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 static bool isUnderSysroot(StringRef Path); 44 45 namespace { 46 class ScriptParser final : ScriptLexer { 47 public: 48 ScriptParser(MemoryBufferRef MB) 49 : ScriptLexer(MB), 50 IsUnderSysroot(isUnderSysroot(MB.getBufferIdentifier())) {} 51 52 void readLinkerScript(); 53 void readVersionScript(); 54 void readDynamicList(); 55 void readDefsym(StringRef Name); 56 57 private: 58 void addFile(StringRef Path); 59 60 void readAsNeeded(); 61 void readEntry(); 62 void readExtern(); 63 void readGroup(); 64 void readInclude(); 65 void readInput(); 66 void readMemory(); 67 void readOutput(); 68 void readOutputArch(); 69 void readOutputFormat(); 70 void readPhdrs(); 71 void readRegionAlias(); 72 void readSearchDir(); 73 void readSections(); 74 void readTarget(); 75 void readVersion(); 76 void readVersionScriptCommand(); 77 78 SymbolAssignment *readSymbolAssignment(StringRef Name); 79 ByteCommand *readByteCommand(StringRef Tok); 80 std::array<uint8_t, 4> readFill(); 81 std::array<uint8_t, 4> parseFill(StringRef Tok); 82 bool readSectionDirective(OutputSection *Cmd, StringRef Tok1, StringRef Tok2); 83 void readSectionAddressType(OutputSection *Cmd); 84 OutputSection *readOverlaySectionDescription(); 85 OutputSection *readOutputSectionDescription(StringRef OutSec); 86 std::vector<BaseCommand *> readOverlay(); 87 std::vector<StringRef> readOutputSectionPhdrs(); 88 InputSectionDescription *readInputSectionDescription(StringRef Tok); 89 StringMatcher readFilePatterns(); 90 std::vector<SectionPattern> readInputSectionsList(); 91 InputSectionDescription *readInputSectionRules(StringRef FilePattern); 92 unsigned readPhdrType(); 93 SortSectionPolicy readSortKind(); 94 SymbolAssignment *readProvideHidden(bool Provide, bool Hidden); 95 SymbolAssignment *readAssignment(StringRef Tok); 96 void readSort(); 97 Expr readAssert(); 98 Expr readConstant(); 99 Expr getPageSize(); 100 101 uint64_t readMemoryAssignment(StringRef, StringRef, StringRef); 102 std::pair<uint32_t, uint32_t> readMemoryAttributes(); 103 104 Expr combine(StringRef Op, Expr L, Expr R); 105 Expr readExpr(); 106 Expr readExpr1(Expr Lhs, int MinPrec); 107 StringRef readParenLiteral(); 108 Expr readPrimary(); 109 Expr readTernary(Expr Cond); 110 Expr readParenExpr(); 111 112 // For parsing version script. 113 std::vector<SymbolVersion> readVersionExtern(); 114 void readAnonymousDeclaration(); 115 void readVersionDeclaration(StringRef VerStr); 116 117 std::pair<std::vector<SymbolVersion>, std::vector<SymbolVersion>> 118 readSymbols(); 119 120 // True if a script being read is in a subdirectory specified by -sysroot. 121 bool IsUnderSysroot; 122 123 // A set to detect an INCLUDE() cycle. 124 StringSet<> Seen; 125 }; 126 } // namespace 127 128 static StringRef unquote(StringRef S) { 129 if (S.startswith("\"")) 130 return S.substr(1, S.size() - 2); 131 return S; 132 } 133 134 static bool isUnderSysroot(StringRef Path) { 135 if (Config->Sysroot == "") 136 return false; 137 for (; !Path.empty(); Path = sys::path::parent_path(Path)) 138 if (sys::fs::equivalent(Config->Sysroot, Path)) 139 return true; 140 return false; 141 } 142 143 // Some operations only support one non absolute value. Move the 144 // absolute one to the right hand side for convenience. 145 static void moveAbsRight(ExprValue &A, ExprValue &B) { 146 if (A.Sec == nullptr || (A.ForceAbsolute && !B.isAbsolute())) 147 std::swap(A, B); 148 if (!B.isAbsolute()) 149 error(A.Loc + ": at least one side of the expression must be absolute"); 150 } 151 152 static ExprValue add(ExprValue A, ExprValue B) { 153 moveAbsRight(A, B); 154 return {A.Sec, A.ForceAbsolute, A.getSectionOffset() + B.getValue(), A.Loc}; 155 } 156 157 static ExprValue sub(ExprValue A, ExprValue B) { 158 // The distance between two symbols in sections is absolute. 159 if (!A.isAbsolute() && !B.isAbsolute()) 160 return A.getValue() - B.getValue(); 161 return {A.Sec, false, A.getSectionOffset() - B.getValue(), A.Loc}; 162 } 163 164 static ExprValue bitAnd(ExprValue A, ExprValue B) { 165 moveAbsRight(A, B); 166 return {A.Sec, A.ForceAbsolute, 167 (A.getValue() & B.getValue()) - A.getSecAddr(), A.Loc}; 168 } 169 170 static ExprValue bitOr(ExprValue A, ExprValue B) { 171 moveAbsRight(A, B); 172 return {A.Sec, A.ForceAbsolute, 173 (A.getValue() | B.getValue()) - A.getSecAddr(), A.Loc}; 174 } 175 176 void ScriptParser::readDynamicList() { 177 Config->HasDynamicList = true; 178 expect("{"); 179 std::vector<SymbolVersion> Locals; 180 std::vector<SymbolVersion> Globals; 181 std::tie(Locals, Globals) = readSymbols(); 182 expect(";"); 183 184 if (!atEOF()) { 185 setError("EOF expected, but got " + next()); 186 return; 187 } 188 if (!Locals.empty()) { 189 setError("\"local:\" scope not supported in --dynamic-list"); 190 return; 191 } 192 193 for (SymbolVersion V : Globals) 194 Config->DynamicList.push_back(V); 195 } 196 197 void ScriptParser::readVersionScript() { 198 readVersionScriptCommand(); 199 if (!atEOF()) 200 setError("EOF expected, but got " + next()); 201 } 202 203 void ScriptParser::readVersionScriptCommand() { 204 if (consume("{")) { 205 readAnonymousDeclaration(); 206 return; 207 } 208 209 while (!atEOF() && !errorCount() && peek() != "}") { 210 StringRef VerStr = next(); 211 if (VerStr == "{") { 212 setError("anonymous version definition is used in " 213 "combination with other version definitions"); 214 return; 215 } 216 expect("{"); 217 readVersionDeclaration(VerStr); 218 } 219 } 220 221 void ScriptParser::readVersion() { 222 expect("{"); 223 readVersionScriptCommand(); 224 expect("}"); 225 } 226 227 void ScriptParser::readLinkerScript() { 228 while (!atEOF()) { 229 StringRef Tok = next(); 230 if (Tok == ";") 231 continue; 232 233 if (Tok == "ENTRY") { 234 readEntry(); 235 } else if (Tok == "EXTERN") { 236 readExtern(); 237 } else if (Tok == "GROUP") { 238 readGroup(); 239 } else if (Tok == "INCLUDE") { 240 readInclude(); 241 } else if (Tok == "INPUT") { 242 readInput(); 243 } else if (Tok == "MEMORY") { 244 readMemory(); 245 } else if (Tok == "OUTPUT") { 246 readOutput(); 247 } else if (Tok == "OUTPUT_ARCH") { 248 readOutputArch(); 249 } else if (Tok == "OUTPUT_FORMAT") { 250 readOutputFormat(); 251 } else if (Tok == "PHDRS") { 252 readPhdrs(); 253 } else if (Tok == "REGION_ALIAS") { 254 readRegionAlias(); 255 } else if (Tok == "SEARCH_DIR") { 256 readSearchDir(); 257 } else if (Tok == "SECTIONS") { 258 readSections(); 259 } else if (Tok == "TARGET") { 260 readTarget(); 261 } else if (Tok == "VERSION") { 262 readVersion(); 263 } else if (SymbolAssignment *Cmd = readAssignment(Tok)) { 264 Script->SectionCommands.push_back(Cmd); 265 } else { 266 setError("unknown directive: " + Tok); 267 } 268 } 269 } 270 271 void ScriptParser::readDefsym(StringRef Name) { 272 if (errorCount()) 273 return; 274 Expr E = readExpr(); 275 if (!atEOF()) 276 setError("EOF expected, but got " + next()); 277 SymbolAssignment *Cmd = make<SymbolAssignment>(Name, E, getCurrentLocation()); 278 Script->SectionCommands.push_back(Cmd); 279 } 280 281 void ScriptParser::addFile(StringRef S) { 282 if (IsUnderSysroot && S.startswith("/")) { 283 SmallString<128> PathData; 284 StringRef Path = (Config->Sysroot + S).toStringRef(PathData); 285 if (sys::fs::exists(Path)) { 286 Driver->addFile(Saver.save(Path), /*WithLOption=*/false); 287 return; 288 } 289 } 290 291 if (S.startswith("/")) { 292 Driver->addFile(S, /*WithLOption=*/false); 293 } else if (S.startswith("=")) { 294 if (Config->Sysroot.empty()) 295 Driver->addFile(S.substr(1), /*WithLOption=*/false); 296 else 297 Driver->addFile(Saver.save(Config->Sysroot + "/" + S.substr(1)), 298 /*WithLOption=*/false); 299 } else if (S.startswith("-l")) { 300 Driver->addLibrary(S.substr(2)); 301 } else if (sys::fs::exists(S)) { 302 Driver->addFile(S, /*WithLOption=*/false); 303 } else { 304 if (Optional<std::string> Path = findFromSearchPaths(S)) 305 Driver->addFile(Saver.save(*Path), /*WithLOption=*/true); 306 else 307 setError("unable to find " + S); 308 } 309 } 310 311 void ScriptParser::readAsNeeded() { 312 expect("("); 313 bool Orig = Config->AsNeeded; 314 Config->AsNeeded = true; 315 while (!errorCount() && !consume(")")) 316 addFile(unquote(next())); 317 Config->AsNeeded = Orig; 318 } 319 320 void ScriptParser::readEntry() { 321 // -e <symbol> takes predecence over ENTRY(<symbol>). 322 expect("("); 323 StringRef Tok = next(); 324 if (Config->Entry.empty()) 325 Config->Entry = Tok; 326 expect(")"); 327 } 328 329 void ScriptParser::readExtern() { 330 expect("("); 331 while (!errorCount() && !consume(")")) 332 Config->Undefined.push_back(unquote(next())); 333 } 334 335 void ScriptParser::readGroup() { 336 bool Orig = InputFile::IsInGroup; 337 InputFile::IsInGroup = true; 338 readInput(); 339 InputFile::IsInGroup = Orig; 340 if (!Orig) 341 ++InputFile::NextGroupId; 342 } 343 344 void ScriptParser::readInclude() { 345 StringRef Tok = unquote(next()); 346 347 if (!Seen.insert(Tok).second) { 348 setError("there is a cycle in linker script INCLUDEs"); 349 return; 350 } 351 352 if (Optional<std::string> Path = searchScript(Tok)) { 353 if (Optional<MemoryBufferRef> MB = readFile(*Path)) 354 tokenize(*MB); 355 return; 356 } 357 setError("cannot find linker script " + Tok); 358 } 359 360 void ScriptParser::readInput() { 361 expect("("); 362 while (!errorCount() && !consume(")")) { 363 if (consume("AS_NEEDED")) 364 readAsNeeded(); 365 else 366 addFile(unquote(next())); 367 } 368 } 369 370 void ScriptParser::readOutput() { 371 // -o <file> takes predecence over OUTPUT(<file>). 372 expect("("); 373 StringRef Tok = next(); 374 if (Config->OutputFile.empty()) 375 Config->OutputFile = unquote(Tok); 376 expect(")"); 377 } 378 379 void ScriptParser::readOutputArch() { 380 // OUTPUT_ARCH is ignored for now. 381 expect("("); 382 while (!errorCount() && !consume(")")) 383 skip(); 384 } 385 386 static std::pair<ELFKind, uint16_t> parseBfdName(StringRef S) { 387 return StringSwitch<std::pair<ELFKind, uint16_t>>(S) 388 .Case("elf32-i386", {ELF32LEKind, EM_386}) 389 .Case("elf32-iamcu", {ELF32LEKind, EM_IAMCU}) 390 .Case("elf32-littlearm", {ELF32LEKind, EM_ARM}) 391 .Case("elf32-x86-64", {ELF32LEKind, EM_X86_64}) 392 .Case("elf64-aarch64", {ELF64LEKind, EM_AARCH64}) 393 .Case("elf64-littleaarch64", {ELF64LEKind, EM_AARCH64}) 394 .Case("elf32-powerpc", {ELF32BEKind, EM_PPC}) 395 .Case("elf64-powerpc", {ELF64BEKind, EM_PPC64}) 396 .Case("elf64-powerpcle", {ELF64LEKind, EM_PPC64}) 397 .Case("elf64-x86-64", {ELF64LEKind, EM_X86_64}) 398 .Cases("elf32-tradbigmips", "elf32-bigmips", {ELF32BEKind, EM_MIPS}) 399 .Case("elf32-ntradbigmips", {ELF32BEKind, EM_MIPS}) 400 .Case("elf32-tradlittlemips", {ELF32LEKind, EM_MIPS}) 401 .Case("elf32-ntradlittlemips", {ELF32LEKind, EM_MIPS}) 402 .Case("elf64-tradbigmips", {ELF64BEKind, EM_MIPS}) 403 .Case("elf64-tradlittlemips", {ELF64LEKind, EM_MIPS}) 404 .Default({ELFNoneKind, EM_NONE}); 405 } 406 407 // Parse OUTPUT_FORMAT(bfdname) or OUTPUT_FORMAT(bfdname, big, little). 408 // Currently we ignore big and little parameters. 409 void ScriptParser::readOutputFormat() { 410 expect("("); 411 412 StringRef Name = unquote(next()); 413 StringRef S = Name; 414 if (S.consume_back("-freebsd")) 415 Config->OSABI = ELFOSABI_FREEBSD; 416 417 std::tie(Config->EKind, Config->EMachine) = parseBfdName(S); 418 if (Config->EMachine == EM_NONE) 419 setError("unknown output format name: " + Name); 420 if (S == "elf32-ntradlittlemips" || S == "elf32-ntradbigmips") 421 Config->MipsN32Abi = true; 422 423 if (consume(")")) 424 return; 425 expect(","); 426 skip(); 427 expect(","); 428 skip(); 429 expect(")"); 430 } 431 432 void ScriptParser::readPhdrs() { 433 expect("{"); 434 435 while (!errorCount() && !consume("}")) { 436 PhdrsCommand Cmd; 437 Cmd.Name = next(); 438 Cmd.Type = readPhdrType(); 439 440 while (!errorCount() && !consume(";")) { 441 if (consume("FILEHDR")) 442 Cmd.HasFilehdr = true; 443 else if (consume("PHDRS")) 444 Cmd.HasPhdrs = true; 445 else if (consume("AT")) 446 Cmd.LMAExpr = readParenExpr(); 447 else if (consume("FLAGS")) 448 Cmd.Flags = readParenExpr()().getValue(); 449 else 450 setError("unexpected header attribute: " + next()); 451 } 452 453 Script->PhdrsCommands.push_back(Cmd); 454 } 455 } 456 457 void ScriptParser::readRegionAlias() { 458 expect("("); 459 StringRef Alias = unquote(next()); 460 expect(","); 461 StringRef Name = next(); 462 expect(")"); 463 464 if (Script->MemoryRegions.count(Alias)) 465 setError("redefinition of memory region '" + Alias + "'"); 466 if (!Script->MemoryRegions.count(Name)) 467 setError("memory region '" + Name + "' is not defined"); 468 Script->MemoryRegions.insert({Alias, Script->MemoryRegions[Name]}); 469 } 470 471 void ScriptParser::readSearchDir() { 472 expect("("); 473 StringRef Tok = next(); 474 if (!Config->Nostdlib) 475 Config->SearchPaths.push_back(unquote(Tok)); 476 expect(")"); 477 } 478 479 // This reads an overlay description. Overlays are used to describe output 480 // sections that use the same virtual memory range and normally would trigger 481 // linker's sections sanity check failures. 482 // https://sourceware.org/binutils/docs/ld/Overlay-Description.html#Overlay-Description 483 std::vector<BaseCommand *> ScriptParser::readOverlay() { 484 // VA and LMA expressions are optional, though for simplicity of 485 // implementation we assume they are not. That is what OVERLAY was designed 486 // for first of all: to allow sections with overlapping VAs at different LMAs. 487 Expr AddrExpr = readExpr(); 488 expect(":"); 489 expect("AT"); 490 Expr LMAExpr = readParenExpr(); 491 expect("{"); 492 493 std::vector<BaseCommand *> V; 494 OutputSection *Prev = nullptr; 495 while (!errorCount() && !consume("}")) { 496 // VA is the same for all sections. The LMAs are consecutive in memory 497 // starting from the base load address specified. 498 OutputSection *OS = readOverlaySectionDescription(); 499 OS->AddrExpr = AddrExpr; 500 if (Prev) 501 OS->LMAExpr = [=] { return Prev->getLMA() + Prev->Size; }; 502 else 503 OS->LMAExpr = LMAExpr; 504 V.push_back(OS); 505 Prev = OS; 506 } 507 508 // According to the specification, at the end of the overlay, the location 509 // counter should be equal to the overlay base address plus size of the 510 // largest section seen in the overlay. 511 // Here we want to create the Dot assignment command to achieve that. 512 Expr MoveDot = [=] { 513 uint64_t Max = 0; 514 for (BaseCommand *Cmd : V) 515 Max = std::max(Max, cast<OutputSection>(Cmd)->Size); 516 return AddrExpr().getValue() + Max; 517 }; 518 V.push_back(make<SymbolAssignment>(".", MoveDot, getCurrentLocation())); 519 return V; 520 } 521 522 void ScriptParser::readSections() { 523 Script->HasSectionsCommand = true; 524 525 // -no-rosegment is used to avoid placing read only non-executable sections in 526 // their own segment. We do the same if SECTIONS command is present in linker 527 // script. See comment for computeFlags(). 528 Config->SingleRoRx = true; 529 530 expect("{"); 531 std::vector<BaseCommand *> V; 532 while (!errorCount() && !consume("}")) { 533 StringRef Tok = next(); 534 if (Tok == "OVERLAY") { 535 for (BaseCommand *Cmd : readOverlay()) 536 V.push_back(Cmd); 537 continue; 538 } else if (Tok == "INCLUDE") { 539 readInclude(); 540 continue; 541 } 542 543 if (BaseCommand *Cmd = readAssignment(Tok)) 544 V.push_back(Cmd); 545 else 546 V.push_back(readOutputSectionDescription(Tok)); 547 } 548 549 if (!atEOF() && consume("INSERT")) { 550 std::vector<BaseCommand *> *Dest = nullptr; 551 if (consume("AFTER")) 552 Dest = &Script->InsertAfterCommands[next()]; 553 else if (consume("BEFORE")) 554 Dest = &Script->InsertBeforeCommands[next()]; 555 else 556 setError("expected AFTER/BEFORE, but got '" + next() + "'"); 557 if (Dest) 558 Dest->insert(Dest->end(), V.begin(), V.end()); 559 return; 560 } 561 562 Script->SectionCommands.insert(Script->SectionCommands.end(), V.begin(), 563 V.end()); 564 } 565 566 void ScriptParser::readTarget() { 567 // TARGET(foo) is an alias for "--format foo". Unlike GNU linkers, 568 // we accept only a limited set of BFD names (i.e. "elf" or "binary") 569 // for --format. We recognize only /^elf/ and "binary" in the linker 570 // script as well. 571 expect("("); 572 StringRef Tok = next(); 573 expect(")"); 574 575 if (Tok.startswith("elf")) 576 Config->FormatBinary = false; 577 else if (Tok == "binary") 578 Config->FormatBinary = true; 579 else 580 setError("unknown target: " + Tok); 581 } 582 583 static int precedence(StringRef Op) { 584 return StringSwitch<int>(Op) 585 .Cases("*", "/", "%", 8) 586 .Cases("+", "-", 7) 587 .Cases("<<", ">>", 6) 588 .Cases("<", "<=", ">", ">=", "==", "!=", 5) 589 .Case("&", 4) 590 .Case("|", 3) 591 .Case("&&", 2) 592 .Case("||", 1) 593 .Default(-1); 594 } 595 596 StringMatcher ScriptParser::readFilePatterns() { 597 std::vector<StringRef> V; 598 while (!errorCount() && !consume(")")) 599 V.push_back(next()); 600 return StringMatcher(V); 601 } 602 603 SortSectionPolicy ScriptParser::readSortKind() { 604 if (consume("SORT") || consume("SORT_BY_NAME")) 605 return SortSectionPolicy::Name; 606 if (consume("SORT_BY_ALIGNMENT")) 607 return SortSectionPolicy::Alignment; 608 if (consume("SORT_BY_INIT_PRIORITY")) 609 return SortSectionPolicy::Priority; 610 if (consume("SORT_NONE")) 611 return SortSectionPolicy::None; 612 return SortSectionPolicy::Default; 613 } 614 615 // Reads SECTIONS command contents in the following form: 616 // 617 // <contents> ::= <elem>* 618 // <elem> ::= <exclude>? <glob-pattern> 619 // <exclude> ::= "EXCLUDE_FILE" "(" <glob-pattern>+ ")" 620 // 621 // For example, 622 // 623 // *(.foo EXCLUDE_FILE (a.o) .bar EXCLUDE_FILE (b.o) .baz) 624 // 625 // is parsed as ".foo", ".bar" with "a.o", and ".baz" with "b.o". 626 // The semantics of that is section .foo in any file, section .bar in 627 // any file but a.o, and section .baz in any file but b.o. 628 std::vector<SectionPattern> ScriptParser::readInputSectionsList() { 629 std::vector<SectionPattern> Ret; 630 while (!errorCount() && peek() != ")") { 631 StringMatcher ExcludeFilePat; 632 if (consume("EXCLUDE_FILE")) { 633 expect("("); 634 ExcludeFilePat = readFilePatterns(); 635 } 636 637 std::vector<StringRef> V; 638 while (!errorCount() && peek() != ")" && peek() != "EXCLUDE_FILE") 639 V.push_back(unquote(next())); 640 641 if (!V.empty()) 642 Ret.push_back({std::move(ExcludeFilePat), StringMatcher(V)}); 643 else 644 setError("section pattern is expected"); 645 } 646 return Ret; 647 } 648 649 // Reads contents of "SECTIONS" directive. That directive contains a 650 // list of glob patterns for input sections. The grammar is as follows. 651 // 652 // <patterns> ::= <section-list> 653 // | <sort> "(" <section-list> ")" 654 // | <sort> "(" <sort> "(" <section-list> ")" ")" 655 // 656 // <sort> ::= "SORT" | "SORT_BY_NAME" | "SORT_BY_ALIGNMENT" 657 // | "SORT_BY_INIT_PRIORITY" | "SORT_NONE" 658 // 659 // <section-list> is parsed by readInputSectionsList(). 660 InputSectionDescription * 661 ScriptParser::readInputSectionRules(StringRef FilePattern) { 662 auto *Cmd = make<InputSectionDescription>(FilePattern); 663 expect("("); 664 665 while (!errorCount() && !consume(")")) { 666 SortSectionPolicy Outer = readSortKind(); 667 SortSectionPolicy Inner = SortSectionPolicy::Default; 668 std::vector<SectionPattern> V; 669 if (Outer != SortSectionPolicy::Default) { 670 expect("("); 671 Inner = readSortKind(); 672 if (Inner != SortSectionPolicy::Default) { 673 expect("("); 674 V = readInputSectionsList(); 675 expect(")"); 676 } else { 677 V = readInputSectionsList(); 678 } 679 expect(")"); 680 } else { 681 V = readInputSectionsList(); 682 } 683 684 for (SectionPattern &Pat : V) { 685 Pat.SortInner = Inner; 686 Pat.SortOuter = Outer; 687 } 688 689 std::move(V.begin(), V.end(), std::back_inserter(Cmd->SectionPatterns)); 690 } 691 return Cmd; 692 } 693 694 InputSectionDescription * 695 ScriptParser::readInputSectionDescription(StringRef Tok) { 696 // Input section wildcard can be surrounded by KEEP. 697 // https://sourceware.org/binutils/docs/ld/Input-Section-Keep.html#Input-Section-Keep 698 if (Tok == "KEEP") { 699 expect("("); 700 StringRef FilePattern = next(); 701 InputSectionDescription *Cmd = readInputSectionRules(FilePattern); 702 expect(")"); 703 Script->KeptSections.push_back(Cmd); 704 return Cmd; 705 } 706 return readInputSectionRules(Tok); 707 } 708 709 void ScriptParser::readSort() { 710 expect("("); 711 expect("CONSTRUCTORS"); 712 expect(")"); 713 } 714 715 Expr ScriptParser::readAssert() { 716 expect("("); 717 Expr E = readExpr(); 718 expect(","); 719 StringRef Msg = unquote(next()); 720 expect(")"); 721 722 return [=] { 723 if (!E().getValue()) 724 error(Msg); 725 return Script->getDot(); 726 }; 727 } 728 729 // Reads a FILL(expr) command. We handle the FILL command as an 730 // alias for =fillexp section attribute, which is different from 731 // what GNU linkers do. 732 // https://sourceware.org/binutils/docs/ld/Output-Section-Data.html 733 std::array<uint8_t, 4> ScriptParser::readFill() { 734 expect("("); 735 std::array<uint8_t, 4> V = parseFill(next()); 736 expect(")"); 737 return V; 738 } 739 740 // Tries to read the special directive for an output section definition which 741 // can be one of following: "(NOLOAD)", "(COPY)", "(INFO)" or "(OVERLAY)". 742 // Tok1 and Tok2 are next 2 tokens peeked. See comment for readSectionAddressType below. 743 bool ScriptParser::readSectionDirective(OutputSection *Cmd, StringRef Tok1, StringRef Tok2) { 744 if (Tok1 != "(") 745 return false; 746 if (Tok2 != "NOLOAD" && Tok2 != "COPY" && Tok2 != "INFO" && Tok2 != "OVERLAY") 747 return false; 748 749 expect("("); 750 if (consume("NOLOAD")) { 751 Cmd->Noload = true; 752 } else { 753 skip(); // This is "COPY", "INFO" or "OVERLAY". 754 Cmd->NonAlloc = true; 755 } 756 expect(")"); 757 return true; 758 } 759 760 // Reads an expression and/or the special directive for an output 761 // section definition. Directive is one of following: "(NOLOAD)", 762 // "(COPY)", "(INFO)" or "(OVERLAY)". 763 // 764 // An output section name can be followed by an address expression 765 // and/or directive. This grammar is not LL(1) because "(" can be 766 // interpreted as either the beginning of some expression or beginning 767 // of directive. 768 // 769 // https://sourceware.org/binutils/docs/ld/Output-Section-Address.html 770 // https://sourceware.org/binutils/docs/ld/Output-Section-Type.html 771 void ScriptParser::readSectionAddressType(OutputSection *Cmd) { 772 if (readSectionDirective(Cmd, peek(), peek2())) 773 return; 774 775 Cmd->AddrExpr = readExpr(); 776 if (peek() == "(" && !readSectionDirective(Cmd, "(", peek2())) 777 setError("unknown section directive: " + peek2()); 778 } 779 780 static Expr checkAlignment(Expr E, std::string &Loc) { 781 return [=] { 782 uint64_t Alignment = std::max((uint64_t)1, E().getValue()); 783 if (!isPowerOf2_64(Alignment)) { 784 error(Loc + ": alignment must be power of 2"); 785 return (uint64_t)1; // Return a dummy value. 786 } 787 return Alignment; 788 }; 789 } 790 791 OutputSection *ScriptParser::readOverlaySectionDescription() { 792 OutputSection *Cmd = 793 Script->createOutputSection(next(), getCurrentLocation()); 794 Cmd->InOverlay = true; 795 expect("{"); 796 while (!errorCount() && !consume("}")) 797 Cmd->SectionCommands.push_back(readInputSectionRules(next())); 798 Cmd->Phdrs = readOutputSectionPhdrs(); 799 return Cmd; 800 } 801 802 OutputSection *ScriptParser::readOutputSectionDescription(StringRef OutSec) { 803 OutputSection *Cmd = 804 Script->createOutputSection(OutSec, getCurrentLocation()); 805 806 size_t SymbolsReferenced = Script->ReferencedSymbols.size(); 807 808 if (peek() != ":") 809 readSectionAddressType(Cmd); 810 expect(":"); 811 812 std::string Location = getCurrentLocation(); 813 if (consume("AT")) 814 Cmd->LMAExpr = readParenExpr(); 815 if (consume("ALIGN")) 816 Cmd->AlignExpr = checkAlignment(readParenExpr(), Location); 817 if (consume("SUBALIGN")) 818 Cmd->SubalignExpr = checkAlignment(readParenExpr(), Location); 819 820 // Parse constraints. 821 if (consume("ONLY_IF_RO")) 822 Cmd->Constraint = ConstraintKind::ReadOnly; 823 if (consume("ONLY_IF_RW")) 824 Cmd->Constraint = ConstraintKind::ReadWrite; 825 expect("{"); 826 827 while (!errorCount() && !consume("}")) { 828 StringRef Tok = next(); 829 if (Tok == ";") { 830 // Empty commands are allowed. Do nothing here. 831 } else if (SymbolAssignment *Assign = readAssignment(Tok)) { 832 Cmd->SectionCommands.push_back(Assign); 833 } else if (ByteCommand *Data = readByteCommand(Tok)) { 834 Cmd->SectionCommands.push_back(Data); 835 } else if (Tok == "CONSTRUCTORS") { 836 // CONSTRUCTORS is a keyword to make the linker recognize C++ ctors/dtors 837 // by name. This is for very old file formats such as ECOFF/XCOFF. 838 // For ELF, we should ignore. 839 } else if (Tok == "FILL") { 840 Cmd->Filler = readFill(); 841 } else if (Tok == "SORT") { 842 readSort(); 843 } else if (Tok == "INCLUDE") { 844 readInclude(); 845 } else if (peek() == "(") { 846 Cmd->SectionCommands.push_back(readInputSectionDescription(Tok)); 847 } else { 848 // We have a file name and no input sections description. It is not a 849 // commonly used syntax, but still acceptable. In that case, all sections 850 // from the file will be included. 851 auto *ISD = make<InputSectionDescription>(Tok); 852 ISD->SectionPatterns.push_back({{}, StringMatcher({"*"})}); 853 Cmd->SectionCommands.push_back(ISD); 854 } 855 } 856 857 if (consume(">")) 858 Cmd->MemoryRegionName = next(); 859 860 if (consume("AT")) { 861 expect(">"); 862 Cmd->LMARegionName = next(); 863 } 864 865 if (Cmd->LMAExpr && !Cmd->LMARegionName.empty()) 866 error("section can't have both LMA and a load region"); 867 868 Cmd->Phdrs = readOutputSectionPhdrs(); 869 870 if (consume("=")) 871 Cmd->Filler = parseFill(next()); 872 else if (peek().startswith("=")) 873 Cmd->Filler = parseFill(next().drop_front()); 874 875 // Consume optional comma following output section command. 876 consume(","); 877 878 if (Script->ReferencedSymbols.size() > SymbolsReferenced) 879 Cmd->ExpressionsUseSymbols = true; 880 return Cmd; 881 } 882 883 // Parses a given string as a octal/decimal/hexadecimal number and 884 // returns it as a big-endian number. Used for `=<fillexp>`. 885 // https://sourceware.org/binutils/docs/ld/Output-Section-Fill.html 886 // 887 // When reading a hexstring, ld.bfd handles it as a blob of arbitrary 888 // size, while ld.gold always handles it as a 32-bit big-endian number. 889 // We are compatible with ld.gold because it's easier to implement. 890 std::array<uint8_t, 4> ScriptParser::parseFill(StringRef Tok) { 891 uint32_t V = 0; 892 if (!to_integer(Tok, V)) 893 setError("invalid filler expression: " + Tok); 894 895 std::array<uint8_t, 4> Buf; 896 write32be(Buf.data(), V); 897 return Buf; 898 } 899 900 SymbolAssignment *ScriptParser::readProvideHidden(bool Provide, bool Hidden) { 901 expect("("); 902 SymbolAssignment *Cmd = readSymbolAssignment(next()); 903 Cmd->Provide = Provide; 904 Cmd->Hidden = Hidden; 905 expect(")"); 906 return Cmd; 907 } 908 909 SymbolAssignment *ScriptParser::readAssignment(StringRef Tok) { 910 // Assert expression returns Dot, so this is equal to ".=." 911 if (Tok == "ASSERT") 912 return make<SymbolAssignment>(".", readAssert(), getCurrentLocation()); 913 914 size_t OldPos = Pos; 915 SymbolAssignment *Cmd = nullptr; 916 if (peek() == "=" || peek() == "+=") 917 Cmd = readSymbolAssignment(Tok); 918 else if (Tok == "PROVIDE") 919 Cmd = readProvideHidden(true, false); 920 else if (Tok == "HIDDEN") 921 Cmd = readProvideHidden(false, true); 922 else if (Tok == "PROVIDE_HIDDEN") 923 Cmd = readProvideHidden(true, true); 924 925 if (Cmd) { 926 Cmd->CommandString = 927 Tok.str() + " " + 928 llvm::join(Tokens.begin() + OldPos, Tokens.begin() + Pos, " "); 929 expect(";"); 930 } 931 return Cmd; 932 } 933 934 SymbolAssignment *ScriptParser::readSymbolAssignment(StringRef Name) { 935 StringRef Op = next(); 936 assert(Op == "=" || Op == "+="); 937 Expr E = readExpr(); 938 if (Op == "+=") { 939 std::string Loc = getCurrentLocation(); 940 E = [=] { return add(Script->getSymbolValue(Name, Loc), E()); }; 941 } 942 return make<SymbolAssignment>(Name, E, getCurrentLocation()); 943 } 944 945 // This is an operator-precedence parser to parse a linker 946 // script expression. 947 Expr ScriptParser::readExpr() { 948 // Our lexer is context-aware. Set the in-expression bit so that 949 // they apply different tokenization rules. 950 bool Orig = InExpr; 951 InExpr = true; 952 Expr E = readExpr1(readPrimary(), 0); 953 InExpr = Orig; 954 return E; 955 } 956 957 Expr ScriptParser::combine(StringRef Op, Expr L, Expr R) { 958 if (Op == "+") 959 return [=] { return add(L(), R()); }; 960 if (Op == "-") 961 return [=] { return sub(L(), R()); }; 962 if (Op == "*") 963 return [=] { return L().getValue() * R().getValue(); }; 964 if (Op == "/") { 965 std::string Loc = getCurrentLocation(); 966 return [=]() -> uint64_t { 967 if (uint64_t RV = R().getValue()) 968 return L().getValue() / RV; 969 error(Loc + ": division by zero"); 970 return 0; 971 }; 972 } 973 if (Op == "%") { 974 std::string Loc = getCurrentLocation(); 975 return [=]() -> uint64_t { 976 if (uint64_t RV = R().getValue()) 977 return L().getValue() % RV; 978 error(Loc + ": modulo by zero"); 979 return 0; 980 }; 981 } 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 L().getValue() != R().getValue(); }; 998 if (Op == "||") 999 return [=] { return L().getValue() || R().getValue(); }; 1000 if (Op == "&&") 1001 return [=] { return L().getValue() && R().getValue(); }; 1002 if (Op == "&") 1003 return [=] { return bitAnd(L(), R()); }; 1004 if (Op == "|") 1005 return [=] { return bitOr(L(), R()); }; 1006 llvm_unreachable("invalid operator"); 1007 } 1008 1009 // This is a part of the operator-precedence parser. This function 1010 // assumes that the remaining token stream starts with an operator. 1011 Expr ScriptParser::readExpr1(Expr Lhs, int MinPrec) { 1012 while (!atEOF() && !errorCount()) { 1013 // Read an operator and an expression. 1014 if (consume("?")) 1015 return readTernary(Lhs); 1016 StringRef Op1 = peek(); 1017 if (precedence(Op1) < MinPrec) 1018 break; 1019 skip(); 1020 Expr Rhs = readPrimary(); 1021 1022 // Evaluate the remaining part of the expression first if the 1023 // next operator has greater precedence than the previous one. 1024 // For example, if we have read "+" and "3", and if the next 1025 // operator is "*", then we'll evaluate 3 * ... part first. 1026 while (!atEOF()) { 1027 StringRef Op2 = peek(); 1028 if (precedence(Op2) <= precedence(Op1)) 1029 break; 1030 Rhs = readExpr1(Rhs, precedence(Op2)); 1031 } 1032 1033 Lhs = combine(Op1, Lhs, Rhs); 1034 } 1035 return Lhs; 1036 } 1037 1038 Expr ScriptParser::getPageSize() { 1039 std::string Location = getCurrentLocation(); 1040 return [=]() -> uint64_t { 1041 if (Target) 1042 return Config->CommonPageSize; 1043 error(Location + ": unable to calculate page size"); 1044 return 4096; // Return a dummy value. 1045 }; 1046 } 1047 1048 Expr ScriptParser::readConstant() { 1049 StringRef S = readParenLiteral(); 1050 if (S == "COMMONPAGESIZE") 1051 return getPageSize(); 1052 if (S == "MAXPAGESIZE") 1053 return [] { return Config->MaxPageSize; }; 1054 setError("unknown constant: " + S); 1055 return [] { return 0; }; 1056 } 1057 1058 // Parses Tok as an integer. It recognizes hexadecimal (prefixed with 1059 // "0x" or suffixed with "H") and decimal numbers. Decimal numbers may 1060 // have "K" (Ki) or "M" (Mi) suffixes. 1061 static Optional<uint64_t> parseInt(StringRef Tok) { 1062 // Hexadecimal 1063 uint64_t Val; 1064 if (Tok.startswith_lower("0x")) { 1065 if (!to_integer(Tok.substr(2), Val, 16)) 1066 return None; 1067 return Val; 1068 } 1069 if (Tok.endswith_lower("H")) { 1070 if (!to_integer(Tok.drop_back(), Val, 16)) 1071 return None; 1072 return Val; 1073 } 1074 1075 // Decimal 1076 if (Tok.endswith_lower("K")) { 1077 if (!to_integer(Tok.drop_back(), Val, 10)) 1078 return None; 1079 return Val * 1024; 1080 } 1081 if (Tok.endswith_lower("M")) { 1082 if (!to_integer(Tok.drop_back(), Val, 10)) 1083 return None; 1084 return Val * 1024 * 1024; 1085 } 1086 if (!to_integer(Tok, Val, 10)) 1087 return None; 1088 return Val; 1089 } 1090 1091 ByteCommand *ScriptParser::readByteCommand(StringRef Tok) { 1092 int Size = StringSwitch<int>(Tok) 1093 .Case("BYTE", 1) 1094 .Case("SHORT", 2) 1095 .Case("LONG", 4) 1096 .Case("QUAD", 8) 1097 .Default(-1); 1098 if (Size == -1) 1099 return nullptr; 1100 1101 size_t OldPos = Pos; 1102 Expr E = readParenExpr(); 1103 std::string CommandString = 1104 Tok.str() + " " + 1105 llvm::join(Tokens.begin() + OldPos, Tokens.begin() + Pos, " "); 1106 return make<ByteCommand>(E, Size, CommandString); 1107 } 1108 1109 StringRef ScriptParser::readParenLiteral() { 1110 expect("("); 1111 bool Orig = InExpr; 1112 InExpr = false; 1113 StringRef Tok = next(); 1114 InExpr = Orig; 1115 expect(")"); 1116 return Tok; 1117 } 1118 1119 static void checkIfExists(OutputSection *Cmd, StringRef Location) { 1120 if (Cmd->Location.empty() && Script->ErrorOnMissingSection) 1121 error(Location + ": undefined section " + Cmd->Name); 1122 } 1123 1124 Expr ScriptParser::readPrimary() { 1125 if (peek() == "(") 1126 return readParenExpr(); 1127 1128 if (consume("~")) { 1129 Expr E = readPrimary(); 1130 return [=] { return ~E().getValue(); }; 1131 } 1132 if (consume("!")) { 1133 Expr E = readPrimary(); 1134 return [=] { return !E().getValue(); }; 1135 } 1136 if (consume("-")) { 1137 Expr E = readPrimary(); 1138 return [=] { return -E().getValue(); }; 1139 } 1140 1141 StringRef Tok = next(); 1142 std::string Location = getCurrentLocation(); 1143 1144 // Built-in functions are parsed here. 1145 // https://sourceware.org/binutils/docs/ld/Builtin-Functions.html. 1146 if (Tok == "ABSOLUTE") { 1147 Expr Inner = readParenExpr(); 1148 return [=] { 1149 ExprValue I = Inner(); 1150 I.ForceAbsolute = true; 1151 return I; 1152 }; 1153 } 1154 if (Tok == "ADDR") { 1155 StringRef Name = readParenLiteral(); 1156 OutputSection *Sec = Script->getOrCreateOutputSection(Name); 1157 Sec->UsedInExpression = true; 1158 return [=]() -> ExprValue { 1159 checkIfExists(Sec, Location); 1160 return {Sec, false, 0, Location}; 1161 }; 1162 } 1163 if (Tok == "ALIGN") { 1164 expect("("); 1165 Expr E = readExpr(); 1166 if (consume(")")) { 1167 E = checkAlignment(E, Location); 1168 return [=] { return alignTo(Script->getDot(), E().getValue()); }; 1169 } 1170 expect(","); 1171 Expr E2 = checkAlignment(readExpr(), Location); 1172 expect(")"); 1173 return [=] { 1174 ExprValue V = E(); 1175 V.Alignment = E2().getValue(); 1176 return V; 1177 }; 1178 } 1179 if (Tok == "ALIGNOF") { 1180 StringRef Name = readParenLiteral(); 1181 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1182 return [=] { 1183 checkIfExists(Cmd, Location); 1184 return Cmd->Alignment; 1185 }; 1186 } 1187 if (Tok == "ASSERT") 1188 return readAssert(); 1189 if (Tok == "CONSTANT") 1190 return readConstant(); 1191 if (Tok == "DATA_SEGMENT_ALIGN") { 1192 expect("("); 1193 Expr E = readExpr(); 1194 expect(","); 1195 readExpr(); 1196 expect(")"); 1197 return [=] { 1198 return alignTo(Script->getDot(), std::max((uint64_t)1, E().getValue())); 1199 }; 1200 } 1201 if (Tok == "DATA_SEGMENT_END") { 1202 expect("("); 1203 expect("."); 1204 expect(")"); 1205 return [] { return Script->getDot(); }; 1206 } 1207 if (Tok == "DATA_SEGMENT_RELRO_END") { 1208 // GNU linkers implements more complicated logic to handle 1209 // DATA_SEGMENT_RELRO_END. We instead ignore the arguments and 1210 // just align to the next page boundary for simplicity. 1211 expect("("); 1212 readExpr(); 1213 expect(","); 1214 readExpr(); 1215 expect(")"); 1216 Expr E = getPageSize(); 1217 return [=] { return alignTo(Script->getDot(), E().getValue()); }; 1218 } 1219 if (Tok == "DEFINED") { 1220 StringRef Name = readParenLiteral(); 1221 return [=] { return Symtab->find(Name) ? 1 : 0; }; 1222 } 1223 if (Tok == "LENGTH") { 1224 StringRef Name = readParenLiteral(); 1225 if (Script->MemoryRegions.count(Name) == 0) { 1226 setError("memory region not defined: " + Name); 1227 return [] { return 0; }; 1228 } 1229 return [=] { return Script->MemoryRegions[Name]->Length; }; 1230 } 1231 if (Tok == "LOADADDR") { 1232 StringRef Name = readParenLiteral(); 1233 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1234 Cmd->UsedInExpression = true; 1235 return [=] { 1236 checkIfExists(Cmd, Location); 1237 return Cmd->getLMA(); 1238 }; 1239 } 1240 if (Tok == "MAX" || Tok == "MIN") { 1241 expect("("); 1242 Expr A = readExpr(); 1243 expect(","); 1244 Expr B = readExpr(); 1245 expect(")"); 1246 if (Tok == "MIN") 1247 return [=] { return std::min(A().getValue(), B().getValue()); }; 1248 return [=] { return std::max(A().getValue(), B().getValue()); }; 1249 } 1250 if (Tok == "ORIGIN") { 1251 StringRef Name = readParenLiteral(); 1252 if (Script->MemoryRegions.count(Name) == 0) { 1253 setError("memory region not defined: " + Name); 1254 return [] { return 0; }; 1255 } 1256 return [=] { return Script->MemoryRegions[Name]->Origin; }; 1257 } 1258 if (Tok == "SEGMENT_START") { 1259 expect("("); 1260 skip(); 1261 expect(","); 1262 Expr E = readExpr(); 1263 expect(")"); 1264 return [=] { return E(); }; 1265 } 1266 if (Tok == "SIZEOF") { 1267 StringRef Name = readParenLiteral(); 1268 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1269 // Linker script does not create an output section if its content is empty. 1270 // We want to allow SIZEOF(.foo) where .foo is a section which happened to 1271 // be empty. 1272 return [=] { return Cmd->Size; }; 1273 } 1274 if (Tok == "SIZEOF_HEADERS") 1275 return [=] { return elf::getHeaderSize(); }; 1276 1277 // Tok is the dot. 1278 if (Tok == ".") 1279 return [=] { return Script->getSymbolValue(Tok, Location); }; 1280 1281 // Tok is a literal number. 1282 if (Optional<uint64_t> Val = parseInt(Tok)) 1283 return [=] { return *Val; }; 1284 1285 // Tok is a symbol name. 1286 if (!isValidCIdentifier(Tok)) 1287 setError("malformed number: " + Tok); 1288 Script->ReferencedSymbols.push_back(Tok); 1289 return [=] { return Script->getSymbolValue(Tok, Location); }; 1290 } 1291 1292 Expr ScriptParser::readTernary(Expr Cond) { 1293 Expr L = readExpr(); 1294 expect(":"); 1295 Expr R = readExpr(); 1296 return [=] { return Cond().getValue() ? L() : R(); }; 1297 } 1298 1299 Expr ScriptParser::readParenExpr() { 1300 expect("("); 1301 Expr E = readExpr(); 1302 expect(")"); 1303 return E; 1304 } 1305 1306 std::vector<StringRef> ScriptParser::readOutputSectionPhdrs() { 1307 std::vector<StringRef> Phdrs; 1308 while (!errorCount() && peek().startswith(":")) { 1309 StringRef Tok = next(); 1310 Phdrs.push_back((Tok.size() == 1) ? next() : Tok.substr(1)); 1311 } 1312 return Phdrs; 1313 } 1314 1315 // Read a program header type name. The next token must be a 1316 // name of a program header type or a constant (e.g. "0x3"). 1317 unsigned ScriptParser::readPhdrType() { 1318 StringRef Tok = next(); 1319 if (Optional<uint64_t> Val = parseInt(Tok)) 1320 return *Val; 1321 1322 unsigned Ret = StringSwitch<unsigned>(Tok) 1323 .Case("PT_NULL", PT_NULL) 1324 .Case("PT_LOAD", PT_LOAD) 1325 .Case("PT_DYNAMIC", PT_DYNAMIC) 1326 .Case("PT_INTERP", PT_INTERP) 1327 .Case("PT_NOTE", PT_NOTE) 1328 .Case("PT_SHLIB", PT_SHLIB) 1329 .Case("PT_PHDR", PT_PHDR) 1330 .Case("PT_TLS", PT_TLS) 1331 .Case("PT_GNU_EH_FRAME", PT_GNU_EH_FRAME) 1332 .Case("PT_GNU_STACK", PT_GNU_STACK) 1333 .Case("PT_GNU_RELRO", PT_GNU_RELRO) 1334 .Case("PT_OPENBSD_RANDOMIZE", PT_OPENBSD_RANDOMIZE) 1335 .Case("PT_OPENBSD_WXNEEDED", PT_OPENBSD_WXNEEDED) 1336 .Case("PT_OPENBSD_BOOTDATA", PT_OPENBSD_BOOTDATA) 1337 .Default(-1); 1338 1339 if (Ret == (unsigned)-1) { 1340 setError("invalid program header type: " + Tok); 1341 return PT_NULL; 1342 } 1343 return Ret; 1344 } 1345 1346 // Reads an anonymous version declaration. 1347 void ScriptParser::readAnonymousDeclaration() { 1348 std::vector<SymbolVersion> Locals; 1349 std::vector<SymbolVersion> Globals; 1350 std::tie(Locals, Globals) = readSymbols(); 1351 1352 for (SymbolVersion V : Locals) { 1353 if (V.Name == "*") 1354 Config->DefaultSymbolVersion = VER_NDX_LOCAL; 1355 else 1356 Config->VersionScriptLocals.push_back(V); 1357 } 1358 1359 for (SymbolVersion V : Globals) 1360 Config->VersionScriptGlobals.push_back(V); 1361 1362 expect(";"); 1363 } 1364 1365 // Reads a non-anonymous version definition, 1366 // e.g. "VerStr { global: foo; bar; local: *; };". 1367 void ScriptParser::readVersionDeclaration(StringRef VerStr) { 1368 // Read a symbol list. 1369 std::vector<SymbolVersion> Locals; 1370 std::vector<SymbolVersion> Globals; 1371 std::tie(Locals, Globals) = readSymbols(); 1372 1373 for (SymbolVersion V : Locals) { 1374 if (V.Name == "*") 1375 Config->DefaultSymbolVersion = VER_NDX_LOCAL; 1376 else 1377 Config->VersionScriptLocals.push_back(V); 1378 } 1379 1380 // Create a new version definition and add that to the global symbols. 1381 VersionDefinition Ver; 1382 Ver.Name = VerStr; 1383 Ver.Globals = Globals; 1384 1385 // User-defined version number starts from 2 because 0 and 1 are 1386 // reserved for VER_NDX_LOCAL and VER_NDX_GLOBAL, respectively. 1387 Ver.Id = Config->VersionDefinitions.size() + 2; 1388 Config->VersionDefinitions.push_back(Ver); 1389 1390 // Each version may have a parent version. For example, "Ver2" 1391 // defined as "Ver2 { global: foo; local: *; } Ver1;" has "Ver1" 1392 // as a parent. This version hierarchy is, probably against your 1393 // instinct, purely for hint; the runtime doesn't care about it 1394 // at all. In LLD, we simply ignore it. 1395 if (peek() != ";") 1396 skip(); 1397 expect(";"); 1398 } 1399 1400 static bool hasWildcard(StringRef S) { 1401 return S.find_first_of("?*[") != StringRef::npos; 1402 } 1403 1404 // Reads a list of symbols, e.g. "{ global: foo; bar; local: *; };". 1405 std::pair<std::vector<SymbolVersion>, std::vector<SymbolVersion>> 1406 ScriptParser::readSymbols() { 1407 std::vector<SymbolVersion> Locals; 1408 std::vector<SymbolVersion> Globals; 1409 std::vector<SymbolVersion> *V = &Globals; 1410 1411 while (!errorCount()) { 1412 if (consume("}")) 1413 break; 1414 if (consumeLabel("local")) { 1415 V = &Locals; 1416 continue; 1417 } 1418 if (consumeLabel("global")) { 1419 V = &Globals; 1420 continue; 1421 } 1422 1423 if (consume("extern")) { 1424 std::vector<SymbolVersion> Ext = readVersionExtern(); 1425 V->insert(V->end(), Ext.begin(), Ext.end()); 1426 } else { 1427 StringRef Tok = next(); 1428 V->push_back({unquote(Tok), false, hasWildcard(Tok)}); 1429 } 1430 expect(";"); 1431 } 1432 return {Locals, Globals}; 1433 } 1434 1435 // Reads an "extern C++" directive, e.g., 1436 // "extern "C++" { ns::*; "f(int, double)"; };" 1437 // 1438 // The last semicolon is optional. E.g. this is OK: 1439 // "extern "C++" { ns::*; "f(int, double)" };" 1440 std::vector<SymbolVersion> ScriptParser::readVersionExtern() { 1441 StringRef Tok = next(); 1442 bool IsCXX = Tok == "\"C++\""; 1443 if (!IsCXX && Tok != "\"C\"") 1444 setError("Unknown language"); 1445 expect("{"); 1446 1447 std::vector<SymbolVersion> Ret; 1448 while (!errorCount() && peek() != "}") { 1449 StringRef Tok = next(); 1450 bool HasWildcard = !Tok.startswith("\"") && hasWildcard(Tok); 1451 Ret.push_back({unquote(Tok), IsCXX, HasWildcard}); 1452 if (consume("}")) 1453 return Ret; 1454 expect(";"); 1455 } 1456 1457 expect("}"); 1458 return Ret; 1459 } 1460 1461 uint64_t ScriptParser::readMemoryAssignment(StringRef S1, StringRef S2, 1462 StringRef S3) { 1463 if (!consume(S1) && !consume(S2) && !consume(S3)) { 1464 setError("expected one of: " + S1 + ", " + S2 + ", or " + S3); 1465 return 0; 1466 } 1467 expect("="); 1468 return readExpr()().getValue(); 1469 } 1470 1471 // Parse the MEMORY command as specified in: 1472 // https://sourceware.org/binutils/docs/ld/MEMORY.html 1473 // 1474 // MEMORY { name [(attr)] : ORIGIN = origin, LENGTH = len ... } 1475 void ScriptParser::readMemory() { 1476 expect("{"); 1477 while (!errorCount() && !consume("}")) { 1478 StringRef Tok = next(); 1479 if (Tok == "INCLUDE") { 1480 readInclude(); 1481 continue; 1482 } 1483 1484 uint32_t Flags = 0; 1485 uint32_t NegFlags = 0; 1486 if (consume("(")) { 1487 std::tie(Flags, NegFlags) = readMemoryAttributes(); 1488 expect(")"); 1489 } 1490 expect(":"); 1491 1492 uint64_t Origin = readMemoryAssignment("ORIGIN", "org", "o"); 1493 expect(","); 1494 uint64_t Length = readMemoryAssignment("LENGTH", "len", "l"); 1495 1496 // Add the memory region to the region map. 1497 MemoryRegion *MR = make<MemoryRegion>(Tok, Origin, Length, Flags, NegFlags); 1498 if (!Script->MemoryRegions.insert({Tok, MR}).second) 1499 setError("region '" + Tok + "' already defined"); 1500 } 1501 } 1502 1503 // This function parses the attributes used to match against section 1504 // flags when placing output sections in a memory region. These flags 1505 // are only used when an explicit memory region name is not used. 1506 std::pair<uint32_t, uint32_t> ScriptParser::readMemoryAttributes() { 1507 uint32_t Flags = 0; 1508 uint32_t NegFlags = 0; 1509 bool Invert = false; 1510 1511 for (char C : next().lower()) { 1512 uint32_t Flag = 0; 1513 if (C == '!') 1514 Invert = !Invert; 1515 else if (C == 'w') 1516 Flag = SHF_WRITE; 1517 else if (C == 'x') 1518 Flag = SHF_EXECINSTR; 1519 else if (C == 'a') 1520 Flag = SHF_ALLOC; 1521 else if (C != 'r') 1522 setError("invalid memory region attribute"); 1523 1524 if (Invert) 1525 NegFlags |= Flag; 1526 else 1527 Flags |= Flag; 1528 } 1529 return {Flags, NegFlags}; 1530 } 1531 1532 void elf::readLinkerScript(MemoryBufferRef MB) { 1533 ScriptParser(MB).readLinkerScript(); 1534 } 1535 1536 void elf::readVersionScript(MemoryBufferRef MB) { 1537 ScriptParser(MB).readVersionScript(); 1538 } 1539 1540 void elf::readDynamicList(MemoryBufferRef MB) { 1541 ScriptParser(MB).readDynamicList(); 1542 } 1543 1544 void elf::readDefsym(StringRef Name, MemoryBufferRef MB) { 1545 ScriptParser(MB).readDefsym(Name); 1546 } 1547