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(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("elf64-powerpc", {ELF64BEKind, EM_PPC64}) 395 .Case("elf64-powerpcle", {ELF64LEKind, EM_PPC64}) 396 .Case("elf64-x86-64", {ELF64LEKind, EM_X86_64}) 397 .Case("elf32-tradbigmips", {ELF32BEKind, EM_MIPS}) 398 .Case("elf32-ntradbigmips", {ELF32BEKind, EM_MIPS}) 399 .Case("elf32-tradlittlemips", {ELF32LEKind, EM_MIPS}) 400 .Case("elf32-ntradlittlemips", {ELF32LEKind, EM_MIPS}) 401 .Case("elf64-tradbigmips", {ELF64BEKind, EM_MIPS}) 402 .Case("elf64-tradlittlemips", {ELF64LEKind, EM_MIPS}) 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(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 // Reads a FILL(expr) command. We handle the FILL command as an 729 // alias for =fillexp section attribute, which is different from 730 // what GNU linkers do. 731 // https://sourceware.org/binutils/docs/ld/Output-Section-Data.html 732 std::array<uint8_t, 4> ScriptParser::readFill() { 733 expect("("); 734 std::array<uint8_t, 4> V = parseFill(next()); 735 expect(")"); 736 return V; 737 } 738 739 // Tries to read the special directive for an output section definition which 740 // can be one of following: "(NOLOAD)", "(COPY)", "(INFO)" or "(OVERLAY)". 741 // Tok1 and Tok2 are next 2 tokens peeked. See comment for readSectionAddressType below. 742 bool ScriptParser::readSectionDirective(OutputSection *Cmd, StringRef Tok1, StringRef Tok2) { 743 if (Tok1 != "(") 744 return false; 745 if (Tok2 != "NOLOAD" && Tok2 != "COPY" && Tok2 != "INFO" && Tok2 != "OVERLAY") 746 return false; 747 748 expect("("); 749 if (consume("NOLOAD")) { 750 Cmd->Noload = true; 751 } else { 752 skip(); // This is "COPY", "INFO" or "OVERLAY". 753 Cmd->NonAlloc = true; 754 } 755 expect(")"); 756 return true; 757 } 758 759 // Reads an expression and/or the special directive for an output 760 // section definition. Directive is one of following: "(NOLOAD)", 761 // "(COPY)", "(INFO)" or "(OVERLAY)". 762 // 763 // An output section name can be followed by an address expression 764 // and/or directive. This grammar is not LL(1) because "(" can be 765 // interpreted as either the beginning of some expression or beginning 766 // of directive. 767 // 768 // https://sourceware.org/binutils/docs/ld/Output-Section-Address.html 769 // https://sourceware.org/binutils/docs/ld/Output-Section-Type.html 770 void ScriptParser::readSectionAddressType(OutputSection *Cmd) { 771 if (readSectionDirective(Cmd, peek(), peek2())) 772 return; 773 774 Cmd->AddrExpr = readExpr(); 775 if (peek() == "(" && !readSectionDirective(Cmd, "(", peek2())) 776 setError("unknown section directive: " + peek2()); 777 } 778 779 static Expr checkAlignment(Expr E, std::string &Loc) { 780 return [=] { 781 uint64_t Alignment = std::max((uint64_t)1, E().getValue()); 782 if (!isPowerOf2_64(Alignment)) { 783 error(Loc + ": alignment must be power of 2"); 784 return (uint64_t)1; // Return a dummy value. 785 } 786 return Alignment; 787 }; 788 } 789 790 OutputSection *ScriptParser::readOverlaySectionDescription() { 791 OutputSection *Cmd = 792 Script->createOutputSection(next(), getCurrentLocation()); 793 Cmd->InOverlay = true; 794 expect("{"); 795 while (!errorCount() && !consume("}")) 796 Cmd->SectionCommands.push_back(readInputSectionRules(next())); 797 Cmd->Phdrs = readOutputSectionPhdrs(); 798 return Cmd; 799 } 800 801 OutputSection *ScriptParser::readOutputSectionDescription(StringRef OutSec) { 802 OutputSection *Cmd = 803 Script->createOutputSection(OutSec, getCurrentLocation()); 804 805 size_t SymbolsReferenced = Script->ReferencedSymbols.size(); 806 807 if (peek() != ":") 808 readSectionAddressType(Cmd); 809 expect(":"); 810 811 std::string Location = getCurrentLocation(); 812 if (consume("AT")) 813 Cmd->LMAExpr = readParenExpr(); 814 if (consume("ALIGN")) 815 Cmd->AlignExpr = checkAlignment(readParenExpr(), Location); 816 if (consume("SUBALIGN")) 817 Cmd->SubalignExpr = checkAlignment(readParenExpr(), Location); 818 819 // Parse constraints. 820 if (consume("ONLY_IF_RO")) 821 Cmd->Constraint = ConstraintKind::ReadOnly; 822 if (consume("ONLY_IF_RW")) 823 Cmd->Constraint = ConstraintKind::ReadWrite; 824 expect("{"); 825 826 while (!errorCount() && !consume("}")) { 827 StringRef Tok = next(); 828 if (Tok == ";") { 829 // Empty commands are allowed. Do nothing here. 830 } else if (SymbolAssignment *Assign = readAssignment(Tok)) { 831 Cmd->SectionCommands.push_back(Assign); 832 } else if (ByteCommand *Data = readByteCommand(Tok)) { 833 Cmd->SectionCommands.push_back(Data); 834 } else if (Tok == "CONSTRUCTORS") { 835 // CONSTRUCTORS is a keyword to make the linker recognize C++ ctors/dtors 836 // by name. This is for very old file formats such as ECOFF/XCOFF. 837 // For ELF, we should ignore. 838 } else if (Tok == "FILL") { 839 Cmd->Filler = readFill(); 840 } else if (Tok == "SORT") { 841 readSort(); 842 } else if (Tok == "INCLUDE") { 843 readInclude(); 844 } else if (peek() == "(") { 845 Cmd->SectionCommands.push_back(readInputSectionDescription(Tok)); 846 } else { 847 // We have a file name and no input sections description. It is not a 848 // commonly used syntax, but still acceptable. In that case, all sections 849 // from the file will be included. 850 auto *ISD = make<InputSectionDescription>(Tok); 851 ISD->SectionPatterns.push_back({{}, StringMatcher({"*"})}); 852 Cmd->SectionCommands.push_back(ISD); 853 } 854 } 855 856 if (consume(">")) 857 Cmd->MemoryRegionName = next(); 858 859 if (consume("AT")) { 860 expect(">"); 861 Cmd->LMARegionName = next(); 862 } 863 864 if (Cmd->LMAExpr && !Cmd->LMARegionName.empty()) 865 error("section can't have both LMA and a load region"); 866 867 Cmd->Phdrs = readOutputSectionPhdrs(); 868 869 if (consume("=")) 870 Cmd->Filler = parseFill(next()); 871 else if (peek().startswith("=")) 872 Cmd->Filler = parseFill(next().drop_front()); 873 874 // Consume optional comma following output section command. 875 consume(","); 876 877 if (Script->ReferencedSymbols.size() > SymbolsReferenced) 878 Cmd->ExpressionsUseSymbols = true; 879 return Cmd; 880 } 881 882 // Parses a given string as a octal/decimal/hexadecimal number and 883 // returns it as a big-endian number. Used for `=<fillexp>`. 884 // https://sourceware.org/binutils/docs/ld/Output-Section-Fill.html 885 // 886 // When reading a hexstring, ld.bfd handles it as a blob of arbitrary 887 // size, while ld.gold always handles it as a 32-bit big-endian number. 888 // We are compatible with ld.gold because it's easier to implement. 889 std::array<uint8_t, 4> ScriptParser::parseFill(StringRef Tok) { 890 uint32_t V = 0; 891 if (!to_integer(Tok, V)) 892 setError("invalid filler expression: " + Tok); 893 894 std::array<uint8_t, 4> Buf; 895 write32be(Buf.data(), V); 896 return Buf; 897 } 898 899 SymbolAssignment *ScriptParser::readProvideHidden(bool Provide, bool Hidden) { 900 expect("("); 901 SymbolAssignment *Cmd = readSymbolAssignment(next()); 902 Cmd->Provide = Provide; 903 Cmd->Hidden = Hidden; 904 expect(")"); 905 return Cmd; 906 } 907 908 SymbolAssignment *ScriptParser::readAssignment(StringRef Tok) { 909 // Assert expression returns Dot, so this is equal to ".=." 910 if (Tok == "ASSERT") 911 return make<SymbolAssignment>(".", readAssert(), getCurrentLocation()); 912 913 size_t OldPos = Pos; 914 SymbolAssignment *Cmd = nullptr; 915 if (peek() == "=" || peek() == "+=") 916 Cmd = readSymbolAssignment(Tok); 917 else if (Tok == "PROVIDE") 918 Cmd = readProvideHidden(true, false); 919 else if (Tok == "HIDDEN") 920 Cmd = readProvideHidden(false, true); 921 else if (Tok == "PROVIDE_HIDDEN") 922 Cmd = readProvideHidden(true, true); 923 924 if (Cmd) { 925 Cmd->CommandString = 926 Tok.str() + " " + 927 llvm::join(Tokens.begin() + OldPos, Tokens.begin() + Pos, " "); 928 expect(";"); 929 } 930 return Cmd; 931 } 932 933 SymbolAssignment *ScriptParser::readSymbolAssignment(StringRef Name) { 934 StringRef Op = next(); 935 assert(Op == "=" || Op == "+="); 936 Expr E = readExpr(); 937 if (Op == "+=") { 938 std::string Loc = getCurrentLocation(); 939 E = [=] { return add(Script->getSymbolValue(Name, Loc), E()); }; 940 } 941 return make<SymbolAssignment>(Name, E, getCurrentLocation()); 942 } 943 944 // This is an operator-precedence parser to parse a linker 945 // script expression. 946 Expr ScriptParser::readExpr() { 947 // Our lexer is context-aware. Set the in-expression bit so that 948 // they apply different tokenization rules. 949 bool Orig = InExpr; 950 InExpr = true; 951 Expr E = readExpr1(readPrimary(), 0); 952 InExpr = Orig; 953 return E; 954 } 955 956 Expr ScriptParser::combine(StringRef Op, Expr L, Expr R) { 957 if (Op == "+") 958 return [=] { return add(L(), R()); }; 959 if (Op == "-") 960 return [=] { return sub(L(), R()); }; 961 if (Op == "*") 962 return [=] { return L().getValue() * R().getValue(); }; 963 if (Op == "/") { 964 std::string Loc = getCurrentLocation(); 965 return [=]() -> uint64_t { 966 if (uint64_t RV = R().getValue()) 967 return L().getValue() / RV; 968 error(Loc + ": division by zero"); 969 return 0; 970 }; 971 } 972 if (Op == "%") { 973 std::string Loc = getCurrentLocation(); 974 return [=]() -> uint64_t { 975 if (uint64_t RV = R().getValue()) 976 return L().getValue() % RV; 977 error(Loc + ": modulo by zero"); 978 return 0; 979 }; 980 } 981 if (Op == "<<") 982 return [=] { return L().getValue() << R().getValue(); }; 983 if (Op == ">>") 984 return [=] { return L().getValue() >> R().getValue(); }; 985 if (Op == "<") 986 return [=] { return L().getValue() < R().getValue(); }; 987 if (Op == ">") 988 return [=] { return L().getValue() > R().getValue(); }; 989 if (Op == ">=") 990 return [=] { return L().getValue() >= R().getValue(); }; 991 if (Op == "<=") 992 return [=] { return L().getValue() <= R().getValue(); }; 993 if (Op == "==") 994 return [=] { return L().getValue() == R().getValue(); }; 995 if (Op == "!=") 996 return [=] { return L().getValue() != R().getValue(); }; 997 if (Op == "||") 998 return [=] { return L().getValue() || R().getValue(); }; 999 if (Op == "&&") 1000 return [=] { return L().getValue() && R().getValue(); }; 1001 if (Op == "&") 1002 return [=] { return bitAnd(L(), R()); }; 1003 if (Op == "|") 1004 return [=] { return bitOr(L(), R()); }; 1005 llvm_unreachable("invalid operator"); 1006 } 1007 1008 // This is a part of the operator-precedence parser. This function 1009 // assumes that the remaining token stream starts with an operator. 1010 Expr ScriptParser::readExpr1(Expr Lhs, int MinPrec) { 1011 while (!atEOF() && !errorCount()) { 1012 // Read an operator and an expression. 1013 if (consume("?")) 1014 return readTernary(Lhs); 1015 StringRef Op1 = peek(); 1016 if (precedence(Op1) < MinPrec) 1017 break; 1018 skip(); 1019 Expr Rhs = readPrimary(); 1020 1021 // Evaluate the remaining part of the expression first if the 1022 // next operator has greater precedence than the previous one. 1023 // For example, if we have read "+" and "3", and if the next 1024 // operator is "*", then we'll evaluate 3 * ... part first. 1025 while (!atEOF()) { 1026 StringRef Op2 = peek(); 1027 if (precedence(Op2) <= precedence(Op1)) 1028 break; 1029 Rhs = readExpr1(Rhs, precedence(Op2)); 1030 } 1031 1032 Lhs = combine(Op1, Lhs, Rhs); 1033 } 1034 return Lhs; 1035 } 1036 1037 Expr ScriptParser::getPageSize() { 1038 std::string Location = getCurrentLocation(); 1039 return [=]() -> uint64_t { 1040 if (Target) 1041 return Target->PageSize; 1042 error(Location + ": unable to calculate page size"); 1043 return 4096; // Return a dummy value. 1044 }; 1045 } 1046 1047 Expr ScriptParser::readConstant() { 1048 StringRef S = readParenLiteral(); 1049 if (S == "COMMONPAGESIZE") 1050 return getPageSize(); 1051 if (S == "MAXPAGESIZE") 1052 return [] { return Config->MaxPageSize; }; 1053 setError("unknown constant: " + S); 1054 return [] { return 0; }; 1055 } 1056 1057 // Parses Tok as an integer. It recognizes hexadecimal (prefixed with 1058 // "0x" or suffixed with "H") and decimal numbers. Decimal numbers may 1059 // have "K" (Ki) or "M" (Mi) suffixes. 1060 static Optional<uint64_t> parseInt(StringRef Tok) { 1061 // Hexadecimal 1062 uint64_t Val; 1063 if (Tok.startswith_lower("0x")) { 1064 if (!to_integer(Tok.substr(2), Val, 16)) 1065 return None; 1066 return Val; 1067 } 1068 if (Tok.endswith_lower("H")) { 1069 if (!to_integer(Tok.drop_back(), Val, 16)) 1070 return None; 1071 return Val; 1072 } 1073 1074 // Decimal 1075 if (Tok.endswith_lower("K")) { 1076 if (!to_integer(Tok.drop_back(), Val, 10)) 1077 return None; 1078 return Val * 1024; 1079 } 1080 if (Tok.endswith_lower("M")) { 1081 if (!to_integer(Tok.drop_back(), Val, 10)) 1082 return None; 1083 return Val * 1024 * 1024; 1084 } 1085 if (!to_integer(Tok, Val, 10)) 1086 return None; 1087 return Val; 1088 } 1089 1090 ByteCommand *ScriptParser::readByteCommand(StringRef Tok) { 1091 int Size = StringSwitch<int>(Tok) 1092 .Case("BYTE", 1) 1093 .Case("SHORT", 2) 1094 .Case("LONG", 4) 1095 .Case("QUAD", 8) 1096 .Default(-1); 1097 if (Size == -1) 1098 return nullptr; 1099 1100 size_t OldPos = Pos; 1101 Expr E = readParenExpr(); 1102 std::string CommandString = 1103 Tok.str() + " " + 1104 llvm::join(Tokens.begin() + OldPos, Tokens.begin() + Pos, " "); 1105 return make<ByteCommand>(E, Size, CommandString); 1106 } 1107 1108 StringRef ScriptParser::readParenLiteral() { 1109 expect("("); 1110 bool Orig = InExpr; 1111 InExpr = false; 1112 StringRef Tok = next(); 1113 InExpr = Orig; 1114 expect(")"); 1115 return Tok; 1116 } 1117 1118 static void checkIfExists(OutputSection *Cmd, StringRef Location) { 1119 if (Cmd->Location.empty() && Script->ErrorOnMissingSection) 1120 error(Location + ": undefined section " + Cmd->Name); 1121 } 1122 1123 Expr ScriptParser::readPrimary() { 1124 if (peek() == "(") 1125 return readParenExpr(); 1126 1127 if (consume("~")) { 1128 Expr E = readPrimary(); 1129 return [=] { return ~E().getValue(); }; 1130 } 1131 if (consume("!")) { 1132 Expr E = readPrimary(); 1133 return [=] { return !E().getValue(); }; 1134 } 1135 if (consume("-")) { 1136 Expr E = readPrimary(); 1137 return [=] { return -E().getValue(); }; 1138 } 1139 1140 StringRef Tok = next(); 1141 std::string Location = getCurrentLocation(); 1142 1143 // Built-in functions are parsed here. 1144 // https://sourceware.org/binutils/docs/ld/Builtin-Functions.html. 1145 if (Tok == "ABSOLUTE") { 1146 Expr Inner = readParenExpr(); 1147 return [=] { 1148 ExprValue I = Inner(); 1149 I.ForceAbsolute = true; 1150 return I; 1151 }; 1152 } 1153 if (Tok == "ADDR") { 1154 StringRef Name = readParenLiteral(); 1155 OutputSection *Sec = Script->getOrCreateOutputSection(Name); 1156 return [=]() -> ExprValue { 1157 checkIfExists(Sec, Location); 1158 return {Sec, false, 0, Location}; 1159 }; 1160 } 1161 if (Tok == "ALIGN") { 1162 expect("("); 1163 Expr E = readExpr(); 1164 if (consume(")")) { 1165 E = checkAlignment(E, Location); 1166 return [=] { return alignTo(Script->getDot(), E().getValue()); }; 1167 } 1168 expect(","); 1169 Expr E2 = checkAlignment(readExpr(), Location); 1170 expect(")"); 1171 return [=] { 1172 ExprValue V = E(); 1173 V.Alignment = E2().getValue(); 1174 return V; 1175 }; 1176 } 1177 if (Tok == "ALIGNOF") { 1178 StringRef Name = readParenLiteral(); 1179 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1180 return [=] { 1181 checkIfExists(Cmd, Location); 1182 return Cmd->Alignment; 1183 }; 1184 } 1185 if (Tok == "ASSERT") 1186 return readAssert(); 1187 if (Tok == "CONSTANT") 1188 return readConstant(); 1189 if (Tok == "DATA_SEGMENT_ALIGN") { 1190 expect("("); 1191 Expr E = readExpr(); 1192 expect(","); 1193 readExpr(); 1194 expect(")"); 1195 return [=] { 1196 return alignTo(Script->getDot(), std::max((uint64_t)1, E().getValue())); 1197 }; 1198 } 1199 if (Tok == "DATA_SEGMENT_END") { 1200 expect("("); 1201 expect("."); 1202 expect(")"); 1203 return [] { return Script->getDot(); }; 1204 } 1205 if (Tok == "DATA_SEGMENT_RELRO_END") { 1206 // GNU linkers implements more complicated logic to handle 1207 // DATA_SEGMENT_RELRO_END. We instead ignore the arguments and 1208 // just align to the next page boundary for simplicity. 1209 expect("("); 1210 readExpr(); 1211 expect(","); 1212 readExpr(); 1213 expect(")"); 1214 Expr E = getPageSize(); 1215 return [=] { return alignTo(Script->getDot(), E().getValue()); }; 1216 } 1217 if (Tok == "DEFINED") { 1218 StringRef Name = readParenLiteral(); 1219 return [=] { return Symtab->find(Name) ? 1 : 0; }; 1220 } 1221 if (Tok == "LENGTH") { 1222 StringRef Name = readParenLiteral(); 1223 if (Script->MemoryRegions.count(Name) == 0) { 1224 setError("memory region not defined: " + Name); 1225 return [] { return 0; }; 1226 } 1227 return [=] { return Script->MemoryRegions[Name]->Length; }; 1228 } 1229 if (Tok == "LOADADDR") { 1230 StringRef Name = readParenLiteral(); 1231 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1232 return [=] { 1233 checkIfExists(Cmd, Location); 1234 return Cmd->getLMA(); 1235 }; 1236 } 1237 if (Tok == "MAX" || Tok == "MIN") { 1238 expect("("); 1239 Expr A = readExpr(); 1240 expect(","); 1241 Expr B = readExpr(); 1242 expect(")"); 1243 if (Tok == "MIN") 1244 return [=] { return std::min(A().getValue(), B().getValue()); }; 1245 return [=] { return std::max(A().getValue(), B().getValue()); }; 1246 } 1247 if (Tok == "ORIGIN") { 1248 StringRef Name = readParenLiteral(); 1249 if (Script->MemoryRegions.count(Name) == 0) { 1250 setError("memory region not defined: " + Name); 1251 return [] { return 0; }; 1252 } 1253 return [=] { return Script->MemoryRegions[Name]->Origin; }; 1254 } 1255 if (Tok == "SEGMENT_START") { 1256 expect("("); 1257 skip(); 1258 expect(","); 1259 Expr E = readExpr(); 1260 expect(")"); 1261 return [=] { return E(); }; 1262 } 1263 if (Tok == "SIZEOF") { 1264 StringRef Name = readParenLiteral(); 1265 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1266 // Linker script does not create an output section if its content is empty. 1267 // We want to allow SIZEOF(.foo) where .foo is a section which happened to 1268 // be empty. 1269 return [=] { return Cmd->Size; }; 1270 } 1271 if (Tok == "SIZEOF_HEADERS") 1272 return [=] { return elf::getHeaderSize(); }; 1273 1274 // Tok is the dot. 1275 if (Tok == ".") 1276 return [=] { return Script->getSymbolValue(Tok, Location); }; 1277 1278 // Tok is a literal number. 1279 if (Optional<uint64_t> Val = parseInt(Tok)) 1280 return [=] { return *Val; }; 1281 1282 // Tok is a symbol name. 1283 if (!isValidCIdentifier(Tok)) 1284 setError("malformed number: " + Tok); 1285 Script->ReferencedSymbols.push_back(Tok); 1286 return [=] { return Script->getSymbolValue(Tok, Location); }; 1287 } 1288 1289 Expr ScriptParser::readTernary(Expr Cond) { 1290 Expr L = readExpr(); 1291 expect(":"); 1292 Expr R = readExpr(); 1293 return [=] { return Cond().getValue() ? L() : R(); }; 1294 } 1295 1296 Expr ScriptParser::readParenExpr() { 1297 expect("("); 1298 Expr E = readExpr(); 1299 expect(")"); 1300 return E; 1301 } 1302 1303 std::vector<StringRef> ScriptParser::readOutputSectionPhdrs() { 1304 std::vector<StringRef> Phdrs; 1305 while (!errorCount() && peek().startswith(":")) { 1306 StringRef Tok = next(); 1307 Phdrs.push_back((Tok.size() == 1) ? next() : Tok.substr(1)); 1308 } 1309 return Phdrs; 1310 } 1311 1312 // Read a program header type name. The next token must be a 1313 // name of a program header type or a constant (e.g. "0x3"). 1314 unsigned ScriptParser::readPhdrType() { 1315 StringRef Tok = next(); 1316 if (Optional<uint64_t> Val = parseInt(Tok)) 1317 return *Val; 1318 1319 unsigned Ret = StringSwitch<unsigned>(Tok) 1320 .Case("PT_NULL", PT_NULL) 1321 .Case("PT_LOAD", PT_LOAD) 1322 .Case("PT_DYNAMIC", PT_DYNAMIC) 1323 .Case("PT_INTERP", PT_INTERP) 1324 .Case("PT_NOTE", PT_NOTE) 1325 .Case("PT_SHLIB", PT_SHLIB) 1326 .Case("PT_PHDR", PT_PHDR) 1327 .Case("PT_TLS", PT_TLS) 1328 .Case("PT_GNU_EH_FRAME", PT_GNU_EH_FRAME) 1329 .Case("PT_GNU_STACK", PT_GNU_STACK) 1330 .Case("PT_GNU_RELRO", PT_GNU_RELRO) 1331 .Case("PT_OPENBSD_RANDOMIZE", PT_OPENBSD_RANDOMIZE) 1332 .Case("PT_OPENBSD_WXNEEDED", PT_OPENBSD_WXNEEDED) 1333 .Case("PT_OPENBSD_BOOTDATA", PT_OPENBSD_BOOTDATA) 1334 .Default(-1); 1335 1336 if (Ret == (unsigned)-1) { 1337 setError("invalid program header type: " + Tok); 1338 return PT_NULL; 1339 } 1340 return Ret; 1341 } 1342 1343 // Reads an anonymous version declaration. 1344 void ScriptParser::readAnonymousDeclaration() { 1345 std::vector<SymbolVersion> Locals; 1346 std::vector<SymbolVersion> Globals; 1347 std::tie(Locals, Globals) = readSymbols(); 1348 1349 for (SymbolVersion V : Locals) { 1350 if (V.Name == "*") 1351 Config->DefaultSymbolVersion = VER_NDX_LOCAL; 1352 else 1353 Config->VersionScriptLocals.push_back(V); 1354 } 1355 1356 for (SymbolVersion V : Globals) 1357 Config->VersionScriptGlobals.push_back(V); 1358 1359 expect(";"); 1360 } 1361 1362 // Reads a non-anonymous version definition, 1363 // e.g. "VerStr { global: foo; bar; local: *; };". 1364 void ScriptParser::readVersionDeclaration(StringRef VerStr) { 1365 // Read a symbol list. 1366 std::vector<SymbolVersion> Locals; 1367 std::vector<SymbolVersion> Globals; 1368 std::tie(Locals, Globals) = readSymbols(); 1369 1370 for (SymbolVersion V : Locals) { 1371 if (V.Name == "*") 1372 Config->DefaultSymbolVersion = VER_NDX_LOCAL; 1373 else 1374 Config->VersionScriptLocals.push_back(V); 1375 } 1376 1377 // Create a new version definition and add that to the global symbols. 1378 VersionDefinition Ver; 1379 Ver.Name = VerStr; 1380 Ver.Globals = Globals; 1381 1382 // User-defined version number starts from 2 because 0 and 1 are 1383 // reserved for VER_NDX_LOCAL and VER_NDX_GLOBAL, respectively. 1384 Ver.Id = Config->VersionDefinitions.size() + 2; 1385 Config->VersionDefinitions.push_back(Ver); 1386 1387 // Each version may have a parent version. For example, "Ver2" 1388 // defined as "Ver2 { global: foo; local: *; } Ver1;" has "Ver1" 1389 // as a parent. This version hierarchy is, probably against your 1390 // instinct, purely for hint; the runtime doesn't care about it 1391 // at all. In LLD, we simply ignore it. 1392 if (peek() != ";") 1393 skip(); 1394 expect(";"); 1395 } 1396 1397 static bool hasWildcard(StringRef S) { 1398 return S.find_first_of("?*[") != StringRef::npos; 1399 } 1400 1401 // Reads a list of symbols, e.g. "{ global: foo; bar; local: *; };". 1402 std::pair<std::vector<SymbolVersion>, std::vector<SymbolVersion>> 1403 ScriptParser::readSymbols() { 1404 std::vector<SymbolVersion> Locals; 1405 std::vector<SymbolVersion> Globals; 1406 std::vector<SymbolVersion> *V = &Globals; 1407 1408 while (!errorCount()) { 1409 if (consume("}")) 1410 break; 1411 if (consumeLabel("local")) { 1412 V = &Locals; 1413 continue; 1414 } 1415 if (consumeLabel("global")) { 1416 V = &Globals; 1417 continue; 1418 } 1419 1420 if (consume("extern")) { 1421 std::vector<SymbolVersion> Ext = readVersionExtern(); 1422 V->insert(V->end(), Ext.begin(), Ext.end()); 1423 } else { 1424 StringRef Tok = next(); 1425 V->push_back({unquote(Tok), false, hasWildcard(Tok)}); 1426 } 1427 expect(";"); 1428 } 1429 return {Locals, Globals}; 1430 } 1431 1432 // Reads an "extern C++" directive, e.g., 1433 // "extern "C++" { ns::*; "f(int, double)"; };" 1434 // 1435 // The last semicolon is optional. E.g. this is OK: 1436 // "extern "C++" { ns::*; "f(int, double)" };" 1437 std::vector<SymbolVersion> ScriptParser::readVersionExtern() { 1438 StringRef Tok = next(); 1439 bool IsCXX = Tok == "\"C++\""; 1440 if (!IsCXX && Tok != "\"C\"") 1441 setError("Unknown language"); 1442 expect("{"); 1443 1444 std::vector<SymbolVersion> Ret; 1445 while (!errorCount() && peek() != "}") { 1446 StringRef Tok = next(); 1447 bool HasWildcard = !Tok.startswith("\"") && hasWildcard(Tok); 1448 Ret.push_back({unquote(Tok), IsCXX, HasWildcard}); 1449 if (consume("}")) 1450 return Ret; 1451 expect(";"); 1452 } 1453 1454 expect("}"); 1455 return Ret; 1456 } 1457 1458 uint64_t ScriptParser::readMemoryAssignment(StringRef S1, StringRef S2, 1459 StringRef S3) { 1460 if (!consume(S1) && !consume(S2) && !consume(S3)) { 1461 setError("expected one of: " + S1 + ", " + S2 + ", or " + S3); 1462 return 0; 1463 } 1464 expect("="); 1465 return readExpr()().getValue(); 1466 } 1467 1468 // Parse the MEMORY command as specified in: 1469 // https://sourceware.org/binutils/docs/ld/MEMORY.html 1470 // 1471 // MEMORY { name [(attr)] : ORIGIN = origin, LENGTH = len ... } 1472 void ScriptParser::readMemory() { 1473 expect("{"); 1474 while (!errorCount() && !consume("}")) { 1475 StringRef Tok = next(); 1476 if (Tok == "INCLUDE") { 1477 readInclude(); 1478 continue; 1479 } 1480 1481 uint32_t Flags = 0; 1482 uint32_t NegFlags = 0; 1483 if (consume("(")) { 1484 std::tie(Flags, NegFlags) = readMemoryAttributes(); 1485 expect(")"); 1486 } 1487 expect(":"); 1488 1489 uint64_t Origin = readMemoryAssignment("ORIGIN", "org", "o"); 1490 expect(","); 1491 uint64_t Length = readMemoryAssignment("LENGTH", "len", "l"); 1492 1493 // Add the memory region to the region map. 1494 MemoryRegion *MR = make<MemoryRegion>(Tok, Origin, Length, Flags, NegFlags); 1495 if (!Script->MemoryRegions.insert({Tok, MR}).second) 1496 setError("region '" + Tok + "' already defined"); 1497 } 1498 } 1499 1500 // This function parses the attributes used to match against section 1501 // flags when placing output sections in a memory region. These flags 1502 // are only used when an explicit memory region name is not used. 1503 std::pair<uint32_t, uint32_t> ScriptParser::readMemoryAttributes() { 1504 uint32_t Flags = 0; 1505 uint32_t NegFlags = 0; 1506 bool Invert = false; 1507 1508 for (char C : next().lower()) { 1509 uint32_t Flag = 0; 1510 if (C == '!') 1511 Invert = !Invert; 1512 else if (C == 'w') 1513 Flag = SHF_WRITE; 1514 else if (C == 'x') 1515 Flag = SHF_EXECINSTR; 1516 else if (C == 'a') 1517 Flag = SHF_ALLOC; 1518 else if (C != 'r') 1519 setError("invalid memory region attribute"); 1520 1521 if (Invert) 1522 NegFlags |= Flag; 1523 else 1524 Flags |= Flag; 1525 } 1526 return {Flags, NegFlags}; 1527 } 1528 1529 void elf::readLinkerScript(MemoryBufferRef MB) { 1530 ScriptParser(MB).readLinkerScript(); 1531 } 1532 1533 void elf::readVersionScript(MemoryBufferRef MB) { 1534 ScriptParser(MB).readVersionScript(); 1535 } 1536 1537 void elf::readDynamicList(MemoryBufferRef MB) { 1538 ScriptParser(MB).readDynamicList(); 1539 } 1540 1541 void elf::readDefsym(StringRef Name, MemoryBufferRef MB) { 1542 ScriptParser(MB).readDefsym(Name); 1543 } 1544