1 //===- ScriptParser.cpp ---------------------------------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This file contains a recursive-descendent parser for linker scripts. 10 // Parsed results are stored to Config and Script global objects. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "ScriptParser.h" 15 #include "Config.h" 16 #include "Driver.h" 17 #include "InputSection.h" 18 #include "LinkerScript.h" 19 #include "OutputSections.h" 20 #include "ScriptLexer.h" 21 #include "Symbols.h" 22 #include "Target.h" 23 #include "lld/Common/Memory.h" 24 #include "llvm/ADT/SmallString.h" 25 #include "llvm/ADT/StringRef.h" 26 #include "llvm/ADT/StringSet.h" 27 #include "llvm/ADT/StringSwitch.h" 28 #include "llvm/BinaryFormat/ELF.h" 29 #include "llvm/Support/Casting.h" 30 #include "llvm/Support/ErrorHandling.h" 31 #include "llvm/Support/FileSystem.h" 32 #include "llvm/Support/Path.h" 33 #include <cassert> 34 #include <limits> 35 #include <vector> 36 37 using namespace llvm; 38 using namespace llvm::ELF; 39 using namespace llvm::support::endian; 40 using namespace lld; 41 using namespace lld::elf; 42 43 namespace { 44 class ScriptParser final : ScriptLexer { 45 public: 46 ScriptParser(MemoryBufferRef MB) : ScriptLexer(MB) { 47 // Initialize IsUnderSysroot 48 if (Config->Sysroot == "") 49 return; 50 StringRef Path = MB.getBufferIdentifier(); 51 for (; !Path.empty(); Path = sys::path::parent_path(Path)) { 52 if (!sys::fs::equivalent(Config->Sysroot, Path)) 53 continue; 54 IsUnderSysroot = true; 55 return; 56 } 57 } 58 59 void readLinkerScript(); 60 void readVersionScript(); 61 void readDynamicList(); 62 void readDefsym(StringRef Name); 63 64 private: 65 void addFile(StringRef Path); 66 67 void readAsNeeded(); 68 void readEntry(); 69 void readExtern(); 70 void readGroup(); 71 void readInclude(); 72 void readInput(); 73 void readMemory(); 74 void readOutput(); 75 void readOutputArch(); 76 void readOutputFormat(); 77 void readPhdrs(); 78 void readRegionAlias(); 79 void readSearchDir(); 80 void readSections(); 81 void readTarget(); 82 void readVersion(); 83 void readVersionScriptCommand(); 84 85 SymbolAssignment *readSymbolAssignment(StringRef Name); 86 ByteCommand *readByteCommand(StringRef Tok); 87 std::array<uint8_t, 4> readFill(); 88 bool readSectionDirective(OutputSection *Cmd, StringRef Tok1, StringRef Tok2); 89 void readSectionAddressType(OutputSection *Cmd); 90 OutputSection *readOverlaySectionDescription(); 91 OutputSection *readOutputSectionDescription(StringRef OutSec); 92 std::vector<BaseCommand *> readOverlay(); 93 std::vector<StringRef> readOutputSectionPhdrs(); 94 InputSectionDescription *readInputSectionDescription(StringRef Tok); 95 StringMatcher readFilePatterns(); 96 std::vector<SectionPattern> readInputSectionsList(); 97 InputSectionDescription *readInputSectionRules(StringRef FilePattern); 98 unsigned readPhdrType(); 99 SortSectionPolicy readSortKind(); 100 SymbolAssignment *readProvideHidden(bool Provide, bool Hidden); 101 SymbolAssignment *readAssignment(StringRef Tok); 102 void readSort(); 103 Expr readAssert(); 104 Expr readConstant(); 105 Expr getPageSize(); 106 107 uint64_t readMemoryAssignment(StringRef, StringRef, StringRef); 108 std::pair<uint32_t, uint32_t> readMemoryAttributes(); 109 110 Expr combine(StringRef Op, Expr L, Expr R); 111 Expr readExpr(); 112 Expr readExpr1(Expr Lhs, int MinPrec); 113 StringRef readParenLiteral(); 114 Expr readPrimary(); 115 Expr readTernary(Expr Cond); 116 Expr readParenExpr(); 117 118 // For parsing version script. 119 std::vector<SymbolVersion> readVersionExtern(); 120 void readAnonymousDeclaration(); 121 void readVersionDeclaration(StringRef VerStr); 122 123 std::pair<std::vector<SymbolVersion>, std::vector<SymbolVersion>> 124 readSymbols(); 125 126 // True if a script being read is in a subdirectory specified by -sysroot. 127 bool IsUnderSysroot = false; 128 129 // A set to detect an INCLUDE() cycle. 130 StringSet<> Seen; 131 }; 132 } // namespace 133 134 static StringRef unquote(StringRef S) { 135 if (S.startswith("\"")) 136 return S.substr(1, S.size() - 2); 137 return S; 138 } 139 140 // Some operations only support one non absolute value. Move the 141 // absolute one to the right hand side for convenience. 142 static void moveAbsRight(ExprValue &A, ExprValue &B) { 143 if (A.Sec == nullptr || (A.ForceAbsolute && !B.isAbsolute())) 144 std::swap(A, B); 145 if (!B.isAbsolute()) 146 error(A.Loc + ": at least one side of the expression must be absolute"); 147 } 148 149 static ExprValue add(ExprValue A, ExprValue B) { 150 moveAbsRight(A, B); 151 return {A.Sec, A.ForceAbsolute, A.getSectionOffset() + B.getValue(), A.Loc}; 152 } 153 154 static ExprValue sub(ExprValue A, ExprValue B) { 155 // The distance between two symbols in sections is absolute. 156 if (!A.isAbsolute() && !B.isAbsolute()) 157 return A.getValue() - B.getValue(); 158 return {A.Sec, false, A.getSectionOffset() - B.getValue(), A.Loc}; 159 } 160 161 static ExprValue bitAnd(ExprValue A, ExprValue B) { 162 moveAbsRight(A, B); 163 return {A.Sec, A.ForceAbsolute, 164 (A.getValue() & B.getValue()) - A.getSecAddr(), A.Loc}; 165 } 166 167 static ExprValue bitOr(ExprValue A, ExprValue B) { 168 moveAbsRight(A, B); 169 return {A.Sec, A.ForceAbsolute, 170 (A.getValue() | B.getValue()) - A.getSecAddr(), A.Loc}; 171 } 172 173 void ScriptParser::readDynamicList() { 174 Config->HasDynamicList = true; 175 expect("{"); 176 std::vector<SymbolVersion> Locals; 177 std::vector<SymbolVersion> Globals; 178 std::tie(Locals, Globals) = readSymbols(); 179 expect(";"); 180 181 if (!atEOF()) { 182 setError("EOF expected, but got " + next()); 183 return; 184 } 185 if (!Locals.empty()) { 186 setError("\"local:\" scope not supported in --dynamic-list"); 187 return; 188 } 189 190 for (SymbolVersion V : Globals) 191 Config->DynamicList.push_back(V); 192 } 193 194 void ScriptParser::readVersionScript() { 195 readVersionScriptCommand(); 196 if (!atEOF()) 197 setError("EOF expected, but got " + next()); 198 } 199 200 void ScriptParser::readVersionScriptCommand() { 201 if (consume("{")) { 202 readAnonymousDeclaration(); 203 return; 204 } 205 206 while (!atEOF() && !errorCount() && peek() != "}") { 207 StringRef VerStr = next(); 208 if (VerStr == "{") { 209 setError("anonymous version definition is used in " 210 "combination with other version definitions"); 211 return; 212 } 213 expect("{"); 214 readVersionDeclaration(VerStr); 215 } 216 } 217 218 void ScriptParser::readVersion() { 219 expect("{"); 220 readVersionScriptCommand(); 221 expect("}"); 222 } 223 224 void ScriptParser::readLinkerScript() { 225 while (!atEOF()) { 226 StringRef Tok = next(); 227 if (Tok == ";") 228 continue; 229 230 if (Tok == "ENTRY") { 231 readEntry(); 232 } else if (Tok == "EXTERN") { 233 readExtern(); 234 } else if (Tok == "GROUP") { 235 readGroup(); 236 } else if (Tok == "INCLUDE") { 237 readInclude(); 238 } else if (Tok == "INPUT") { 239 readInput(); 240 } else if (Tok == "MEMORY") { 241 readMemory(); 242 } else if (Tok == "OUTPUT") { 243 readOutput(); 244 } else if (Tok == "OUTPUT_ARCH") { 245 readOutputArch(); 246 } else if (Tok == "OUTPUT_FORMAT") { 247 readOutputFormat(); 248 } else if (Tok == "PHDRS") { 249 readPhdrs(); 250 } else if (Tok == "REGION_ALIAS") { 251 readRegionAlias(); 252 } else if (Tok == "SEARCH_DIR") { 253 readSearchDir(); 254 } else if (Tok == "SECTIONS") { 255 readSections(); 256 } else if (Tok == "TARGET") { 257 readTarget(); 258 } else if (Tok == "VERSION") { 259 readVersion(); 260 } else if (SymbolAssignment *Cmd = readAssignment(Tok)) { 261 Script->SectionCommands.push_back(Cmd); 262 } else { 263 setError("unknown directive: " + Tok); 264 } 265 } 266 } 267 268 void ScriptParser::readDefsym(StringRef Name) { 269 if (errorCount()) 270 return; 271 Expr E = readExpr(); 272 if (!atEOF()) 273 setError("EOF expected, but got " + next()); 274 SymbolAssignment *Cmd = make<SymbolAssignment>(Name, E, getCurrentLocation()); 275 Script->SectionCommands.push_back(Cmd); 276 } 277 278 void ScriptParser::addFile(StringRef S) { 279 if (IsUnderSysroot && S.startswith("/")) { 280 SmallString<128> PathData; 281 StringRef Path = (Config->Sysroot + S).toStringRef(PathData); 282 if (sys::fs::exists(Path)) { 283 Driver->addFile(Saver.save(Path), /*WithLOption=*/false); 284 return; 285 } 286 } 287 288 if (S.startswith("/")) { 289 Driver->addFile(S, /*WithLOption=*/false); 290 } else if (S.startswith("=")) { 291 if (Config->Sysroot.empty()) 292 Driver->addFile(S.substr(1), /*WithLOption=*/false); 293 else 294 Driver->addFile(Saver.save(Config->Sysroot + "/" + S.substr(1)), 295 /*WithLOption=*/false); 296 } else if (S.startswith("-l")) { 297 Driver->addLibrary(S.substr(2)); 298 } else if (sys::fs::exists(S)) { 299 Driver->addFile(S, /*WithLOption=*/false); 300 } else { 301 if (Optional<std::string> Path = findFromSearchPaths(S)) 302 Driver->addFile(Saver.save(*Path), /*WithLOption=*/true); 303 else 304 setError("unable to find " + S); 305 } 306 } 307 308 void ScriptParser::readAsNeeded() { 309 expect("("); 310 bool Orig = Config->AsNeeded; 311 Config->AsNeeded = true; 312 while (!errorCount() && !consume(")")) 313 addFile(unquote(next())); 314 Config->AsNeeded = Orig; 315 } 316 317 void ScriptParser::readEntry() { 318 // -e <symbol> takes predecence over ENTRY(<symbol>). 319 expect("("); 320 StringRef Tok = next(); 321 if (Config->Entry.empty()) 322 Config->Entry = Tok; 323 expect(")"); 324 } 325 326 void ScriptParser::readExtern() { 327 expect("("); 328 while (!errorCount() && !consume(")")) 329 Config->Undefined.push_back(unquote(next())); 330 } 331 332 void ScriptParser::readGroup() { 333 bool Orig = InputFile::IsInGroup; 334 InputFile::IsInGroup = true; 335 readInput(); 336 InputFile::IsInGroup = Orig; 337 if (!Orig) 338 ++InputFile::NextGroupId; 339 } 340 341 void ScriptParser::readInclude() { 342 StringRef Tok = unquote(next()); 343 344 if (!Seen.insert(Tok).second) { 345 setError("there is a cycle in linker script INCLUDEs"); 346 return; 347 } 348 349 if (Optional<std::string> Path = searchScript(Tok)) { 350 if (Optional<MemoryBufferRef> MB = readFile(*Path)) 351 tokenize(*MB); 352 return; 353 } 354 setError("cannot find linker script " + Tok); 355 } 356 357 void ScriptParser::readInput() { 358 expect("("); 359 while (!errorCount() && !consume(")")) { 360 if (consume("AS_NEEDED")) 361 readAsNeeded(); 362 else 363 addFile(unquote(next())); 364 } 365 } 366 367 void ScriptParser::readOutput() { 368 // -o <file> takes predecence over OUTPUT(<file>). 369 expect("("); 370 StringRef Tok = next(); 371 if (Config->OutputFile.empty()) 372 Config->OutputFile = unquote(Tok); 373 expect(")"); 374 } 375 376 void ScriptParser::readOutputArch() { 377 // OUTPUT_ARCH is ignored for now. 378 expect("("); 379 while (!errorCount() && !consume(")")) 380 skip(); 381 } 382 383 static std::pair<ELFKind, uint16_t> parseBfdName(StringRef S) { 384 return StringSwitch<std::pair<ELFKind, uint16_t>>(S) 385 .Case("elf32-i386", {ELF32LEKind, EM_386}) 386 .Case("elf32-iamcu", {ELF32LEKind, EM_IAMCU}) 387 .Case("elf32-littlearm", {ELF32LEKind, EM_ARM}) 388 .Case("elf32-x86-64", {ELF32LEKind, EM_X86_64}) 389 .Case("elf64-aarch64", {ELF64LEKind, EM_AARCH64}) 390 .Case("elf64-littleaarch64", {ELF64LEKind, EM_AARCH64}) 391 .Case("elf32-powerpc", {ELF32BEKind, EM_PPC}) 392 .Case("elf64-powerpc", {ELF64BEKind, EM_PPC64}) 393 .Case("elf64-powerpcle", {ELF64LEKind, EM_PPC64}) 394 .Case("elf64-x86-64", {ELF64LEKind, EM_X86_64}) 395 .Cases("elf32-tradbigmips", "elf32-bigmips", {ELF32BEKind, EM_MIPS}) 396 .Case("elf32-ntradbigmips", {ELF32BEKind, EM_MIPS}) 397 .Case("elf32-tradlittlemips", {ELF32LEKind, EM_MIPS}) 398 .Case("elf32-ntradlittlemips", {ELF32LEKind, EM_MIPS}) 399 .Case("elf64-tradbigmips", {ELF64BEKind, EM_MIPS}) 400 .Case("elf64-tradlittlemips", {ELF64LEKind, EM_MIPS}) 401 .Case("elf32-littleriscv", {ELF32LEKind, EM_RISCV}) 402 .Case("elf64-littleriscv", {ELF64LEKind, EM_RISCV}) 403 .Default({ELFNoneKind, EM_NONE}); 404 } 405 406 // Parse OUTPUT_FORMAT(bfdname) or OUTPUT_FORMAT(bfdname, big, little). 407 // Currently we ignore big and little parameters. 408 void ScriptParser::readOutputFormat() { 409 expect("("); 410 411 StringRef Name = unquote(next()); 412 StringRef S = Name; 413 if (S.consume_back("-freebsd")) 414 Config->OSABI = ELFOSABI_FREEBSD; 415 416 std::tie(Config->EKind, Config->EMachine) = parseBfdName(S); 417 if (Config->EMachine == EM_NONE) 418 setError("unknown output format name: " + Name); 419 if (S == "elf32-ntradlittlemips" || S == "elf32-ntradbigmips") 420 Config->MipsN32Abi = true; 421 422 if (consume(")")) 423 return; 424 expect(","); 425 skip(); 426 expect(","); 427 skip(); 428 expect(")"); 429 } 430 431 void ScriptParser::readPhdrs() { 432 expect("{"); 433 434 while (!errorCount() && !consume("}")) { 435 PhdrsCommand Cmd; 436 Cmd.Name = next(); 437 Cmd.Type = readPhdrType(); 438 439 while (!errorCount() && !consume(";")) { 440 if (consume("FILEHDR")) 441 Cmd.HasFilehdr = true; 442 else if (consume("PHDRS")) 443 Cmd.HasPhdrs = true; 444 else if (consume("AT")) 445 Cmd.LMAExpr = readParenExpr(); 446 else if (consume("FLAGS")) 447 Cmd.Flags = readParenExpr()().getValue(); 448 else 449 setError("unexpected header attribute: " + next()); 450 } 451 452 Script->PhdrsCommands.push_back(Cmd); 453 } 454 } 455 456 void ScriptParser::readRegionAlias() { 457 expect("("); 458 StringRef Alias = unquote(next()); 459 expect(","); 460 StringRef Name = next(); 461 expect(")"); 462 463 if (Script->MemoryRegions.count(Alias)) 464 setError("redefinition of memory region '" + Alias + "'"); 465 if (!Script->MemoryRegions.count(Name)) 466 setError("memory region '" + Name + "' is not defined"); 467 Script->MemoryRegions.insert({Alias, Script->MemoryRegions[Name]}); 468 } 469 470 void ScriptParser::readSearchDir() { 471 expect("("); 472 StringRef Tok = next(); 473 if (!Config->Nostdlib) 474 Config->SearchPaths.push_back(unquote(Tok)); 475 expect(")"); 476 } 477 478 // This reads an overlay description. Overlays are used to describe output 479 // sections that use the same virtual memory range and normally would trigger 480 // linker's sections sanity check failures. 481 // https://sourceware.org/binutils/docs/ld/Overlay-Description.html#Overlay-Description 482 std::vector<BaseCommand *> ScriptParser::readOverlay() { 483 // VA and LMA expressions are optional, though for simplicity of 484 // implementation we assume they are not. That is what OVERLAY was designed 485 // for first of all: to allow sections with overlapping VAs at different LMAs. 486 Expr AddrExpr = readExpr(); 487 expect(":"); 488 expect("AT"); 489 Expr LMAExpr = readParenExpr(); 490 expect("{"); 491 492 std::vector<BaseCommand *> V; 493 OutputSection *Prev = nullptr; 494 while (!errorCount() && !consume("}")) { 495 // VA is the same for all sections. The LMAs are consecutive in memory 496 // starting from the base load address specified. 497 OutputSection *OS = readOverlaySectionDescription(); 498 OS->AddrExpr = AddrExpr; 499 if (Prev) 500 OS->LMAExpr = [=] { return Prev->getLMA() + Prev->Size; }; 501 else 502 OS->LMAExpr = LMAExpr; 503 V.push_back(OS); 504 Prev = OS; 505 } 506 507 // According to the specification, at the end of the overlay, the location 508 // counter should be equal to the overlay base address plus size of the 509 // largest section seen in the overlay. 510 // Here we want to create the Dot assignment command to achieve that. 511 Expr MoveDot = [=] { 512 uint64_t Max = 0; 513 for (BaseCommand *Cmd : V) 514 Max = std::max(Max, cast<OutputSection>(Cmd)->Size); 515 return AddrExpr().getValue() + Max; 516 }; 517 V.push_back(make<SymbolAssignment>(".", MoveDot, getCurrentLocation())); 518 return V; 519 } 520 521 void ScriptParser::readSections() { 522 Script->HasSectionsCommand = true; 523 524 // -no-rosegment is used to avoid placing read only non-executable sections in 525 // their own segment. We do the same if SECTIONS command is present in linker 526 // script. See comment for computeFlags(). 527 Config->SingleRoRx = true; 528 529 expect("{"); 530 std::vector<BaseCommand *> V; 531 while (!errorCount() && !consume("}")) { 532 StringRef Tok = next(); 533 if (Tok == "OVERLAY") { 534 for (BaseCommand *Cmd : readOverlay()) 535 V.push_back(Cmd); 536 continue; 537 } else if (Tok == "INCLUDE") { 538 readInclude(); 539 continue; 540 } 541 542 if (BaseCommand *Cmd = readAssignment(Tok)) 543 V.push_back(Cmd); 544 else 545 V.push_back(readOutputSectionDescription(Tok)); 546 } 547 548 if (!atEOF() && consume("INSERT")) { 549 std::vector<BaseCommand *> *Dest = nullptr; 550 if (consume("AFTER")) 551 Dest = &Script->InsertAfterCommands[next()]; 552 else if (consume("BEFORE")) 553 Dest = &Script->InsertBeforeCommands[next()]; 554 else 555 setError("expected AFTER/BEFORE, but got '" + next() + "'"); 556 if (Dest) 557 Dest->insert(Dest->end(), V.begin(), V.end()); 558 return; 559 } 560 561 Script->SectionCommands.insert(Script->SectionCommands.end(), V.begin(), 562 V.end()); 563 } 564 565 void ScriptParser::readTarget() { 566 // TARGET(foo) is an alias for "--format foo". Unlike GNU linkers, 567 // we accept only a limited set of BFD names (i.e. "elf" or "binary") 568 // for --format. We recognize only /^elf/ and "binary" in the linker 569 // script as well. 570 expect("("); 571 StringRef Tok = next(); 572 expect(")"); 573 574 if (Tok.startswith("elf")) 575 Config->FormatBinary = false; 576 else if (Tok == "binary") 577 Config->FormatBinary = true; 578 else 579 setError("unknown target: " + Tok); 580 } 581 582 static int precedence(StringRef Op) { 583 return StringSwitch<int>(Op) 584 .Cases("*", "/", "%", 8) 585 .Cases("+", "-", 7) 586 .Cases("<<", ">>", 6) 587 .Cases("<", "<=", ">", ">=", "==", "!=", 5) 588 .Case("&", 4) 589 .Case("|", 3) 590 .Case("&&", 2) 591 .Case("||", 1) 592 .Default(-1); 593 } 594 595 StringMatcher ScriptParser::readFilePatterns() { 596 std::vector<StringRef> V; 597 while (!errorCount() && !consume(")")) 598 V.push_back(next()); 599 return StringMatcher(V); 600 } 601 602 SortSectionPolicy ScriptParser::readSortKind() { 603 if (consume("SORT") || consume("SORT_BY_NAME")) 604 return SortSectionPolicy::Name; 605 if (consume("SORT_BY_ALIGNMENT")) 606 return SortSectionPolicy::Alignment; 607 if (consume("SORT_BY_INIT_PRIORITY")) 608 return SortSectionPolicy::Priority; 609 if (consume("SORT_NONE")) 610 return SortSectionPolicy::None; 611 return SortSectionPolicy::Default; 612 } 613 614 // Reads SECTIONS command contents in the following form: 615 // 616 // <contents> ::= <elem>* 617 // <elem> ::= <exclude>? <glob-pattern> 618 // <exclude> ::= "EXCLUDE_FILE" "(" <glob-pattern>+ ")" 619 // 620 // For example, 621 // 622 // *(.foo EXCLUDE_FILE (a.o) .bar EXCLUDE_FILE (b.o) .baz) 623 // 624 // is parsed as ".foo", ".bar" with "a.o", and ".baz" with "b.o". 625 // The semantics of that is section .foo in any file, section .bar in 626 // any file but a.o, and section .baz in any file but b.o. 627 std::vector<SectionPattern> ScriptParser::readInputSectionsList() { 628 std::vector<SectionPattern> Ret; 629 while (!errorCount() && peek() != ")") { 630 StringMatcher ExcludeFilePat; 631 if (consume("EXCLUDE_FILE")) { 632 expect("("); 633 ExcludeFilePat = readFilePatterns(); 634 } 635 636 std::vector<StringRef> V; 637 while (!errorCount() && peek() != ")" && peek() != "EXCLUDE_FILE") 638 V.push_back(unquote(next())); 639 640 if (!V.empty()) 641 Ret.push_back({std::move(ExcludeFilePat), StringMatcher(V)}); 642 else 643 setError("section pattern is expected"); 644 } 645 return Ret; 646 } 647 648 // Reads contents of "SECTIONS" directive. That directive contains a 649 // list of glob patterns for input sections. The grammar is as follows. 650 // 651 // <patterns> ::= <section-list> 652 // | <sort> "(" <section-list> ")" 653 // | <sort> "(" <sort> "(" <section-list> ")" ")" 654 // 655 // <sort> ::= "SORT" | "SORT_BY_NAME" | "SORT_BY_ALIGNMENT" 656 // | "SORT_BY_INIT_PRIORITY" | "SORT_NONE" 657 // 658 // <section-list> is parsed by readInputSectionsList(). 659 InputSectionDescription * 660 ScriptParser::readInputSectionRules(StringRef FilePattern) { 661 auto *Cmd = make<InputSectionDescription>(FilePattern); 662 expect("("); 663 664 while (!errorCount() && !consume(")")) { 665 SortSectionPolicy Outer = readSortKind(); 666 SortSectionPolicy Inner = SortSectionPolicy::Default; 667 std::vector<SectionPattern> V; 668 if (Outer != SortSectionPolicy::Default) { 669 expect("("); 670 Inner = readSortKind(); 671 if (Inner != SortSectionPolicy::Default) { 672 expect("("); 673 V = readInputSectionsList(); 674 expect(")"); 675 } else { 676 V = readInputSectionsList(); 677 } 678 expect(")"); 679 } else { 680 V = readInputSectionsList(); 681 } 682 683 for (SectionPattern &Pat : V) { 684 Pat.SortInner = Inner; 685 Pat.SortOuter = Outer; 686 } 687 688 std::move(V.begin(), V.end(), std::back_inserter(Cmd->SectionPatterns)); 689 } 690 return Cmd; 691 } 692 693 InputSectionDescription * 694 ScriptParser::readInputSectionDescription(StringRef Tok) { 695 // Input section wildcard can be surrounded by KEEP. 696 // https://sourceware.org/binutils/docs/ld/Input-Section-Keep.html#Input-Section-Keep 697 if (Tok == "KEEP") { 698 expect("("); 699 StringRef FilePattern = next(); 700 InputSectionDescription *Cmd = readInputSectionRules(FilePattern); 701 expect(")"); 702 Script->KeptSections.push_back(Cmd); 703 return Cmd; 704 } 705 return readInputSectionRules(Tok); 706 } 707 708 void ScriptParser::readSort() { 709 expect("("); 710 expect("CONSTRUCTORS"); 711 expect(")"); 712 } 713 714 Expr ScriptParser::readAssert() { 715 expect("("); 716 Expr E = readExpr(); 717 expect(","); 718 StringRef Msg = unquote(next()); 719 expect(")"); 720 721 return [=] { 722 if (!E().getValue()) 723 error(Msg); 724 return Script->getDot(); 725 }; 726 } 727 728 // Tries to read the special directive for an output section definition which 729 // can be one of following: "(NOLOAD)", "(COPY)", "(INFO)" or "(OVERLAY)". 730 // Tok1 and Tok2 are next 2 tokens peeked. See comment for readSectionAddressType below. 731 bool ScriptParser::readSectionDirective(OutputSection *Cmd, StringRef Tok1, StringRef Tok2) { 732 if (Tok1 != "(") 733 return false; 734 if (Tok2 != "NOLOAD" && Tok2 != "COPY" && Tok2 != "INFO" && Tok2 != "OVERLAY") 735 return false; 736 737 expect("("); 738 if (consume("NOLOAD")) { 739 Cmd->Noload = true; 740 } else { 741 skip(); // This is "COPY", "INFO" or "OVERLAY". 742 Cmd->NonAlloc = true; 743 } 744 expect(")"); 745 return true; 746 } 747 748 // Reads an expression and/or the special directive for an output 749 // section definition. Directive is one of following: "(NOLOAD)", 750 // "(COPY)", "(INFO)" or "(OVERLAY)". 751 // 752 // An output section name can be followed by an address expression 753 // and/or directive. This grammar is not LL(1) because "(" can be 754 // interpreted as either the beginning of some expression or beginning 755 // of directive. 756 // 757 // https://sourceware.org/binutils/docs/ld/Output-Section-Address.html 758 // https://sourceware.org/binutils/docs/ld/Output-Section-Type.html 759 void ScriptParser::readSectionAddressType(OutputSection *Cmd) { 760 if (readSectionDirective(Cmd, peek(), peek2())) 761 return; 762 763 Cmd->AddrExpr = readExpr(); 764 if (peek() == "(" && !readSectionDirective(Cmd, "(", peek2())) 765 setError("unknown section directive: " + peek2()); 766 } 767 768 static Expr checkAlignment(Expr E, std::string &Loc) { 769 return [=] { 770 uint64_t Alignment = std::max((uint64_t)1, E().getValue()); 771 if (!isPowerOf2_64(Alignment)) { 772 error(Loc + ": alignment must be power of 2"); 773 return (uint64_t)1; // Return a dummy value. 774 } 775 return Alignment; 776 }; 777 } 778 779 OutputSection *ScriptParser::readOverlaySectionDescription() { 780 OutputSection *Cmd = 781 Script->createOutputSection(next(), getCurrentLocation()); 782 Cmd->InOverlay = true; 783 expect("{"); 784 while (!errorCount() && !consume("}")) 785 Cmd->SectionCommands.push_back(readInputSectionRules(next())); 786 Cmd->Phdrs = readOutputSectionPhdrs(); 787 return Cmd; 788 } 789 790 OutputSection *ScriptParser::readOutputSectionDescription(StringRef OutSec) { 791 OutputSection *Cmd = 792 Script->createOutputSection(OutSec, getCurrentLocation()); 793 794 size_t SymbolsReferenced = Script->ReferencedSymbols.size(); 795 796 if (peek() != ":") 797 readSectionAddressType(Cmd); 798 expect(":"); 799 800 std::string Location = getCurrentLocation(); 801 if (consume("AT")) 802 Cmd->LMAExpr = readParenExpr(); 803 if (consume("ALIGN")) 804 Cmd->AlignExpr = checkAlignment(readParenExpr(), Location); 805 if (consume("SUBALIGN")) 806 Cmd->SubalignExpr = checkAlignment(readParenExpr(), Location); 807 808 // Parse constraints. 809 if (consume("ONLY_IF_RO")) 810 Cmd->Constraint = ConstraintKind::ReadOnly; 811 if (consume("ONLY_IF_RW")) 812 Cmd->Constraint = ConstraintKind::ReadWrite; 813 expect("{"); 814 815 while (!errorCount() && !consume("}")) { 816 StringRef Tok = next(); 817 if (Tok == ";") { 818 // Empty commands are allowed. Do nothing here. 819 } else if (SymbolAssignment *Assign = readAssignment(Tok)) { 820 Cmd->SectionCommands.push_back(Assign); 821 } else if (ByteCommand *Data = readByteCommand(Tok)) { 822 Cmd->SectionCommands.push_back(Data); 823 } else if (Tok == "CONSTRUCTORS") { 824 // CONSTRUCTORS is a keyword to make the linker recognize C++ ctors/dtors 825 // by name. This is for very old file formats such as ECOFF/XCOFF. 826 // For ELF, we should ignore. 827 } else if (Tok == "FILL") { 828 // We handle the FILL command as an alias for =fillexp section attribute, 829 // which is different from what GNU linkers do. 830 // https://sourceware.org/binutils/docs/ld/Output-Section-Data.html 831 Cmd->Filler = readFill(); 832 } else if (Tok == "SORT") { 833 readSort(); 834 } else if (Tok == "INCLUDE") { 835 readInclude(); 836 } else if (peek() == "(") { 837 Cmd->SectionCommands.push_back(readInputSectionDescription(Tok)); 838 } else { 839 // We have a file name and no input sections description. It is not a 840 // commonly used syntax, but still acceptable. In that case, all sections 841 // from the file will be included. 842 auto *ISD = make<InputSectionDescription>(Tok); 843 ISD->SectionPatterns.push_back({{}, StringMatcher({"*"})}); 844 Cmd->SectionCommands.push_back(ISD); 845 } 846 } 847 848 if (consume(">")) 849 Cmd->MemoryRegionName = next(); 850 851 if (consume("AT")) { 852 expect(">"); 853 Cmd->LMARegionName = next(); 854 } 855 856 if (Cmd->LMAExpr && !Cmd->LMARegionName.empty()) 857 error("section can't have both LMA and a load region"); 858 859 Cmd->Phdrs = readOutputSectionPhdrs(); 860 861 if (peek() == "=" || peek().startswith("=")) { 862 InExpr = true; 863 consume("="); 864 Cmd->Filler = readFill(); 865 InExpr = false; 866 } 867 868 // Consume optional comma following output section command. 869 consume(","); 870 871 if (Script->ReferencedSymbols.size() > SymbolsReferenced) 872 Cmd->ExpressionsUseSymbols = true; 873 return Cmd; 874 } 875 876 // Reads a `=<fillexp>` expression and returns its value as a big-endian number. 877 // https://sourceware.org/binutils/docs/ld/Output-Section-Fill.html 878 // We do not support using symbols in such expressions. 879 // 880 // When reading a hexstring, ld.bfd handles it as a blob of arbitrary 881 // size, while ld.gold always handles it as a 32-bit big-endian number. 882 // We are compatible with ld.gold because it's easier to implement. 883 std::array<uint8_t, 4> ScriptParser::readFill() { 884 uint64_t Value = readExpr()().Val; 885 if (Value > UINT32_MAX) 886 setError("filler expression result does not fit 32-bit: 0x" + 887 Twine::utohexstr(Value)); 888 889 std::array<uint8_t, 4> Buf; 890 write32be(Buf.data(), (uint32_t)Value); 891 return Buf; 892 } 893 894 SymbolAssignment *ScriptParser::readProvideHidden(bool Provide, bool Hidden) { 895 expect("("); 896 SymbolAssignment *Cmd = readSymbolAssignment(next()); 897 Cmd->Provide = Provide; 898 Cmd->Hidden = Hidden; 899 expect(")"); 900 return Cmd; 901 } 902 903 SymbolAssignment *ScriptParser::readAssignment(StringRef Tok) { 904 // Assert expression returns Dot, so this is equal to ".=." 905 if (Tok == "ASSERT") 906 return make<SymbolAssignment>(".", readAssert(), getCurrentLocation()); 907 908 size_t OldPos = Pos; 909 SymbolAssignment *Cmd = nullptr; 910 if (peek() == "=" || peek() == "+=") 911 Cmd = readSymbolAssignment(Tok); 912 else if (Tok == "PROVIDE") 913 Cmd = readProvideHidden(true, false); 914 else if (Tok == "HIDDEN") 915 Cmd = readProvideHidden(false, true); 916 else if (Tok == "PROVIDE_HIDDEN") 917 Cmd = readProvideHidden(true, true); 918 919 if (Cmd) { 920 Cmd->CommandString = 921 Tok.str() + " " + 922 llvm::join(Tokens.begin() + OldPos, Tokens.begin() + Pos, " "); 923 expect(";"); 924 } 925 return Cmd; 926 } 927 928 SymbolAssignment *ScriptParser::readSymbolAssignment(StringRef Name) { 929 StringRef Op = next(); 930 assert(Op == "=" || Op == "+="); 931 Expr E = readExpr(); 932 if (Op == "+=") { 933 std::string Loc = getCurrentLocation(); 934 E = [=] { return add(Script->getSymbolValue(Name, Loc), E()); }; 935 } 936 return make<SymbolAssignment>(Name, E, getCurrentLocation()); 937 } 938 939 // This is an operator-precedence parser to parse a linker 940 // script expression. 941 Expr ScriptParser::readExpr() { 942 // Our lexer is context-aware. Set the in-expression bit so that 943 // they apply different tokenization rules. 944 bool Orig = InExpr; 945 InExpr = true; 946 Expr E = readExpr1(readPrimary(), 0); 947 InExpr = Orig; 948 return E; 949 } 950 951 Expr ScriptParser::combine(StringRef Op, Expr L, Expr R) { 952 if (Op == "+") 953 return [=] { return add(L(), R()); }; 954 if (Op == "-") 955 return [=] { return sub(L(), R()); }; 956 if (Op == "*") 957 return [=] { return L().getValue() * R().getValue(); }; 958 if (Op == "/") { 959 std::string Loc = getCurrentLocation(); 960 return [=]() -> uint64_t { 961 if (uint64_t RV = R().getValue()) 962 return L().getValue() / RV; 963 error(Loc + ": division by zero"); 964 return 0; 965 }; 966 } 967 if (Op == "%") { 968 std::string Loc = getCurrentLocation(); 969 return [=]() -> uint64_t { 970 if (uint64_t RV = R().getValue()) 971 return L().getValue() % RV; 972 error(Loc + ": modulo by zero"); 973 return 0; 974 }; 975 } 976 if (Op == "<<") 977 return [=] { return L().getValue() << R().getValue(); }; 978 if (Op == ">>") 979 return [=] { return L().getValue() >> R().getValue(); }; 980 if (Op == "<") 981 return [=] { return L().getValue() < R().getValue(); }; 982 if (Op == ">") 983 return [=] { return L().getValue() > R().getValue(); }; 984 if (Op == ">=") 985 return [=] { return L().getValue() >= R().getValue(); }; 986 if (Op == "<=") 987 return [=] { return L().getValue() <= R().getValue(); }; 988 if (Op == "==") 989 return [=] { return L().getValue() == R().getValue(); }; 990 if (Op == "!=") 991 return [=] { return L().getValue() != R().getValue(); }; 992 if (Op == "||") 993 return [=] { return L().getValue() || R().getValue(); }; 994 if (Op == "&&") 995 return [=] { return L().getValue() && R().getValue(); }; 996 if (Op == "&") 997 return [=] { return bitAnd(L(), R()); }; 998 if (Op == "|") 999 return [=] { return bitOr(L(), R()); }; 1000 llvm_unreachable("invalid operator"); 1001 } 1002 1003 // This is a part of the operator-precedence parser. This function 1004 // assumes that the remaining token stream starts with an operator. 1005 Expr ScriptParser::readExpr1(Expr Lhs, int MinPrec) { 1006 while (!atEOF() && !errorCount()) { 1007 // Read an operator and an expression. 1008 if (consume("?")) 1009 return readTernary(Lhs); 1010 StringRef Op1 = peek(); 1011 if (precedence(Op1) < MinPrec) 1012 break; 1013 skip(); 1014 Expr Rhs = readPrimary(); 1015 1016 // Evaluate the remaining part of the expression first if the 1017 // next operator has greater precedence than the previous one. 1018 // For example, if we have read "+" and "3", and if the next 1019 // operator is "*", then we'll evaluate 3 * ... part first. 1020 while (!atEOF()) { 1021 StringRef Op2 = peek(); 1022 if (precedence(Op2) <= precedence(Op1)) 1023 break; 1024 Rhs = readExpr1(Rhs, precedence(Op2)); 1025 } 1026 1027 Lhs = combine(Op1, Lhs, Rhs); 1028 } 1029 return Lhs; 1030 } 1031 1032 Expr ScriptParser::getPageSize() { 1033 std::string Location = getCurrentLocation(); 1034 return [=]() -> uint64_t { 1035 if (Target) 1036 return Config->CommonPageSize; 1037 error(Location + ": unable to calculate page size"); 1038 return 4096; // Return a dummy value. 1039 }; 1040 } 1041 1042 Expr ScriptParser::readConstant() { 1043 StringRef S = readParenLiteral(); 1044 if (S == "COMMONPAGESIZE") 1045 return getPageSize(); 1046 if (S == "MAXPAGESIZE") 1047 return [] { return Config->MaxPageSize; }; 1048 setError("unknown constant: " + S); 1049 return [] { return 0; }; 1050 } 1051 1052 // Parses Tok as an integer. It recognizes hexadecimal (prefixed with 1053 // "0x" or suffixed with "H") and decimal numbers. Decimal numbers may 1054 // have "K" (Ki) or "M" (Mi) suffixes. 1055 static Optional<uint64_t> parseInt(StringRef Tok) { 1056 // Hexadecimal 1057 uint64_t Val; 1058 if (Tok.startswith_lower("0x")) { 1059 if (!to_integer(Tok.substr(2), Val, 16)) 1060 return None; 1061 return Val; 1062 } 1063 if (Tok.endswith_lower("H")) { 1064 if (!to_integer(Tok.drop_back(), Val, 16)) 1065 return None; 1066 return Val; 1067 } 1068 1069 // Decimal 1070 if (Tok.endswith_lower("K")) { 1071 if (!to_integer(Tok.drop_back(), Val, 10)) 1072 return None; 1073 return Val * 1024; 1074 } 1075 if (Tok.endswith_lower("M")) { 1076 if (!to_integer(Tok.drop_back(), Val, 10)) 1077 return None; 1078 return Val * 1024 * 1024; 1079 } 1080 if (!to_integer(Tok, Val, 10)) 1081 return None; 1082 return Val; 1083 } 1084 1085 ByteCommand *ScriptParser::readByteCommand(StringRef Tok) { 1086 int Size = StringSwitch<int>(Tok) 1087 .Case("BYTE", 1) 1088 .Case("SHORT", 2) 1089 .Case("LONG", 4) 1090 .Case("QUAD", 8) 1091 .Default(-1); 1092 if (Size == -1) 1093 return nullptr; 1094 1095 size_t OldPos = Pos; 1096 Expr E = readParenExpr(); 1097 std::string CommandString = 1098 Tok.str() + " " + 1099 llvm::join(Tokens.begin() + OldPos, Tokens.begin() + Pos, " "); 1100 return make<ByteCommand>(E, Size, CommandString); 1101 } 1102 1103 StringRef ScriptParser::readParenLiteral() { 1104 expect("("); 1105 bool Orig = InExpr; 1106 InExpr = false; 1107 StringRef Tok = next(); 1108 InExpr = Orig; 1109 expect(")"); 1110 return Tok; 1111 } 1112 1113 static void checkIfExists(OutputSection *Cmd, StringRef Location) { 1114 if (Cmd->Location.empty() && Script->ErrorOnMissingSection) 1115 error(Location + ": undefined section " + Cmd->Name); 1116 } 1117 1118 Expr ScriptParser::readPrimary() { 1119 if (peek() == "(") 1120 return readParenExpr(); 1121 1122 if (consume("~")) { 1123 Expr E = readPrimary(); 1124 return [=] { return ~E().getValue(); }; 1125 } 1126 if (consume("!")) { 1127 Expr E = readPrimary(); 1128 return [=] { return !E().getValue(); }; 1129 } 1130 if (consume("-")) { 1131 Expr E = readPrimary(); 1132 return [=] { return -E().getValue(); }; 1133 } 1134 1135 StringRef Tok = next(); 1136 std::string Location = getCurrentLocation(); 1137 1138 // Built-in functions are parsed here. 1139 // https://sourceware.org/binutils/docs/ld/Builtin-Functions.html. 1140 if (Tok == "ABSOLUTE") { 1141 Expr Inner = readParenExpr(); 1142 return [=] { 1143 ExprValue I = Inner(); 1144 I.ForceAbsolute = true; 1145 return I; 1146 }; 1147 } 1148 if (Tok == "ADDR") { 1149 StringRef Name = readParenLiteral(); 1150 OutputSection *Sec = Script->getOrCreateOutputSection(Name); 1151 Sec->UsedInExpression = true; 1152 return [=]() -> ExprValue { 1153 checkIfExists(Sec, Location); 1154 return {Sec, false, 0, Location}; 1155 }; 1156 } 1157 if (Tok == "ALIGN") { 1158 expect("("); 1159 Expr E = readExpr(); 1160 if (consume(")")) { 1161 E = checkAlignment(E, Location); 1162 return [=] { return alignTo(Script->getDot(), E().getValue()); }; 1163 } 1164 expect(","); 1165 Expr E2 = checkAlignment(readExpr(), Location); 1166 expect(")"); 1167 return [=] { 1168 ExprValue V = E(); 1169 V.Alignment = E2().getValue(); 1170 return V; 1171 }; 1172 } 1173 if (Tok == "ALIGNOF") { 1174 StringRef Name = readParenLiteral(); 1175 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1176 return [=] { 1177 checkIfExists(Cmd, Location); 1178 return Cmd->Alignment; 1179 }; 1180 } 1181 if (Tok == "ASSERT") 1182 return readAssert(); 1183 if (Tok == "CONSTANT") 1184 return readConstant(); 1185 if (Tok == "DATA_SEGMENT_ALIGN") { 1186 expect("("); 1187 Expr E = readExpr(); 1188 expect(","); 1189 readExpr(); 1190 expect(")"); 1191 return [=] { 1192 return alignTo(Script->getDot(), std::max((uint64_t)1, E().getValue())); 1193 }; 1194 } 1195 if (Tok == "DATA_SEGMENT_END") { 1196 expect("("); 1197 expect("."); 1198 expect(")"); 1199 return [] { return Script->getDot(); }; 1200 } 1201 if (Tok == "DATA_SEGMENT_RELRO_END") { 1202 // GNU linkers implements more complicated logic to handle 1203 // DATA_SEGMENT_RELRO_END. We instead ignore the arguments and 1204 // just align to the next page boundary for simplicity. 1205 expect("("); 1206 readExpr(); 1207 expect(","); 1208 readExpr(); 1209 expect(")"); 1210 Expr E = getPageSize(); 1211 return [=] { return alignTo(Script->getDot(), E().getValue()); }; 1212 } 1213 if (Tok == "DEFINED") { 1214 StringRef Name = readParenLiteral(); 1215 return [=] { return Symtab->find(Name) ? 1 : 0; }; 1216 } 1217 if (Tok == "LENGTH") { 1218 StringRef Name = readParenLiteral(); 1219 if (Script->MemoryRegions.count(Name) == 0) { 1220 setError("memory region not defined: " + Name); 1221 return [] { return 0; }; 1222 } 1223 return [=] { return Script->MemoryRegions[Name]->Length; }; 1224 } 1225 if (Tok == "LOADADDR") { 1226 StringRef Name = readParenLiteral(); 1227 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1228 Cmd->UsedInExpression = true; 1229 return [=] { 1230 checkIfExists(Cmd, Location); 1231 return Cmd->getLMA(); 1232 }; 1233 } 1234 if (Tok == "MAX" || Tok == "MIN") { 1235 expect("("); 1236 Expr A = readExpr(); 1237 expect(","); 1238 Expr B = readExpr(); 1239 expect(")"); 1240 if (Tok == "MIN") 1241 return [=] { return std::min(A().getValue(), B().getValue()); }; 1242 return [=] { return std::max(A().getValue(), B().getValue()); }; 1243 } 1244 if (Tok == "ORIGIN") { 1245 StringRef Name = readParenLiteral(); 1246 if (Script->MemoryRegions.count(Name) == 0) { 1247 setError("memory region not defined: " + Name); 1248 return [] { return 0; }; 1249 } 1250 return [=] { return Script->MemoryRegions[Name]->Origin; }; 1251 } 1252 if (Tok == "SEGMENT_START") { 1253 expect("("); 1254 skip(); 1255 expect(","); 1256 Expr E = readExpr(); 1257 expect(")"); 1258 return [=] { return E(); }; 1259 } 1260 if (Tok == "SIZEOF") { 1261 StringRef Name = readParenLiteral(); 1262 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1263 // Linker script does not create an output section if its content is empty. 1264 // We want to allow SIZEOF(.foo) where .foo is a section which happened to 1265 // be empty. 1266 return [=] { return Cmd->Size; }; 1267 } 1268 if (Tok == "SIZEOF_HEADERS") 1269 return [=] { return elf::getHeaderSize(); }; 1270 1271 // Tok is the dot. 1272 if (Tok == ".") 1273 return [=] { return Script->getSymbolValue(Tok, Location); }; 1274 1275 // Tok is a literal number. 1276 if (Optional<uint64_t> Val = parseInt(Tok)) 1277 return [=] { return *Val; }; 1278 1279 // Tok is a symbol name. 1280 if (!isValidCIdentifier(Tok)) 1281 setError("malformed number: " + Tok); 1282 Script->ReferencedSymbols.push_back(Tok); 1283 return [=] { return Script->getSymbolValue(Tok, Location); }; 1284 } 1285 1286 Expr ScriptParser::readTernary(Expr Cond) { 1287 Expr L = readExpr(); 1288 expect(":"); 1289 Expr R = readExpr(); 1290 return [=] { return Cond().getValue() ? L() : R(); }; 1291 } 1292 1293 Expr ScriptParser::readParenExpr() { 1294 expect("("); 1295 Expr E = readExpr(); 1296 expect(")"); 1297 return E; 1298 } 1299 1300 std::vector<StringRef> ScriptParser::readOutputSectionPhdrs() { 1301 std::vector<StringRef> Phdrs; 1302 while (!errorCount() && peek().startswith(":")) { 1303 StringRef Tok = next(); 1304 Phdrs.push_back((Tok.size() == 1) ? next() : Tok.substr(1)); 1305 } 1306 return Phdrs; 1307 } 1308 1309 // Read a program header type name. The next token must be a 1310 // name of a program header type or a constant (e.g. "0x3"). 1311 unsigned ScriptParser::readPhdrType() { 1312 StringRef Tok = next(); 1313 if (Optional<uint64_t> Val = parseInt(Tok)) 1314 return *Val; 1315 1316 unsigned Ret = StringSwitch<unsigned>(Tok) 1317 .Case("PT_NULL", PT_NULL) 1318 .Case("PT_LOAD", PT_LOAD) 1319 .Case("PT_DYNAMIC", PT_DYNAMIC) 1320 .Case("PT_INTERP", PT_INTERP) 1321 .Case("PT_NOTE", PT_NOTE) 1322 .Case("PT_SHLIB", PT_SHLIB) 1323 .Case("PT_PHDR", PT_PHDR) 1324 .Case("PT_TLS", PT_TLS) 1325 .Case("PT_GNU_EH_FRAME", PT_GNU_EH_FRAME) 1326 .Case("PT_GNU_STACK", PT_GNU_STACK) 1327 .Case("PT_GNU_RELRO", PT_GNU_RELRO) 1328 .Case("PT_OPENBSD_RANDOMIZE", PT_OPENBSD_RANDOMIZE) 1329 .Case("PT_OPENBSD_WXNEEDED", PT_OPENBSD_WXNEEDED) 1330 .Case("PT_OPENBSD_BOOTDATA", PT_OPENBSD_BOOTDATA) 1331 .Default(-1); 1332 1333 if (Ret == (unsigned)-1) { 1334 setError("invalid program header type: " + Tok); 1335 return PT_NULL; 1336 } 1337 return Ret; 1338 } 1339 1340 // Reads an anonymous version declaration. 1341 void ScriptParser::readAnonymousDeclaration() { 1342 std::vector<SymbolVersion> Locals; 1343 std::vector<SymbolVersion> Globals; 1344 std::tie(Locals, Globals) = readSymbols(); 1345 1346 for (SymbolVersion V : Locals) { 1347 if (V.Name == "*") 1348 Config->DefaultSymbolVersion = VER_NDX_LOCAL; 1349 else 1350 Config->VersionScriptLocals.push_back(V); 1351 } 1352 1353 for (SymbolVersion V : Globals) 1354 Config->VersionScriptGlobals.push_back(V); 1355 1356 expect(";"); 1357 } 1358 1359 // Reads a non-anonymous version definition, 1360 // e.g. "VerStr { global: foo; bar; local: *; };". 1361 void ScriptParser::readVersionDeclaration(StringRef VerStr) { 1362 // Read a symbol list. 1363 std::vector<SymbolVersion> Locals; 1364 std::vector<SymbolVersion> Globals; 1365 std::tie(Locals, Globals) = readSymbols(); 1366 1367 for (SymbolVersion V : Locals) { 1368 if (V.Name == "*") 1369 Config->DefaultSymbolVersion = VER_NDX_LOCAL; 1370 else 1371 Config->VersionScriptLocals.push_back(V); 1372 } 1373 1374 // Create a new version definition and add that to the global symbols. 1375 VersionDefinition Ver; 1376 Ver.Name = VerStr; 1377 Ver.Globals = Globals; 1378 1379 // User-defined version number starts from 2 because 0 and 1 are 1380 // reserved for VER_NDX_LOCAL and VER_NDX_GLOBAL, respectively. 1381 Ver.Id = Config->VersionDefinitions.size() + 2; 1382 Config->VersionDefinitions.push_back(Ver); 1383 1384 // Each version may have a parent version. For example, "Ver2" 1385 // defined as "Ver2 { global: foo; local: *; } Ver1;" has "Ver1" 1386 // as a parent. This version hierarchy is, probably against your 1387 // instinct, purely for hint; the runtime doesn't care about it 1388 // at all. In LLD, we simply ignore it. 1389 if (peek() != ";") 1390 skip(); 1391 expect(";"); 1392 } 1393 1394 static bool hasWildcard(StringRef S) { 1395 return S.find_first_of("?*[") != StringRef::npos; 1396 } 1397 1398 // Reads a list of symbols, e.g. "{ global: foo; bar; local: *; };". 1399 std::pair<std::vector<SymbolVersion>, std::vector<SymbolVersion>> 1400 ScriptParser::readSymbols() { 1401 std::vector<SymbolVersion> Locals; 1402 std::vector<SymbolVersion> Globals; 1403 std::vector<SymbolVersion> *V = &Globals; 1404 1405 while (!errorCount()) { 1406 if (consume("}")) 1407 break; 1408 if (consumeLabel("local")) { 1409 V = &Locals; 1410 continue; 1411 } 1412 if (consumeLabel("global")) { 1413 V = &Globals; 1414 continue; 1415 } 1416 1417 if (consume("extern")) { 1418 std::vector<SymbolVersion> Ext = readVersionExtern(); 1419 V->insert(V->end(), Ext.begin(), Ext.end()); 1420 } else { 1421 StringRef Tok = next(); 1422 V->push_back({unquote(Tok), false, hasWildcard(Tok)}); 1423 } 1424 expect(";"); 1425 } 1426 return {Locals, Globals}; 1427 } 1428 1429 // Reads an "extern C++" directive, e.g., 1430 // "extern "C++" { ns::*; "f(int, double)"; };" 1431 // 1432 // The last semicolon is optional. E.g. this is OK: 1433 // "extern "C++" { ns::*; "f(int, double)" };" 1434 std::vector<SymbolVersion> ScriptParser::readVersionExtern() { 1435 StringRef Tok = next(); 1436 bool IsCXX = Tok == "\"C++\""; 1437 if (!IsCXX && Tok != "\"C\"") 1438 setError("Unknown language"); 1439 expect("{"); 1440 1441 std::vector<SymbolVersion> Ret; 1442 while (!errorCount() && peek() != "}") { 1443 StringRef Tok = next(); 1444 Ret.push_back( 1445 {unquote(Tok), IsCXX, !Tok.startswith("\"") && hasWildcard(Tok)}); 1446 if (consume("}")) 1447 return Ret; 1448 expect(";"); 1449 } 1450 1451 expect("}"); 1452 return Ret; 1453 } 1454 1455 uint64_t ScriptParser::readMemoryAssignment(StringRef S1, StringRef S2, 1456 StringRef S3) { 1457 if (!consume(S1) && !consume(S2) && !consume(S3)) { 1458 setError("expected one of: " + S1 + ", " + S2 + ", or " + S3); 1459 return 0; 1460 } 1461 expect("="); 1462 return readExpr()().getValue(); 1463 } 1464 1465 // Parse the MEMORY command as specified in: 1466 // https://sourceware.org/binutils/docs/ld/MEMORY.html 1467 // 1468 // MEMORY { name [(attr)] : ORIGIN = origin, LENGTH = len ... } 1469 void ScriptParser::readMemory() { 1470 expect("{"); 1471 while (!errorCount() && !consume("}")) { 1472 StringRef Tok = next(); 1473 if (Tok == "INCLUDE") { 1474 readInclude(); 1475 continue; 1476 } 1477 1478 uint32_t Flags = 0; 1479 uint32_t NegFlags = 0; 1480 if (consume("(")) { 1481 std::tie(Flags, NegFlags) = readMemoryAttributes(); 1482 expect(")"); 1483 } 1484 expect(":"); 1485 1486 uint64_t Origin = readMemoryAssignment("ORIGIN", "org", "o"); 1487 expect(","); 1488 uint64_t Length = readMemoryAssignment("LENGTH", "len", "l"); 1489 1490 // Add the memory region to the region map. 1491 MemoryRegion *MR = make<MemoryRegion>(Tok, Origin, Length, Flags, NegFlags); 1492 if (!Script->MemoryRegions.insert({Tok, MR}).second) 1493 setError("region '" + Tok + "' already defined"); 1494 } 1495 } 1496 1497 // This function parses the attributes used to match against section 1498 // flags when placing output sections in a memory region. These flags 1499 // are only used when an explicit memory region name is not used. 1500 std::pair<uint32_t, uint32_t> ScriptParser::readMemoryAttributes() { 1501 uint32_t Flags = 0; 1502 uint32_t NegFlags = 0; 1503 bool Invert = false; 1504 1505 for (char C : next().lower()) { 1506 uint32_t Flag = 0; 1507 if (C == '!') 1508 Invert = !Invert; 1509 else if (C == 'w') 1510 Flag = SHF_WRITE; 1511 else if (C == 'x') 1512 Flag = SHF_EXECINSTR; 1513 else if (C == 'a') 1514 Flag = SHF_ALLOC; 1515 else if (C != 'r') 1516 setError("invalid memory region attribute"); 1517 1518 if (Invert) 1519 NegFlags |= Flag; 1520 else 1521 Flags |= Flag; 1522 } 1523 return {Flags, NegFlags}; 1524 } 1525 1526 void elf::readLinkerScript(MemoryBufferRef MB) { 1527 ScriptParser(MB).readLinkerScript(); 1528 } 1529 1530 void elf::readVersionScript(MemoryBufferRef MB) { 1531 ScriptParser(MB).readVersionScript(); 1532 } 1533 1534 void elf::readDynamicList(MemoryBufferRef MB) { 1535 ScriptParser(MB).readDynamicList(); 1536 } 1537 1538 void elf::readDefsym(StringRef Name, MemoryBufferRef MB) { 1539 ScriptParser(MB).readDefsym(Name); 1540 } 1541