 bool callee_regs_used[4] = {}; 886 int insn_cnt = bpf_prog->len; 887 bool tail_call_seen = false; 888 bool seen_exit = false; 889 u8 temp[BPF_MAX_INSN_SIZE + BPF_INSN_SAFETY]; 890 int i, excnt = 0; 891 int ilen, proglen = 0; 892 u8 *prog = temp; 893 int err; 894 895 detect_reg_usage(insn, insn_cnt, callee_regs_used, 896 &tail_call_seen); 897 898 /* tail call's presence in current prog implies it is reachable */ 899 tail_call_reachable |= tail_call_seen; 900 901 emit_prologue(&prog, bpf_prog->aux->stack_depth, 902 bpf_prog_was_classic(bpf_prog), tail_call_reachable, 903 bpf_prog->aux->func_idx != 0); 904 push_callee_regs(&prog, callee_regs_used); 905 906 ilen = prog - temp; 907 if (rw_image) 908 memcpy(rw_image + proglen, temp, ilen); 909 proglen += ilen; 910 addrs[0] = proglen; 911 prog = temp; 912 913 for (i = 1; i <= insn_cnt; i++, insn++) { 914 const s32 imm32 = insn->imm; 915 u32 dst_reg = insn->dst_reg; 916 u32 src_reg = insn->src_reg; 917 u8 b2 = 0, b3 = 0; 918 u8 *start_of_ldx; 919 s64 jmp_offset; 920 u8 jmp_cond; 921 u8 *func; 922 int nops; 923 924 switch (insn->code) { 925 /* ALU */ 926 case BPF_ALU | BPF_ADD | BPF_X: 927 case BPF_ALU | BPF_SUB | BPF_X: 928 case BPF_ALU | BPF_AND | BPF_X: 929 case BPF_ALU | BPF_OR | BPF_X: 930 case BPF_ALU | BPF_XOR | BPF_X: 931 case BPF_ALU64 | BPF_ADD | BPF_X: 932 case BPF_ALU64 | BPF_SUB | BPF_X: 933 case BPF_ALU64 | BPF_AND | BPF_X: 934 case BPF_ALU64 | BPF_OR | BPF_X: 935 case BPF_ALU64 | BPF_XOR | BPF_X: 936 maybe_emit_mod(&prog, dst_reg, src_reg, 937 BPF_CLASS(insn->code) == BPF_ALU64); 938 b2 = simple_alu_opcodes[BPF_OP(insn->code)]; 939 EMIT2(b2, add_2reg(0xC0, dst_reg, src_reg)); 940 break; 941 942 case BPF_ALU64 | BPF_MOV | BPF_X: 943 case BPF_ALU | BPF_MOV | BPF_X: 944 emit_mov_reg(&prog, 945 BPF_CLASS(insn->code) == BPF_ALU64, 946 dst_reg, src_reg); 947 break; 948 949 /* neg dst */ 950 case BPF_ALU | BPF_NEG: 951 case BPF_ALU64 | BPF_NEG: 952 maybe_emit_1mod(&prog, dst_reg, 953 BPF_CLASS(insn->code) == BPF_ALU64); 954 EMIT2(0xF7, add_1reg(0xD8, dst_reg)); 955 break; 956 957 case BPF_ALU | BPF_ADD | BPF_K: 958 case BPF_ALU | BPF_SUB | BPF_K: 959 case BPF_ALU | BPF_AND | BPF_K: 960 case BPF_ALU | BPF_OR | BPF_K: 961 case BPF_ALU | BPF_XOR | BPF_K: 962 case BPF_ALU64 | BPF_ADD | BPF_K: 963 case BPF_ALU64 | BPF_SUB | BPF_K: 964 case BPF_ALU64 | BPF_AND | BPF_K: 965 case BPF_ALU64 | BPF_OR | BPF_K: 966 case BPF_ALU64 | BPF_XOR | BPF_K: 967 maybe_emit_1mod(&prog, dst_reg, 968 BPF_CLASS(insn->code) == BPF_ALU64); 969 970 /* 971 * b3 holds 'normal' opcode, b2 short form only valid 972 * in case dst is eax/rax. 973 */ 974 switch (BPF_OP(insn->code)) { 975 case BPF_ADD: 976 b3 = 0xC0; 977 b2 = 0x05; 978 break; 979 case BPF_SUB: 980 b3 = 0xE8; 981 b2 = 0x2D; 982 break; 983 case BPF_AND: 984 b3 = 0xE0; 985 b2 = 0x25; 986 break; 987 case BPF_OR: 988 b3 = 0xC8; 989 b2 = 0x0D; 990 break; 991 case BPF_XOR: 992 b3 = 0xF0; 993 b2 = 0x35; 994 break; 995 } 996 997 if (is_imm8(imm32)) 998 EMIT3(0x83, add_1reg(b3, dst_reg), imm32); 999 else if (is_axreg(dst_reg)) 1000 EMIT1_off32(b2, imm32); 1001 else 1002 EMIT2_off32(0x81, add_1reg(b3, dst_reg), imm32); 1003 break; 1004 1005 case BPF_ALU64 | BPF_MOV | BPF_K: 1006 case BPF_ALU | BPF_MOV | BPF_K: 1007 emit_mov_imm32(&prog, BPF_CLASS(insn->code) == BPF_ALU64, 1008 dst_reg, imm32); 1009 break; 1010 1011 case BPF_LD | BPF_IMM | BPF_DW: 1012 emit_mov_imm64(&prog, dst_reg, insn[1].imm, insn[0].imm); 1013 insn++; 1014 i++; 1015 break; 1016 1017 /* dst %= src, dst /= src, dst %= imm32, dst /= imm32 */ 1018 case BPF_ALU | BPF_MOD | BPF_X: 1019 case BPF_ALU | BPF_DIV | BPF_X: 1020 case BPF_ALU | BPF_MOD | BPF_K: 1021 case BPF_ALU | BPF_DIV | BPF_K: 1022 case BPF_ALU64 | BPF_MOD | BPF_X: 1023 case BPF_ALU64 | BPF_DIV | BPF_X: 1024 case BPF_ALU64 | BPF_MOD | BPF_K: 1025 case BPF_ALU64 | BPF_DIV | BPF_K: { 1026 bool is64 = BPF_CLASS(insn->code) == BPF_ALU64; 1027 1028 if (dst_reg != BPF_REG_0) 1029 EMIT1(0x50); /* push rax */ 1030 if (dst_reg != BPF_REG_3) 1031 EMIT1(0x52); /* push rdx */ 1032 1033 if (BPF_SRC(insn->code) == BPF_X) { 1034 if (src_reg == BPF_REG_0 || 1035 src_reg == BPF_REG_3) { 1036 /* mov r11, src_reg */ 1037 EMIT_mov(AUX_REG, src_reg); 1038 src_reg = AUX_REG; 1039 } 1040 } else { 1041 /* mov r11, imm32 */ 1042 EMIT3_off32(0x49, 0xC7, 0xC3, imm32); 1043 src_reg = AUX_REG; 1044 } 1045 1046 if (dst_reg != BPF_REG_0) 1047 /* mov rax, dst_reg */ 1048 emit_mov_reg(&prog, is64, BPF_REG_0, dst_reg); 1049 1050 /* 1051 * xor edx, edx 1052 * equivalent to 'xor rdx, rdx', but one byte less 1053 */ 1054 EMIT2(0x31, 0xd2); 1055 1056 /* div src_reg */ 1057 maybe_emit_1mod(&prog, src_reg, is64); 1058 EMIT2(0xF7, add_1reg(0xF0, src_reg)); 1059 1060 if (BPF_OP(insn->code) == BPF_MOD && 1061 dst_reg != BPF_REG_3) 1062 /* mov dst_reg, rdx */ 1063 emit_mov_reg(&prog, is64, dst_reg, BPF_REG_3); 1064 else if (BPF_OP(insn->code) == BPF_DIV && 1065 dst_reg != BPF_REG_0) 1066 /* mov dst_reg, rax */ 1067 emit_mov_reg(&prog, is64, dst_reg, BPF_REG_0); 1068 1069 if (dst_reg != BPF_REG_3) 1070 EMIT1(0x5A); /* pop rdx */ 1071 if (dst_reg != BPF_REG_0) 1072 EMIT1(0x58); /* pop rax */ 1073 break; 1074 } 1075 1076 case BPF_ALU | BPF_MUL | BPF_K: 1077 case BPF_ALU64 | BPF_MUL | BPF_K: 1078 maybe_emit_mod(&prog, dst_reg, dst_reg, 1079 BPF_CLASS(insn->code) == BPF_ALU64); 1080 1081 if (is_imm8(imm32)) 1082 /* imul dst_reg, dst_reg, imm8 */ 1083 EMIT3(0x6B, add_2reg(0xC0, dst_reg, dst_reg), 1084 imm32); 1085 else 1086 /* imul dst_reg, dst_reg, imm32 */ 1087 EMIT2_off32(0x69, 1088 add_2reg(0xC0, dst_reg, dst_reg), 1089 imm32); 1090 break; 1091 1092 case BPF_ALU | BPF_MUL | BPF_X: 1093 case BPF_ALU64 | BPF_MUL | BPF_X: 1094 maybe_emit_mod(&prog, src_reg, dst_reg, 1095 BPF_CLASS(insn->code) == BPF_ALU64); 1096 1097 /* imul dst_reg, src_reg */ 1098 EMIT3(0x0F, 0xAF, add_2reg(0xC0, src_reg, dst_reg)); 1099 break; 1100 1101 /* Shifts */ 1102 case BPF_ALU | BPF_LSH | BPF_K: 1103 case BPF_ALU | BPF_RSH | BPF_K: 1104 case BPF_ALU | BPF_ARSH | BPF_K: 1105 case BPF_ALU64 | BPF_LSH | BPF_K: 1106 case BPF_ALU64 | BPF_RSH | BPF_K: 1107 case BPF_ALU64 | BPF_ARSH | BPF_K: 1108 maybe_emit_1mod(&prog, dst_reg, 1109 BPF_CLASS(insn->code) == BPF_ALU64); 1110 1111 b3 = simple_alu_opcodes[BPF_OP(insn->code)]; 1112 if (imm32 == 1) 1113 EMIT2(0xD1, add_1reg(b3, dst_reg)); 1114 else 1115 EMIT3(0xC1, add_1reg(b3, dst_reg), imm32); 1116 break; 1117 1118 case BPF_ALU | BPF_LSH | BPF_X: 1119 case BPF_ALU | BPF_RSH | BPF_X: 1120 case BPF_ALU | BPF_ARSH | BPF_X: 1121 case BPF_ALU64 | BPF_LSH | BPF_X: 1122 case BPF_ALU64 | BPF_RSH | BPF_X: 1123 case BPF_ALU64 | BPF_ARSH | BPF_X: 1124 1125 /* Check for bad case when dst_reg == rcx */ 1126 if (dst_reg == BPF_REG_4) { 1127 /* mov r11, dst_reg */ 1128 EMIT_mov(AUX_REG, dst_reg); 1129 dst_reg = AUX_REG; 1130 } 1131 1132 if (src_reg != BPF_REG_4) { /* common case */ 1133 EMIT1(0x51); /* push rcx */ 1134 1135 /* mov rcx, src_reg */ 1136 EMIT_mov(BPF_REG_4, src_reg); 1137 } 1138 1139 /* shl %rax, %cl | shr %rax, %cl | sar %rax, %cl */ 1140 maybe_emit_1mod(&prog, dst_reg, 1141 BPF_CLASS(insn->code) == BPF_ALU64); 1142 1143 b3 = simple_alu_opcodes[BPF_OP(insn->code)]; 1144 EMIT2(0xD3, add_1reg(b3, dst_reg)); 1145 1146 if (src_reg != BPF_REG_4) 1147 EMIT1(0x59); /* pop rcx */ 1148 1149 if (insn->dst_reg == BPF_REG_4) 1150 /* mov dst_reg, r11 */ 1151 EMIT_mov(insn->dst_reg, AUX_REG); 1152 break; 1153 1154 case BPF_ALU | BPF_END | BPF_FROM_BE: 1155 switch (imm32) { 1156 case 16: 1157 /* Emit 'ror %ax, 8' to swap lower 2 bytes */ 1158 EMIT1(0x66); 1159 if (is_ereg(dst_reg)) 1160 EMIT1(0x41); 1161 EMIT3(0xC1, add_1reg(0xC8, dst_reg), 8); 1162 1163 /* Emit 'movzwl eax, ax' */ 1164 if (is_ereg(dst_reg)) 1165 EMIT3(0x45, 0x0F, 0xB7); 1166 else 1167 EMIT2(0x0F, 0xB7); 1168 EMIT1(add_2reg(0xC0, dst_reg, dst_reg)); 1169 break; 1170 case 32: 1171 /* Emit 'bswap eax' to swap lower 4 bytes */ 1172 if (is_ereg(dst_reg)) 1173 EMIT2(0x41, 0x0F); 1174 else 1175 EMIT1(0x0F); 1176 EMIT1(add_1reg(0xC8, dst_reg)); 1177 break; 1178 case 64: 1179 /* Emit 'bswap rax' to swap 8 bytes */ 1180 EMIT3(add_1mod(0x48, dst_reg), 0x0F, 1181 add_1reg(0xC8, dst_reg)); 1182 break; 1183 } 1184 break; 1185 1186 case BPF_ALU | BPF_END | BPF_FROM_LE: 1187 switch (imm32) { 1188 case 16: 1189 /* 1190 * Emit 'movzwl eax, ax' to zero extend 16-bit 1191 * into 64 bit 1192 */ 1193 if (is_ereg(dst_reg)) 1194 EMIT3(0x45, 0x0F, 0xB7); 1195 else 1196 EMIT2(0x0F, 0xB7); 1197 EMIT1(add_2reg(0xC0, dst_reg, dst_reg)); 1198 break; 1199 case 32: 1200 /* Emit 'mov eax, eax' to clear upper 32-bits */ 1201 if (is_ereg(dst_reg)) 1202 EMIT1(0x45); 1203 EMIT2(0x89, add_2reg(0xC0, dst_reg, dst_reg)); 1204 break; 1205 case 64: 1206 /* nop */ 1207 break; 1208 } 1209 break; 1210 1211 /* speculation barrier */ 1212 case BPF_ST | BPF_NOSPEC: 1213 if (boot_cpu_has(X86_FEATURE_XMM2)) 1214 EMIT_LFENCE(); 1215 break; 1216 1217 /* ST: *(u8*)(dst_reg + off) = imm */ 1218 case BPF_ST | BPF_MEM | BPF_B: 1219 if (is_ereg(dst_reg)) 1220 EMIT2(0x41, 0xC6); 1221 else 1222 EMIT1(0xC6); 1223 goto st; 1224 case BPF_ST | BPF_MEM | BPF_H: 1225 if (is_ereg(dst_reg)) 1226 EMIT3(0x66, 0x41, 0xC7); 1227 else 1228 EMIT2(0x66, 0xC7); 1229 goto st; 1230 case BPF_ST | BPF_MEM | BPF_W: 1231 if (is_ereg(dst_reg)) 1232 EMIT2(0x41, 0xC7); 1233 else 1234 EMIT1(0xC7); 1235 goto st; 1236 case BPF_ST | BPF_MEM | BPF_DW: 1237 EMIT2(add_1mod(0x48, dst_reg), 0xC7); 1238 1239 st: if (is_imm8(insn->off)) 1240 EMIT2(add_1reg(0x40, dst_reg), insn->off); 1241 else 1242 EMIT1_off32(add_1reg(0x80, dst_reg), insn->off); 1243 1244 EMIT(imm32, bpf_size_to_x86_bytes(BPF_SIZE(insn->code))); 1245 break; 1246 1247 /* STX: *(u8*)(dst_reg + off) = src_reg */ 1248 case BPF_STX | BPF_MEM | BPF_B: 1249 case BPF_STX | BPF_MEM | BPF_H: 1250 case BPF_STX | BPF_MEM | BPF_W: 1251 case BPF_STX | BPF_MEM | BPF_DW: 1252 emit_stx(&prog, BPF_SIZE(insn->code), dst_reg, src_reg, insn->off); 1253 break; 1254 1255 /* LDX: dst_reg = *(u8*)(src_reg + off) */ 1256 case BPF_LDX | BPF_MEM | BPF_B: 1257 case BPF_LDX | BPF_PROBE_MEM | BPF_B: 1258 case BPF_LDX | BPF_MEM | BPF_H: 1259 case BPF_LDX | BPF_PROBE_MEM | BPF_H: 1260 case BPF_LDX | BPF_MEM | BPF_W: 1261 case BPF_LDX | BPF_PROBE_MEM | BPF_W: 1262 case BPF_LDX | BPF_MEM | BPF_DW: 1263 case BPF_LDX | BPF_PROBE_MEM | BPF_DW: 1264 if (BPF_MODE(insn->code) == BPF_PROBE_MEM) { 1265 /* Though the verifier prevents negative insn->off in BPF_PROBE_MEM 1266 * add abs(insn->off) to the limit to make sure that negative 1267 * offset won't be an issue. 1268 * insn->off is s16, so it won't affect valid pointers. 1269 */ 1270 u64 limit = TASK_SIZE_MAX + PAGE_SIZE + abs(insn->off); 1271 u8 *end_of_jmp1, *end_of_jmp2; 1272 1273 /* Conservatively check that src_reg + insn->off is a kernel address: 1274 * 1. src_reg + insn->off >= limit 1275 * 2. src_reg + insn->off doesn't become small positive. 1276 * Cannot do src_reg + insn->off >= limit in one branch, 1277 * since it needs two spare registers, but JIT has only one. 1278 */ 1279 1280 /* movabsq r11, limit */ 1281 EMIT2(add_1mod(0x48, AUX_REG), add_1reg(0xB8, AUX_REG)); 1282 EMIT((u32)limit, 4); 1283 EMIT(limit >> 32, 4); 1284 /* cmp src_reg, r11 */ 1285 maybe_emit_mod(&prog, src_reg, AUX_REG, true); 1286 EMIT2(0x39, add_2reg(0xC0, src_reg, AUX_REG)); 1287 /* if unsigned '<' goto end_of_jmp2 */ 1288 EMIT2(X86_JB, 0); 1289 end_of_jmp1 = prog; 1290 1291 /* mov r11, src_reg */ 1292 emit_mov_reg(&prog, true, AUX_REG, src_reg); 1293 /* add r11, insn->off */ 1294 maybe_emit_1mod(&prog, AUX_REG, true); 1295 EMIT2_off32(0x81, add_1reg(0xC0, AUX_REG), insn->off); 1296 /* jmp if not carry to start_of_ldx 1297 * Otherwise ERR_PTR(-EINVAL) + 128 will be the user addr 1298 * that has to be rejected. 1299 */ 1300 EMIT2(0x73 /* JNC */, 0); 1301 end_of_jmp2 = prog; 1302 1303 /* xor dst_reg, dst_reg */ 1304 emit_mov_imm32(&prog, false, dst_reg, 0); 1305 /* jmp byte_after_ldx */ 1306 EMIT2(0xEB, 0); 1307 1308 /* populate jmp_offset for JB above to jump to xor dst_reg */ 1309 end_of_jmp1[-1] = end_of_jmp2 - end_of_jmp1; 1310 /* populate jmp_offset for JNC above to jump to start_of_ldx */ 1311 start_of_ldx = prog; 1312 end_of_jmp2[-1] = start_of_ldx - end_of_jmp2; 1313 } 1314 emit_ldx(&prog, BPF_SIZE(insn->code), dst_reg, src_reg, insn->off); 1315 if (BPF_MODE(insn->code) == BPF_PROBE_MEM) { 1316 struct exception_table_entry *ex; 1317 u8 *_insn = image + proglen + (start_of_ldx - temp); 1318 s64 delta; 1319 1320 /* populate jmp_offset for JMP above */ 1321 start_of_ldx[-1] = prog - start_of_ldx; 1322 1323 if (!bpf_prog->aux->extable) 1324 break; 1325 1326 if (excnt >= bpf_prog->aux->num_exentries) { 1327 pr_err("ex gen bug\n"); 1328 return -EFAULT; 1329 } 1330 ex = &bpf_prog->aux->extable[excnt++]; 1331 1332 delta = _insn - (u8 *)&ex->insn; 1333 if (!is_simm32(delta)) { 1334 pr_err("extable->insn doesn't fit into 32-bit\n"); 1335 return -EFAULT; 1336 } 1337 /* switch ex to rw buffer for writes */ 1338 ex = (void *)rw_image + ((void *)ex - (void *)image); 1339 1340 ex->insn = delta; 1341 1342 ex->data = EX_TYPE_BPF; 1343 1344 if (dst_reg > BPF_REG_9) { 1345 pr_err("verifier error\n"); 1346 return -EFAULT; 1347 } 1348 /* 1349 * Compute size of x86 insn and its target dest x86 register. 1350 * ex_handler_bpf() will use lower 8 bits to adjust 1351 * pt_regs->ip to jump over this x86 instruction 1352 * and upper bits to figure out which pt_regs to zero out. 1353 * End result: x86 insn "mov rbx, qword ptr [rax+0x14]" 1354 * of 4 bytes will be ignored and rbx will be zero inited. 1355 */ 1356 ex->fixup = (prog - start_of_ldx) | (reg2pt_regs[dst_reg] << 8); 1357 } 1358 break; 1359 1360 case BPF_STX | BPF_ATOMIC | BPF_W: 1361 case BPF_STX | BPF_ATOMIC | BPF_DW: 1362 if (insn->imm == (BPF_AND | BPF_FETCH) || 1363 insn->imm == (BPF_OR | BPF_FETCH) || 1364 insn->imm == (BPF_XOR | BPF_FETCH)) { 1365 bool is64 = BPF_SIZE(insn->code) == BPF_DW; 1366 u32 real_src_reg = src_reg; 1367 u32 real_dst_reg = dst_reg; 1368 u8 *branch_target; 1369 1370 /* 1371 * Can't be implemented with a single x86 insn. 1372 * Need to do a CMPXCHG loop. 1373 */ 1374 1375 /* Will need RAX as a CMPXCHG operand so save R0 */ 1376 emit_mov_reg(&prog, true, BPF_REG_AX, BPF_REG_0); 1377 if (src_reg == BPF_REG_0) 1378 real_src_reg = BPF_REG_AX; 1379 if (dst_reg == BPF_REG_0) 1380 real_dst_reg = BPF_REG_AX; 1381 1382 branch_target = prog; 1383 /* Load old value */ 1384 emit_ldx(&prog, BPF_SIZE(insn->code), 1385 BPF_REG_0, real_dst_reg, insn->off); 1386 /* 1387 * Perform the (commutative) operation locally, 1388 * put the result in the AUX_REG. 1389 */ 1390 emit_mov_reg(&prog, is64, AUX_REG, BPF_REG_0); 1391 maybe_emit_mod(&prog, AUX_REG, real_src_reg, is64); 1392 EMIT2(simple_alu_opcodes[BPF_OP(insn->imm)], 1393 add_2reg(0xC0, AUX_REG, real_src_reg)); 1394 /* Attempt to swap in new value */ 1395 err = emit_atomic(&prog, BPF_CMPXCHG, 1396 real_dst_reg, AUX_REG, 1397 insn->off, 1398 BPF_SIZE(insn->code)); 1399 if (WARN_ON(err)) 1400 return err; 1401 /* 1402 * ZF tells us whether we won the race. If it's 1403 * cleared we need to try again. 1404 */ 1405 EMIT2(X86_JNE, -(prog - branch_target) - 2); 1406 /* Return the pre-modification value */ 1407 emit_mov_reg(&prog, is64, real_src_reg, BPF_REG_0); 1408 /* Restore R0 after clobbering RAX */ 1409 emit_mov_reg(&prog, true, BPF_REG_0, BPF_REG_AX); 1410 break; 1411 } 1412 1413 err = emit_atomic(&prog, insn->imm, dst_reg, src_reg, 1414 insn->off, BPF_SIZE(insn->code)); 1415 if (err) 1416 return err; 1417 break; 1418 1419 /* call */ 1420 case BPF_JMP | BPF_CALL: 1421 func = (u8 *) __bpf_call_base + imm32; 1422 if (tail_call_reachable) { 1423 EMIT3_off32(0x48, 0x8B, 0x85, 1424 -(bpf_prog->aux->stack_depth + 8)); 1425 if (!imm32 || emit_call(&prog, func, image + addrs[i - 1] + 7)) 1426 return -EINVAL; 1427 } else { 1428 if (!imm32 || emit_call(&prog, func, image + addrs[i - 1])) 1429 return -EINVAL; 1430 } 1431 break; 1432 1433 case BPF_JMP | BPF_TAIL_CALL: 1434 if (imm32) 1435 emit_bpf_tail_call_direct(&bpf_prog->aux->poke_tab[imm32 - 1], 1436 &prog, image + addrs[i - 1], 1437 callee_regs_used, 1438 bpf_prog->aux->stack_depth, 1439 ctx); 1440 else 1441 emit_bpf_tail_call_indirect(&prog, 1442 callee_regs_used, 1443 bpf_prog->aux->stack_depth, 1444 image + addrs[i - 1], 1445 ctx); 1446 break; 1447 1448 /* cond jump */ 1449 case BPF_JMP | BPF_JEQ | BPF_X: 1450 case BPF_JMP | BPF_JNE | BPF_X: 1451 case BPF_JMP | BPF_JGT | BPF_X: 1452 case BPF_JMP | BPF_JLT | BPF_X: 1453 case BPF_JMP | BPF_JGE | BPF_X: 1454 case BPF_JMP | BPF_JLE | BPF_X: 1455 case BPF_JMP | BPF_JSGT | BPF_X: 1456 case BPF_JMP | BPF_JSLT | BPF_X: 1457 case BPF_JMP | BPF_JSGE | BPF_X: 1458 case BPF_JMP | BPF_JSLE | BPF_X: 1459 case BPF_JMP32 | BPF_JEQ | BPF_X: 1460 case BPF_JMP32 | BPF_JNE | BPF_X: 1461 case BPF_JMP32 | BPF_JGT | BPF_X: 1462 case BPF_JMP32 | BPF_JLT | BPF_X: 1463 case BPF_JMP32 | BPF_JGE | BPF_X: 1464 case BPF_JMP32 | BPF_JLE | BPF_X: 1465 case BPF_JMP32 | BPF_JSGT | BPF_X: 1466 case BPF_JMP32 | BPF_JSLT | BPF_X: 1467 case BPF_JMP32 | BPF_JSGE | BPF_X: 1468 case BPF_JMP32 | BPF_JSLE | BPF_X: 1469 /* cmp dst_reg, src_reg */ 1470 maybe_emit_mod(&prog, dst_reg, src_reg, 1471 BPF_CLASS(insn->code) == BPF_JMP); 1472 EMIT2(0x39, add_2reg(0xC0, dst_reg, src_reg)); 1473 goto emit_cond_jmp; 1474 1475 case BPF_JMP | BPF_JSET | BPF_X: 1476 case BPF_JMP32 | BPF_JSET | BPF_X: 1477 /* test dst_reg, src_reg */ 1478 maybe_emit_mod(&prog, dst_reg, src_reg, 1479 BPF_CLASS(insn->code) == BPF_JMP); 1480 EMIT2(0x85, add_2reg(0xC0, dst_reg, src_reg)); 1481 goto emit_cond_jmp; 1482 1483 case BPF_JMP | BPF_JSET | BPF_K: 1484 case BPF_JMP32 | BPF_JSET | BPF_K: 1485 /* test dst_reg, imm32 */ 1486 maybe_emit_1mod(&prog, dst_reg, 1487 BPF_CLASS(insn->code) == BPF_JMP); 1488 EMIT2_off32(0xF7, add_1reg(0xC0, dst_reg), imm32); 1489 goto emit_cond_jmp; 1490 1491 case BPF_JMP | BPF_JEQ | BPF_K: 1492 case BPF_JMP | BPF_JNE | BPF_K: 1493 case BPF_JMP | BPF_JGT | BPF_K: 1494 case BPF_JMP | BPF_JLT | BPF_K: 1495 case BPF_JMP | BPF_JGE | BPF_K: 1496 case BPF_JMP | BPF_JLE | BPF_K: 1497 case BPF_JMP | BPF_JSGT | BPF_K: 1498 case BPF_JMP | BPF_JSLT | BPF_K: 1499 case BPF_JMP | BPF_JSGE | BPF_K: 1500 case BPF_JMP | BPF_JSLE | BPF_K: 1501 case BPF_JMP32 | BPF_JEQ | BPF_K: 1502 case BPF_JMP32 | BPF_JNE | BPF_K: 1503 case BPF_JMP32 | BPF_JGT | BPF_K: 1504 case BPF_JMP32 | BPF_JLT | BPF_K: 1505 case BPF_JMP32 | BPF_JGE | BPF_K: 1506 case BPF_JMP32 | BPF_JLE | BPF_K: 1507 case BPF_JMP32 | BPF_JSGT | BPF_K: 1508 case BPF_JMP32 | BPF_JSLT | BPF_K: 1509 case BPF_JMP32 | BPF_JSGE | BPF_K: 1510 case BPF_JMP32 | BPF_JSLE | BPF_K: 1511 /* test dst_reg, dst_reg to save one extra byte */ 1512 if (imm32 == 0) { 1513 maybe_emit_mod(&prog, dst_reg, dst_reg, 1514 BPF_CLASS(insn->code) == BPF_JMP); 1515 EMIT2(0x85, add_2reg(0xC0, dst_reg, dst_reg)); 1516 goto emit_cond_jmp; 1517 } 1518 1519 /* cmp dst_reg, imm8/32 */ 1520 maybe_emit_1mod(&prog, dst_reg, 1521 BPF_CLASS(insn->code) == BPF_JMP); 1522 1523 if (is_imm8(imm32)) 1524 EMIT3(0x83, add_1reg(0xF8, dst_reg), imm32); 1525 else 1526 EMIT2_off32(0x81, add_1reg(0xF8, dst_reg), imm32); 1527 1528 emit_cond_jmp: /* Convert BPF opcode to x86 */ 1529 switch (BPF_OP(insn->code)) { 1530 case BPF_JEQ: 1531 jmp_cond = X86_JE; 1532 break; 1533 case BPF_JSET: 1534 case BPF_JNE: 1535 jmp_cond = X86_JNE; 1536 break; 1537 case BPF_JGT: 1538 /* GT is unsigned '>', JA in x86 */ 1539 jmp_cond = X86_JA; 1540 break; 1541 case BPF_JLT: 1542 /* LT is unsigned '<', JB in x86 */ 1543 jmp_cond = X86_JB; 1544 break; 1545 case BPF_JGE: 1546 /* GE is unsigned '>=', JAE in x86 */ 1547 jmp_cond = X86_JAE; 1548 break; 1549 case BPF_JLE: 1550 /* LE is unsigned '<=', JBE in x86 */ 1551 jmp_cond = X86_JBE; 1552 break; 1553 case BPF_JSGT: 1554 /* Signed '>', GT in x86 */ 1555 jmp_cond = X86_JG; 1556 break; 1557 case BPF_JSLT: 1558 /* Signed '<', LT in x86 */ 1559 jmp_cond = X86_JL; 1560 break; 1561 case BPF_JSGE: 1562 /* Signed '>=', GE in x86 */ 1563 jmp_cond = X86_JGE; 1564 break; 1565 case BPF_JSLE: 1566 /* Signed '<=', LE in x86 */ 1567 jmp_cond = X86_JLE; 1568 break; 1569 default: /* to silence GCC warning */ 1570 return -EFAULT; 1571 } 1572 jmp_offset = addrs[i + insn->off] - addrs[i]; 1573 if (is_imm8(jmp_offset)) { 1574 if (jmp_padding) { 1575 /* To keep the jmp_offset valid, the extra bytes are 1576 * padded before the jump insn, so we subtract the 1577 * 2 bytes of jmp_cond insn from INSN_SZ_DIFF. 1578 * 1579 * If the previous pass already emits an imm8 1580 * jmp_cond, then this BPF insn won't shrink, so 1581 * "nops" is 0. 1582 * 1583 * On the other hand, if the previous pass emits an 1584 * imm32 jmp_cond, the extra 4 bytes(*) is padded to 1585 * keep the image from shrinking further. 1586 * 1587 * (*) imm32 jmp_cond is 6 bytes, and imm8 jmp_cond 1588 * is 2 bytes, so the size difference is 4 bytes. 1589 */ 1590 nops = INSN_SZ_DIFF - 2; 1591 if (nops != 0 && nops != 4) { 1592 pr_err("unexpected jmp_cond padding: %d bytes\n", 1593 nops); 1594 return -EFAULT; 1595 } 1596 emit_nops(&prog, nops); 1597 } 1598 EMIT2(jmp_cond, jmp_offset); 1599 } else if (is_simm32(jmp_offset)) { 1600 EMIT2_off32(0x0F, jmp_cond + 0x10, jmp_offset); 1601 } else { 1602 pr_err("cond_jmp gen bug %llx\n", jmp_offset); 1603 return -EFAULT; 1604 } 1605 1606 break; 1607 1608 case BPF_JMP | BPF_JA: 1609 if (insn->off == -1) 1610 /* -1 jmp instructions will always jump 1611 * backwards two bytes. Explicitly handling 1612 * this case avoids wasting too many passes 1613 * when there are long sequences of replaced 1614 * dead code. 1615 */ 1616 jmp_offset = -2; 1617 else 1618 jmp_offset = addrs[i + insn->off] - addrs[i]; 1619 1620 if (!jmp_offset) { 1621 /* 1622 * If jmp_padding is enabled, the extra nops will 1623 * be inserted. Otherwise, optimize out nop jumps. 1624 */ 1625 if (jmp_padding) { 1626 /* There are 3 possible conditions. 1627 * (1) This BPF_JA is already optimized out in 1628 * the previous run, so there is no need 1629 * to pad any extra byte (0 byte). 1630 * (2) The previous pass emits an imm8 jmp, 1631 * so we pad 2 bytes to match the previous 1632 * insn size. 1633 * (3) Similarly, the previous pass emits an 1634 * imm32 jmp, and 5 bytes is padded. 1635 */ 1636 nops = INSN_SZ_DIFF; 1637 if (nops != 0 && nops != 2 && nops != 5) { 1638 pr_err("unexpected nop jump padding: %d bytes\n", 1639 nops); 1640 return -EFAULT; 1641 } 1642 emit_nops(&prog, nops); 1643 } 1644 break; 1645 } 1646 emit_jmp: 1647 if (is_imm8(jmp_offset)) { 1648 if (jmp_padding) { 1649 /* To avoid breaking jmp_offset, the extra bytes 1650 * are padded before the actual jmp insn, so 1651 * 2 bytes is subtracted from INSN_SZ_DIFF. 1652 * 1653 * If the previous pass already emits an imm8 1654 * jmp, there is nothing to pad (0 byte). 1655 * 1656 * If it emits an imm32 jmp (5 bytes) previously 1657 * and now an imm8 jmp (2 bytes), then we pad 1658 * (5 - 2 = 3) bytes to stop the image from 1659 * shrinking further. 1660 */ 1661 nops = INSN_SZ_DIFF - 2; 1662 if (nops != 0 && nops != 3) { 1663 pr_err("unexpected jump padding: %d bytes\n", 1664 nops); 1665 return -EFAULT; 1666 } 1667 emit_nops(&prog, INSN_SZ_DIFF - 2); 1668 } 1669 EMIT2(0xEB, jmp_offset); 1670 } else if (is_simm32(jmp_offset)) { 1671 EMIT1_off32(0xE9, jmp_offset); 1672 } else { 1673 pr_err("jmp gen bug %llx\n", jmp_offset); 1674 return -EFAULT; 1675 } 1676 break; 1677 1678 case BPF_JMP | BPF_EXIT: 1679 if (seen_exit) { 1680 jmp_offset = ctx->cleanup_addr - addrs[i]; 1681 goto emit_jmp; 1682 } 1683 seen_exit = true; 1684 /* Update cleanup_addr */ 1685 ctx->cleanup_addr = proglen; 1686 pop_callee_regs(&prog, callee_regs_used); 1687 EMIT1(0xC9); /* leave */ 1688 EMIT1(0xC3); /* ret */ 1689 break; 1690 1691 default: 1692 /* 1693 * By design x86-64 JIT should support all BPF instructions. 1694 * This error will be seen if new instruction was added 1695 * to the interpreter, but not to the JIT, or if there is 1696 * junk in bpf_prog. 1697 */ 1698 pr_err("bpf_jit: unknown opcode %02x\n", insn->code); 1699 return -EINVAL; 1700 } 1701 1702 ilen = prog - temp; 1703 if (ilen > BPF_MAX_INSN_SIZE) { 1704 pr_err("bpf_jit: fatal insn size error\n"); 1705 return -EFAULT; 1706 } 1707 1708 if (image) { 1709 /* 1710 * When populating the image, assert that: 1711 * 1712 * i) We do not write beyond the allocated space, and 1713 * ii) addrs[i] did not change from the prior run, in order 1714 * to validate assumptions made for computing branch 1715 * displacements. 1716 */ 1717 if (unlikely(proglen + ilen > oldproglen || 1718 proglen + ilen != addrs[i])) { 1719 pr_err("bpf_jit: fatal error\n"); 1720 return -EFAULT; 1721 } 1722 memcpy(rw_image + proglen, temp, ilen); 1723 } 1724 proglen += ilen; 1725 addrs[i] = proglen; 1726 prog = temp; 1727 } 1728 1729 if (image && excnt != bpf_prog->aux->num_exentries) { 1730 pr_err("extable is not populated\n"); 1731 return -EFAULT; 1732 } 1733 return proglen; 1734 } 1735 1736 static void save_regs(const struct btf_func_model *m, u8 **prog, int nr_args, 1737 int stack_size) 1738 { 1739 int i; 1740 /* Store function arguments to stack. 1741 * For a function that accepts two pointers the sequence will be: 1742 * mov QWORD PTR [rbp-0x10],rdi 1743 * mov QWORD PTR [rbp-0x8],rsi 1744 */ 1745 for (i = 0; i < min(nr_args, 6); i++) 1746 emit_stx(prog, bytes_to_bpf_size(m->arg_size[i]), 1747 BPF_REG_FP, 1748 i == 5 ? X86_REG_R9 : BPF_REG_1 + i, 1749 -(stack_size - i * 8)); 1750 } 1751 1752 static void restore_regs(const struct btf_func_model *m, u8 **prog, int nr_args, 1753 int stack_size) 1754 { 1755 int i; 1756 1757 /* Restore function arguments from stack. 1758 * For a function that accepts two pointers the sequence will be: 1759 * EMIT4(0x48, 0x8B, 0x7D, 0xF0); mov rdi,QWORD PTR [rbp-0x10] 1760 * EMIT4(0x48, 0x8B, 0x75, 0xF8); mov rsi,QWORD PTR [rbp-0x8] 1761 */ 1762 for (i = 0; i < min(nr_args, 6); i++) 1763 emit_ldx(prog, bytes_to_bpf_size(m->arg_size[i]), 1764 i == 5 ? X86_REG_R9 : BPF_REG_1 + i, 1765 BPF_REG_FP, 1766 -(stack_size - i * 8)); 1767 } 1768 1769 static int invoke_bpf_prog(const struct btf_func_model *m, u8 **pprog, 1770 struct bpf_tramp_link *l, int stack_size, 1771 int run_ctx_off, bool save_ret) 1772 { 1773 void (*exit)(struct bpf_prog *prog, u64 start, 1774 struct bpf_tramp_run_ctx *run_ctx) = __bpf_prog_exit; 1775 u64 (*enter)(struct bpf_prog *prog, 1776 struct bpf_tramp_run_ctx *run_ctx) = __bpf_prog_enter; 1777 u8 *prog = *pprog; 1778 u8 *jmp_insn; 1779 int ctx_cookie_off = offsetof(struct bpf_tramp_run_ctx, bpf_cookie); 1780 struct bpf_prog *p = l->link.prog; 1781 u64 cookie = l->cookie; 1782 1783 /* mov rdi, cookie */ 1784 emit_mov_imm64(&prog, BPF_REG_1, (long) cookie >> 32, (u32) (long) cookie); 1785 1786 /* Prepare struct bpf_tramp_run_ctx. 1787 * 1788 * bpf_tramp_run_ctx is already preserved by 1789 * arch_prepare_bpf_trampoline(). 1790 * 1791 * mov QWORD PTR [rbp - run_ctx_off + ctx_cookie_off], rdi 1792 */ 1793 emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_1, -run_ctx_off + ctx_cookie_off); 1794 1795 if (p->aux->sleepable) { 1796 enter = __bpf_prog_enter_sleepable; 1797 exit = __bpf_prog_exit_sleepable; 1798 } else if (p->expected_attach_type == BPF_LSM_CGROUP) { 1799 enter = __bpf_prog_enter_lsm_cgroup; 1800 exit = __bpf_prog_exit_lsm_cgroup; 1801 } 1802 1803 /* arg1: mov rdi, progs[i] */ 1804 emit_mov_imm64(&prog, BPF_REG_1, (long) p >> 32, (u32) (long) p); 1805 /* arg2: lea rsi, [rbp - ctx_cookie_off] */ 1806 EMIT4(0x48, 0x8D, 0x75, -run_ctx_off); 1807 1808 if (emit_call(&prog, enter, prog)) 1809 return -EINVAL; 1810 /* remember prog start time returned by __bpf_prog_enter */ 1811 emit_mov_reg(&prog, true, BPF_REG_6, BPF_REG_0); 1812 1813 /* if (__bpf_prog_enter*(prog) == 0) 1814 * goto skip_exec_of_prog; 1815 */ 1816 EMIT3(0x48, 0x85, 0xC0); /* test rax,rax */ 1817 /* emit 2 nops that will be replaced with JE insn */ 1818 jmp_insn = prog; 1819 emit_nops(&prog, 2); 1820 1821 /* arg1: lea rdi, [rbp - stack_size] */ 1822 EMIT4(0x48, 0x8D, 0x7D, -stack_size); 1823 /* arg2: progs[i]->insnsi for interpreter */ 1824 if (!p->jited) 1825 emit_mov_imm64(&prog, BPF_REG_2, 1826 (long) p->insnsi >> 32, 1827 (u32) (long) p->insnsi); 1828 /* call JITed bpf program or interpreter */ 1829 if (emit_call(&prog, p->bpf_func, prog)) 1830 return -EINVAL; 1831 1832 /* 1833 * BPF_TRAMP_MODIFY_RETURN trampolines can modify the return 1834 * of the previous call which is then passed on the stack to 1835 * the next BPF program. 1836 * 1837 * BPF_TRAMP_FENTRY trampoline may need to return the return 1838 * value of BPF_PROG_TYPE_STRUCT_OPS prog. 1839 */ 1840 if (save_ret) 1841 emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_0, -8); 1842 1843 /* replace 2 nops with JE insn, since jmp target is known */ 1844 jmp_insn[0] = X86_JE; 1845 jmp_insn[1] = prog - jmp_insn - 2; 1846 1847 /* arg1: mov rdi, progs[i] */ 1848 emit_mov_imm64(&prog, BPF_REG_1, (long) p >> 32, (u32) (long) p); 1849 /* arg2: mov rsi, rbx <- start time in nsec */ 1850 emit_mov_reg(&prog, true, BPF_REG_2, BPF_REG_6); 1851 /* arg3: lea rdx, [rbp - run_ctx_off] */ 1852 EMIT4(0x48, 0x8D, 0x55, -run_ctx_off); 1853 if (emit_call(&prog, exit, prog)) 1854 return -EINVAL; 1855 1856 *pprog = prog; 1857 return 0; 1858 } 1859 1860 static void emit_align(u8 **pprog, u32 align) 1861 { 1862 u8 *target, *prog = *pprog; 1863 1864 target = PTR_ALIGN(prog, align); 1865 if (target != prog) 1866 emit_nops(&prog, target - prog); 1867 1868 *pprog = prog; 1869 } 1870 1871 static int emit_cond_near_jump(u8 **pprog, void *func, void *ip, u8 jmp_cond) 1872 { 1873 u8 *prog = *pprog; 1874 s64 offset; 1875 1876 offset = func - (ip + 2 + 4); 1877 if (!is_simm32(offset)) { 1878 pr_err("Target %p is out of range\n", func); 1879 return -EINVAL; 1880 } 1881 EMIT2_off32(0x0F, jmp_cond + 0x10, offset); 1882 *pprog = prog; 1883 return 0; 1884 } 1885 1886 static int invoke_bpf(const struct btf_func_model *m, u8 **pprog, 1887 struct bpf_tramp_links *tl, int stack_size, 1888 int run_ctx_off, bool save_ret) 1889 { 1890 int i; 1891 u8 *prog = *pprog; 1892 1893 for (i = 0; i < tl->nr_links; i++) { 1894 if (invoke_bpf_prog(m, &prog, tl->links[i], stack_size, 1895 run_ctx_off, save_ret)) 1896 return -EINVAL; 1897 } 1898 *pprog = prog; 1899 return 0; 1900 } 1901 1902 static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog, 1903 struct bpf_tramp_links *tl, int stack_size, 1904 int run_ctx_off, u8 **branches) 1905 { 1906 u8 *prog = *pprog; 1907 int i; 1908 1909 /* The first fmod_ret program will receive a garbage return value. 1910 * Set this to 0 to avoid confusing the program. 1911 */ 1912 emit_mov_imm32(&prog, false, BPF_REG_0, 0); 1913 emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_0, -8); 1914 for (i = 0; i < tl->nr_links; i++) { 1915 if (invoke_bpf_prog(m, &prog, tl->links[i], stack_size, run_ctx_off, true)) 1916 return -EINVAL; 1917 1918 /* mod_ret prog stored return value into [rbp - 8]. Emit: 1919 * if (*(u64 *)(rbp - 8) != 0) 1920 * goto do_fexit; 1921 */ 1922 /* cmp QWORD PTR [rbp - 0x8], 0x0 */ 1923 EMIT4(0x48, 0x83, 0x7d, 0xf8); EMIT1(0x00); 1924 1925 /* Save the location of the branch and Generate 6 nops 1926 * (4 bytes for an offset and 2 bytes for the jump) These nops 1927 * are replaced with a conditional jump once do_fexit (i.e. the 1928 * start of the fexit invocation) is finalized. 1929 */ 1930 branches[i] = prog; 1931 emit_nops(&prog, 4 + 2); 1932 } 1933 1934 *pprog = prog; 1935 return 0; 1936 } 1937 1938 static bool is_valid_bpf_tramp_flags(unsigned int flags) 1939 { 1940 if ((flags & BPF_TRAMP_F_RESTORE_REGS) && 1941 (flags & BPF_TRAMP_F_SKIP_FRAME)) 1942 return false; 1943 1944 /* 1945 * BPF_TRAMP_F_RET_FENTRY_RET is only used by bpf_struct_ops, 1946 * and it must be used alone. 1947 */ 1948 if ((flags & BPF_TRAMP_F_RET_FENTRY_RET) && 1949 (flags & ~BPF_TRAMP_F_RET_FENTRY_RET)) 1950 return false; 1951 1952 return true; 1953 } 1954 1955 /* Example: 1956 * __be16 eth_type_trans(struct sk_buff *skb, struct net_device *dev); 1957 * its 'struct btf_func_model' will be nr_args=2 1958 * The assembly code when eth_type_trans is executing after trampoline: 1959 * 1960 * push rbp 1961 * mov rbp, rsp 1962 * sub rsp, 16 // space for skb and dev 1963 * push rbx // temp regs to pass start time 1964 * mov qword ptr [rbp - 16], rdi // save skb pointer to stack 1965 * mov qword ptr [rbp - 8], rsi // save dev pointer to stack 1966 * call __bpf_prog_enter // rcu_read_lock and preempt_disable 1967 * mov rbx, rax // remember start time in bpf stats are enabled 1968 * lea rdi, [rbp - 16] // R1==ctx of bpf prog 1969 * call addr_of_jited_FENTRY_prog 1970 * movabsq rdi, 64bit_addr_of_struct_bpf_prog // unused if bpf stats are off 1971 * mov rsi, rbx // prog start time 1972 * call __bpf_prog_exit // rcu_read_unlock, preempt_enable and stats math 1973 * mov rdi, qword ptr [rbp - 16] // restore skb pointer from stack 1974 * mov rsi, qword ptr [rbp - 8] // restore dev pointer from stack 1975 * pop rbx 1976 * leave 1977 * ret 1978 * 1979 * eth_type_trans has 5 byte nop at the beginning. These 5 bytes will be 1980 * replaced with 'call generated_bpf_trampoline'. When it returns 1981 * eth_type_trans will continue executing with original skb and dev pointers. 1982 * 1983 * The assembly code when eth_type_trans is called from trampoline: 1984 * 1985 * push rbp 1986 * mov rbp, rsp 1987 * sub rsp, 24 // space for skb, dev, return value 1988 * push rbx // temp regs to pass start time 1989 * mov qword ptr [rbp - 24], rdi // save skb pointer to stack 1990 * mov qword ptr [rbp - 16], rsi // save dev pointer to stack 1991 * call __bpf_prog_enter // rcu_read_lock and preempt_disable 1992 * mov rbx, rax // remember start time if bpf stats are enabled 1993 * lea rdi, [rbp - 24] // R1==ctx of bpf prog 1994 * call addr_of_jited_FENTRY_prog // bpf prog can access skb and dev 1995 * movabsq rdi, 64bit_addr_of_struct_bpf_prog // unused if bpf stats are off 1996 * mov rsi, rbx // prog start time 1997 * call __bpf_prog_exit // rcu_read_unlock, preempt_enable and stats math 1998 * mov rdi, qword ptr [rbp - 24] // restore skb pointer from stack 1999 * mov rsi, qword ptr [rbp - 16] // restore dev pointer from stack 2000 * call eth_type_trans+5 // execute body of eth_type_trans 2001 * mov qword ptr [rbp - 8], rax // save return value 2002 * call __bpf_prog_enter // rcu_read_lock and preempt_disable 2003 * mov rbx, rax // remember start time in bpf stats are enabled 2004 * lea rdi, [rbp - 24] // R1==ctx of bpf prog 2005 * call addr_of_jited_FEXIT_prog // bpf prog can access skb, dev, return value 2006 * movabsq rdi, 64bit_addr_of_struct_bpf_prog // unused if bpf stats are off 2007 * mov rsi, rbx // prog start time 2008 * call __bpf_prog_exit // rcu_read_unlock, preempt_enable and stats math 2009 * mov rax, qword ptr [rbp - 8] // restore eth_type_trans's return value 2010 * pop rbx 2011 * leave 2012 * add rsp, 8 // skip eth_type_trans's frame 2013 * ret // return to its caller 2014 */ 2015 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *image_end, 2016 const struct btf_func_model *m, u32 flags, 2017 struct bpf_tramp_links *tlinks, 2018 void *orig_call) 2019 { 2020 int ret, i, nr_args = m->nr_args; 2021 int regs_off, ip_off, args_off, stack_size = nr_args * 8, run_ctx_off; 2022 struct bpf_tramp_links *fentry = &tlinks[BPF_TRAMP_FENTRY]; 2023 struct bpf_tramp_links *fexit = &tlinks[BPF_TRAMP_FEXIT]; 2024 struct bpf_tramp_links *fmod_ret = &tlinks[BPF_TRAMP_MODIFY_RETURN]; 2025 u8 **branches = NULL; 2026 u8 *prog; 2027 bool save_ret; 2028 2029 /* x86-64 supports up to 6 arguments. 7+ can be added in the future */ 2030 if (nr_args > 6) 2031 return -ENOTSUPP; 2032 2033 if (!is_valid_bpf_tramp_flags(flags)) 2034 return -EINVAL; 2035 2036 /* Generated trampoline stack layout: 2037 * 2038 * RBP + 8 [ return address ] 2039 * RBP + 0 [ RBP ] 2040 * 2041 * RBP - 8 [ return value ] BPF_TRAMP_F_CALL_ORIG or 2042 * BPF_TRAMP_F_RET_FENTRY_RET flags 2043 * 2044 * [ reg_argN ] always 2045 * [ ... ] 2046 * RBP - regs_off [ reg_arg1 ] program's ctx pointer 2047 * 2048 * RBP - args_off [ args count ] always 2049 * 2050 * RBP - ip_off [ traced function ] BPF_TRAMP_F_IP_ARG flag 2051 * 2052 * RBP - run_ctx_off [ bpf_tramp_run_ctx ] 2053 */ 2054 2055 /* room for return value of orig_call or fentry prog */ 2056 save_ret = flags & (BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_RET_FENTRY_RET); 2057 if (save_ret) 2058 stack_size += 8; 2059 2060 regs_off = stack_size; 2061 2062 /* args count */ 2063 stack_size += 8; 2064 args_off = stack_size; 2065 2066 if (flags & BPF_TRAMP_F_IP_ARG) 2067 stack_size += 8; /* room for IP address argument */ 2068 2069 ip_off = stack_size; 2070 2071 stack_size += (sizeof(struct bpf_tramp_run_ctx) + 7) & ~0x7; 2072 run_ctx_off = stack_size; 2073 2074 if (flags & BPF_TRAMP_F_SKIP_FRAME) { 2075 /* skip patched call instruction and point orig_call to actual 2076 * body of the kernel function. 2077 */ 2078 if (is_endbr(*(u32 *)orig_call)) 2079 orig_call += ENDBR_INSN_SIZE; 2080 orig_call += X86_PATCH_SIZE; 2081 } 2082 2083 prog = image; 2084 2085 EMIT_ENDBR(); 2086 EMIT1(0x55); /* push rbp */ 2087 EMIT3(0x48, 0x89, 0xE5); /* mov rbp, rsp */ 2088 EMIT4(0x48, 0x83, 0xEC, stack_size); /* sub rsp, stack_size */ 2089 EMIT1(0x53); /* push rbx */ 2090 2091 /* Store number of arguments of the traced function: 2092 * mov rax, nr_args 2093 * mov QWORD PTR [rbp - args_off], rax 2094 */ 2095 emit_mov_imm64(&prog, BPF_REG_0, 0, (u32) nr_args); 2096 emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_0, -args_off); 2097 2098 if (flags & BPF_TRAMP_F_IP_ARG) { 2099 /* Store IP address of the traced function: 2100 * mov rax, QWORD PTR [rbp + 8] 2101 * sub rax, X86_PATCH_SIZE 2102 * mov QWORD PTR [rbp - ip_off], rax 2103 */ 2104 emit_ldx(&prog, BPF_DW, BPF_REG_0, BPF_REG_FP, 8); 2105 EMIT4(0x48, 0x83, 0xe8, X86_PATCH_SIZE); 2106 emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_0, -ip_off); 2107 } 2108 2109 save_regs(m, &prog, nr_args, regs_off); 2110 2111 if (flags & BPF_TRAMP_F_CALL_ORIG) { 2112 /* arg1: mov rdi, im */ 2113 emit_mov_imm64(&prog, BPF_REG_1, (long) im >> 32, (u32) (long) im); 2114 if (emit_call(&prog, __bpf_tramp_enter, prog)) { 2115 ret = -EINVAL; 2116 goto cleanup; 2117 } 2118 } 2119 2120 if (fentry->nr_links) 2121 if (invoke_bpf(m, &prog, fentry, regs_off, run_ctx_off, 2122 flags & BPF_TRAMP_F_RET_FENTRY_RET)) 2123 return -EINVAL; 2124 2125 if (fmod_ret->nr_links) { 2126 branches = kcalloc(fmod_ret->nr_links, sizeof(u8 *), 2127 GFP_KERNEL); 2128 if (!branches) 2129 return -ENOMEM; 2130 2131 if (invoke_bpf_mod_ret(m, &prog, fmod_ret, regs_off, 2132 run_ctx_off, branches)) { 2133 ret = -EINVAL; 2134 goto cleanup; 2135 } 2136 } 2137 2138 if (flags & BPF_TRAMP_F_CALL_ORIG) { 2139 restore_regs(m, &prog, nr_args, regs_off); 2140 2141 /* call original function */ 2142 if (emit_call(&prog, orig_call, prog)) { 2143 ret = -EINVAL; 2144 goto cleanup; 2145 } 2146 /* remember return value in a stack for bpf prog to access */ 2147 emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_0, -8); 2148 im->ip_after_call = prog; 2149 memcpy(prog, x86_nops[5], X86_PATCH_SIZE); 2150 prog += X86_PATCH_SIZE; 2151 } 2152 2153 if (fmod_ret->nr_links) { 2154 /* From Intel 64 and IA-32 Architectures Optimization 2155 * Reference Manual, 3.4.1.4 Code Alignment, Assembly/Compiler 2156 * Coding Rule 11: All branch targets should be 16-byte 2157 * aligned. 2158 */ 2159 emit_align(&prog, 16); 2160 /* Update the branches saved in invoke_bpf_mod_ret with the 2161 * aligned address of do_fexit. 2162 */ 2163 for (i = 0; i < fmod_ret->nr_links; i++) 2164 emit_cond_near_jump(&branches[i], prog, branches[i], 2165 X86_JNE); 2166 } 2167 2168 if (fexit->nr_links) 2169 if (invoke_bpf(m, &prog, fexit, regs_off, run_ctx_off, false)) { 2170 ret = -EINVAL; 2171 goto cleanup; 2172 } 2173 2174 if (flags & BPF_TRAMP_F_RESTORE_REGS) 2175 restore_regs(m, &prog, nr_args, regs_off); 2176 2177 /* This needs to be done regardless. If there were fmod_ret programs, 2178 * the return value is only updated on the stack and still needs to be 2179 * restored to R0. 2180 */ 2181 if (flags & BPF_TRAMP_F_CALL_ORIG) { 2182 im->ip_epilogue = prog; 2183 /* arg1: mov rdi, im */ 2184 emit_mov_imm64(&prog, BPF_REG_1, (long) im >> 32, (u32) (long) im); 2185 if (emit_call(&prog, __bpf_tramp_exit, prog)) { 2186 ret = -EINVAL; 2187 goto cleanup; 2188 } 2189 } 2190 /* restore return value of orig_call or fentry prog back into RAX */ 2191 if (save_ret) 2192 emit_ldx(&prog, BPF_DW, BPF_REG_0, BPF_REG_FP, -8); 2193 2194 EMIT1(0x5B); /* pop rbx */ 2195 EMIT1(0xC9); /* leave */ 2196 if (flags & BPF_TRAMP_F_SKIP_FRAME) 2197 /* skip our return address and return to parent */ 2198 EMIT4(0x48, 0x83, 0xC4, 8); /* add rsp, 8 */ 2199 EMIT1(0xC3); /* ret */ 2200 /* Make sure the trampoline generation logic doesn't overflow */ 2201 if (WARN_ON_ONCE(prog > (u8 *)image_end - BPF_INSN_SAFETY)) { 2202 ret = -EFAULT; 2203 goto cleanup; 2204 } 2205 ret = prog - (u8 *)image; 2206 2207 cleanup: 2208 kfree(branches); 2209 return ret; 2210 } 2211 2212 static int emit_bpf_dispatcher(u8 **pprog, int a, int b, s64 *progs) 2213 { 2214 u8 *jg_reloc, *prog = *pprog; 2215 int pivot, err, jg_bytes = 1; 2216 s64 jg_offset; 2217 2218 if (a == b) { 2219 /* Leaf node of recursion, i.e. not a range of indices 2220 * anymore. 2221 */ 2222 EMIT1(add_1mod(0x48, BPF_REG_3)); /* cmp rdx,func */ 2223 if (!is_simm32(progs[a])) 2224 return -1; 2225 EMIT2_off32(0x81, add_1reg(0xF8, BPF_REG_3), 2226 progs[a]); 2227 err = emit_cond_near_jump(&prog, /* je func */ 2228 (void *)progs[a], prog, 2229 X86_JE); 2230 if (err) 2231 return err; 2232 2233 emit_indirect_jump(&prog, 2 /* rdx */, prog); 2234 2235 *pprog = prog; 2236 return 0; 2237 } 2238 2239 /* Not a leaf node, so we pivot, and recursively descend into 2240 * the lower and upper ranges. 2241 */ 2242 pivot = (b - a) / 2; 2243 EMIT1(add_1mod(0x48, BPF_REG_3)); /* cmp rdx,func */ 2244 if (!is_simm32(progs[a + pivot])) 2245 return -1; 2246 EMIT2_off32(0x81, add_1reg(0xF8, BPF_REG_3), progs[a + pivot]); 2247 2248 if (pivot > 2) { /* jg upper_part */ 2249 /* Require near jump. */ 2250 jg_bytes = 4; 2251 EMIT2_off32(0x0F, X86_JG + 0x10, 0); 2252 } else { 2253 EMIT2(X86_JG, 0); 2254 } 2255 jg_reloc = prog; 2256 2257 err = emit_bpf_dispatcher(&prog, a, a + pivot, /* emit lower_part */ 2258 progs); 2259 if (err) 2260 return err; 2261 2262 /* From Intel 64 and IA-32 Architectures Optimization 2263 * Reference Manual, 3.4.1.4 Code Alignment, Assembly/Compiler 2264 * Coding Rule 11: All branch targets should be 16-byte 2265 * aligned. 2266 */ 2267 emit_align(&prog, 16); 2268 jg_offset = prog - jg_reloc; 2269 emit_code(jg_reloc - jg_bytes, jg_offset, jg_bytes); 2270 2271 err = emit_bpf_dispatcher(&prog, a + pivot + 1, /* emit upper_part */ 2272 b, progs); 2273 if (err) 2274 return err; 2275 2276 *pprog = prog; 2277 return 0; 2278 } 2279 2280 static int cmp_ips(const void *a, const void *b) 2281 { 2282 const s64 *ipa = a; 2283 const s64 *ipb = b; 2284 2285 if (*ipa > *ipb) 2286 return 1; 2287 if (*ipa < *ipb) 2288 return -1; 2289 return 0; 2290 } 2291 2292 int arch_prepare_bpf_dispatcher(void *image, s64 *funcs, int num_funcs) 2293 { 2294 u8 *prog = image; 2295 2296 sort(funcs, num_funcs, sizeof(funcs[0]), cmp_ips, NULL); 2297 return emit_bpf_dispatcher(&prog, 0, num_funcs - 1, funcs); 2298 } 2299 2300 struct x64_jit_data { 2301 struct bpf_binary_header *rw_header; 2302 struct bpf_binary_header *header; 2303 int *addrs; 2304 u8 *image; 2305 int proglen; 2306 struct jit_context ctx; 2307 }; 2308 2309 #define MAX_PASSES 20 2310 #define PADDING_PASSES (MAX_PASSES - 5) 2311 2312 struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) 2313 { 2314 struct bpf_binary_header *rw_header = NULL; 2315 struct bpf_binary_header *header = NULL; 2316 struct bpf_prog *tmp, *orig_prog = prog; 2317 struct x64_jit_data *jit_data; 2318 int proglen, oldproglen = 0; 2319 struct jit_context ctx = {}; 2320 bool tmp_blinded = false; 2321 bool extra_pass = false; 2322 bool padding = false; 2323 u8 *rw_image = NULL; 2324 u8 *image = NULL; 2325 int *addrs; 2326 int pass; 2327 int i; 2328 2329 if (!prog->jit_requested) 2330 return orig_prog; 2331 2332 tmp = bpf_jit_blind_constants(prog); 2333 /* 2334 * If blinding was requested and we failed during blinding, 2335 * we must fall back to the interpreter. 2336 */ 2337 if (IS_ERR(tmp)) 2338 return orig_prog; 2339 if (tmp != prog) { 2340 tmp_blinded = true; 2341 prog = tmp; 2342 } 2343 2344 jit_data = prog->aux->jit_data; 2345 if (!jit_data) { 2346 jit_data = kzalloc(sizeof(*jit_data), GFP_KERNEL); 2347 if (!jit_data) { 2348 prog = orig_prog; 2349 goto out; 2350 } 2351 prog->aux->jit_data = jit_data; 2352 } 2353 addrs = jit_data->addrs; 2354 if (addrs) { 2355 ctx = jit_data->ctx; 2356 oldproglen = jit_data->proglen; 2357 image = jit_data->image; 2358 header = jit_data->header; 2359 rw_header = jit_data->rw_header; 2360 rw_image = (void *)rw_header + ((void *)image - (void *)header); 2361 extra_pass = true; 2362 padding = true; 2363 goto skip_init_addrs; 2364 } 2365 addrs = kvmalloc_array(prog->len + 1, sizeof(*addrs), GFP_KERNEL); 2366 if (!addrs) { 2367 prog = orig_prog; 2368 goto out_addrs; 2369 } 2370 2371 /* 2372 * Before first pass, make a rough estimation of addrs[] 2373 * each BPF instruction is translated to less than 64 bytes 2374 */ 2375 for (proglen = 0, i = 0; i <= prog->len; i++) { 2376 proglen += 64; 2377 addrs[i] = proglen; 2378 } 2379 ctx.cleanup_addr = proglen; 2380 skip_init_addrs: 2381 2382 /* 2383 * JITed image shrinks with every pass and the loop iterates 2384 * until the image stops shrinking. Very large BPF programs 2385 * may converge on the last pass. In such case do one more 2386 * pass to emit the final image. 2387 */ 2388 for (pass = 0; pass < MAX_PASSES || image; pass++) { 2389 if (!padding && pass >= PADDING_PASSES) 2390 padding = true; 2391 proglen = do_jit(prog, addrs, image, rw_image, oldproglen, &ctx, padding); 2392 if (proglen <= 0) { 2393 out_image: 2394 image = NULL; 2395 if (header) { 2396 bpf_arch_text_copy(&header->size, &rw_header->size, 2397 sizeof(rw_header->size)); 2398 bpf_jit_binary_pack_free(header, rw_header); 2399 } 2400 /* Fall back to interpreter mode */ 2401 prog = orig_prog; 2402 if (extra_pass) { 2403 prog->bpf_func = NULL; 2404 prog->jited = 0; 2405 prog->jited_len = 0; 2406 } 2407 goto out_addrs; 2408 } 2409 if (image) { 2410 if (proglen != oldproglen) { 2411 pr_err("bpf_jit: proglen=%d != oldproglen=%d\n", 2412 proglen, oldproglen); 2413 goto out_image; 2414 } 2415 break; 2416 } 2417 if (proglen == oldproglen) { 2418 /* 2419 * The number of entries in extable is the number of BPF_LDX 2420 * insns that access kernel memory via "pointer to BTF type". 2421 * The verifier changed their opcode from LDX|MEM|size 2422 * to LDX|PROBE_MEM|size to make JITing easier. 2423 */ 2424 u32 align = __alignof__(struct exception_table_entry); 2425 u32 extable_size = prog->aux->num_exentries * 2426 sizeof(struct exception_table_entry); 2427 2428 /* allocate module memory for x86 insns and extable */ 2429 header = bpf_jit_binary_pack_alloc(roundup(proglen, align) + extable_size, 2430 &image, align, &rw_header, &rw_image, 2431 jit_fill_hole); 2432 if (!header) { 2433 prog = orig_prog; 2434 goto out_addrs; 2435 } 2436 prog->aux->extable = (void *) image + roundup(proglen, align); 2437 } 2438 oldproglen = proglen; 2439 cond_resched(); 2440 } 2441 2442 if (bpf_jit_enable > 1) 2443 bpf_jit_dump(prog->len, proglen, pass + 1, image); 2444 2445 if (image) { 2446 if (!prog->is_func || extra_pass) { 2447 /* 2448 * bpf_jit_binary_pack_finalize fails in two scenarios: 2449 * 1) header is not pointing to proper module memory; 2450 * 2) the arch doesn't support bpf_arch_text_copy(). 2451 * 2452 * Both cases are serious bugs and justify WARN_ON. 2453 */ 2454 if (WARN_ON(bpf_jit_binary_pack_finalize(prog, header, rw_header))) { 2455 /* header has been freed */ 2456 header = NULL; 2457 goto out_image; 2458 } 2459 2460 bpf_tail_call_direct_fixup(prog); 2461 } else { 2462 jit_data->addrs = addrs; 2463 jit_data->ctx = ctx; 2464 jit_data->proglen = proglen; 2465 jit_data->image = image; 2466 jit_data->header = header; 2467 jit_data->rw_header = rw_header; 2468 } 2469 prog->bpf_func = (void *)image; 2470 prog->jited = 1; 2471 prog->jited_len = proglen; 2472 } else { 2473 prog = orig_prog; 2474 } 2475 2476 if (!image || !prog->is_func || extra_pass) { 2477 if (image) 2478 bpf_prog_fill_jited_linfo(prog, addrs + 1); 2479 out_addrs: 2480 kvfree(addrs); 2481 kfree(jit_data); 2482 prog->aux->jit_data = NULL; 2483 } 2484 out: 2485 if (tmp_blinded) 2486 bpf_jit_prog_release_other(prog, prog == orig_prog ? 2487 tmp : orig_prog); 2488 return prog; 2489 } 2490 2491 bool bpf_jit_supports_kfunc_call(void) 2492 { 2493 return true; 2494 } 2495 2496 void *bpf_arch_text_copy(void *dst, void *src, size_t len) 2497 { 2498 if (text_poke_copy(dst, src, len) == NULL) 2499 return ERR_PTR(-EINVAL); 2500 return dst; 2501 } 2502 2503 /* Indicate the JIT backend supports mixing bpf2bpf and tailcalls. */ 2504 bool bpf_jit_supports_subprog_tailcalls(void) 2505 { 2506 return true; 2507 } 2508