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 std::tuple<ELFKind, uint16_t, bool> readBfdName(); 97 void readSort(); 98 Expr readAssert(); 99 Expr readConstant(); 100 Expr getPageSize(); 101 102 uint64_t readMemoryAssignment(StringRef, StringRef, StringRef); 103 std::pair<uint32_t, uint32_t> readMemoryAttributes(); 104 105 Expr combine(StringRef Op, Expr L, Expr R); 106 Expr readExpr(); 107 Expr readExpr1(Expr Lhs, int MinPrec); 108 StringRef readParenLiteral(); 109 Expr readPrimary(); 110 Expr readTernary(Expr Cond); 111 Expr readParenExpr(); 112 113 // For parsing version script. 114 std::vector<SymbolVersion> readVersionExtern(); 115 void readAnonymousDeclaration(); 116 void readVersionDeclaration(StringRef VerStr); 117 118 std::pair<std::vector<SymbolVersion>, std::vector<SymbolVersion>> 119 readSymbols(); 120 121 // True if a script being read is in a subdirectory specified by -sysroot. 122 bool IsUnderSysroot; 123 124 // A set to detect an INCLUDE() cycle. 125 StringSet<> Seen; 126 }; 127 } // namespace 128 129 static StringRef unquote(StringRef S) { 130 if (S.startswith("\"")) 131 return S.substr(1, S.size() - 2); 132 return S; 133 } 134 135 static bool isUnderSysroot(StringRef Path) { 136 if (Config->Sysroot == "") 137 return false; 138 for (; !Path.empty(); Path = sys::path::parent_path(Path)) 139 if (sys::fs::equivalent(Config->Sysroot, Path)) 140 return true; 141 return false; 142 } 143 144 // Some operations only support one non absolute value. Move the 145 // absolute one to the right hand side for convenience. 146 static void moveAbsRight(ExprValue &A, ExprValue &B) { 147 if (A.Sec == nullptr || (A.ForceAbsolute && !B.isAbsolute())) 148 std::swap(A, B); 149 if (!B.isAbsolute()) 150 error(A.Loc + ": at least one side of the expression must be absolute"); 151 } 152 153 static ExprValue add(ExprValue A, ExprValue B) { 154 moveAbsRight(A, B); 155 return {A.Sec, A.ForceAbsolute, A.getSectionOffset() + B.getValue(), A.Loc}; 156 } 157 158 static ExprValue sub(ExprValue A, ExprValue B) { 159 // The distance between two symbols in sections is absolute. 160 if (!A.isAbsolute() && !B.isAbsolute()) 161 return A.getValue() - B.getValue(); 162 return {A.Sec, false, A.getSectionOffset() - B.getValue(), A.Loc}; 163 } 164 165 static ExprValue bitAnd(ExprValue A, ExprValue B) { 166 moveAbsRight(A, B); 167 return {A.Sec, A.ForceAbsolute, 168 (A.getValue() & B.getValue()) - A.getSecAddr(), A.Loc}; 169 } 170 171 static ExprValue bitOr(ExprValue A, ExprValue B) { 172 moveAbsRight(A, B); 173 return {A.Sec, A.ForceAbsolute, 174 (A.getValue() | B.getValue()) - A.getSecAddr(), A.Loc}; 175 } 176 177 void ScriptParser::readDynamicList() { 178 Config->HasDynamicList = true; 179 expect("{"); 180 std::vector<SymbolVersion> Locals; 181 std::vector<SymbolVersion> Globals; 182 std::tie(Locals, Globals) = readSymbols(); 183 expect(";"); 184 185 if (!atEOF()) { 186 setError("EOF expected, but got " + next()); 187 return; 188 } 189 if (!Locals.empty()) { 190 setError("\"local:\" scope not supported in --dynamic-list"); 191 return; 192 } 193 194 for (SymbolVersion V : Globals) 195 Config->DynamicList.push_back(V); 196 } 197 198 void ScriptParser::readVersionScript() { 199 readVersionScriptCommand(); 200 if (!atEOF()) 201 setError("EOF expected, but got " + next()); 202 } 203 204 void ScriptParser::readVersionScriptCommand() { 205 if (consume("{")) { 206 readAnonymousDeclaration(); 207 return; 208 } 209 210 while (!atEOF() && !errorCount() && peek() != "}") { 211 StringRef VerStr = next(); 212 if (VerStr == "{") { 213 setError("anonymous version definition is used in " 214 "combination with other version definitions"); 215 return; 216 } 217 expect("{"); 218 readVersionDeclaration(VerStr); 219 } 220 } 221 222 void ScriptParser::readVersion() { 223 expect("{"); 224 readVersionScriptCommand(); 225 expect("}"); 226 } 227 228 void ScriptParser::readLinkerScript() { 229 while (!atEOF()) { 230 StringRef Tok = next(); 231 if (Tok == ";") 232 continue; 233 234 if (Tok == "ENTRY") { 235 readEntry(); 236 } else if (Tok == "EXTERN") { 237 readExtern(); 238 } else if (Tok == "GROUP") { 239 readGroup(); 240 } else if (Tok == "INCLUDE") { 241 readInclude(); 242 } else if (Tok == "INPUT") { 243 readInput(); 244 } else if (Tok == "MEMORY") { 245 readMemory(); 246 } else if (Tok == "OUTPUT") { 247 readOutput(); 248 } else if (Tok == "OUTPUT_ARCH") { 249 readOutputArch(); 250 } else if (Tok == "OUTPUT_FORMAT") { 251 readOutputFormat(); 252 } else if (Tok == "PHDRS") { 253 readPhdrs(); 254 } else if (Tok == "REGION_ALIAS") { 255 readRegionAlias(); 256 } else if (Tok == "SEARCH_DIR") { 257 readSearchDir(); 258 } else if (Tok == "SECTIONS") { 259 readSections(); 260 } else if (Tok == "TARGET") { 261 readTarget(); 262 } else if (Tok == "VERSION") { 263 readVersion(); 264 } else if (SymbolAssignment *Cmd = readAssignment(Tok)) { 265 Script->SectionCommands.push_back(Cmd); 266 } else { 267 setError("unknown directive: " + Tok); 268 } 269 } 270 } 271 272 void ScriptParser::readDefsym(StringRef Name) { 273 if (errorCount()) 274 return; 275 Expr E = readExpr(); 276 if (!atEOF()) 277 setError("EOF expected, but got " + next()); 278 SymbolAssignment *Cmd = make<SymbolAssignment>(Name, E, getCurrentLocation()); 279 Script->SectionCommands.push_back(Cmd); 280 } 281 282 void ScriptParser::addFile(StringRef S) { 283 if (IsUnderSysroot && S.startswith("/")) { 284 SmallString<128> PathData; 285 StringRef Path = (Config->Sysroot + S).toStringRef(PathData); 286 if (sys::fs::exists(Path)) { 287 Driver->addFile(Saver.save(Path), /*WithLOption=*/false); 288 return; 289 } 290 } 291 292 if (S.startswith("/")) { 293 Driver->addFile(S, /*WithLOption=*/false); 294 } else if (S.startswith("=")) { 295 if (Config->Sysroot.empty()) 296 Driver->addFile(S.substr(1), /*WithLOption=*/false); 297 else 298 Driver->addFile(Saver.save(Config->Sysroot + "/" + S.substr(1)), 299 /*WithLOption=*/false); 300 } else if (S.startswith("-l")) { 301 Driver->addLibrary(S.substr(2)); 302 } else if (sys::fs::exists(S)) { 303 Driver->addFile(S, /*WithLOption=*/false); 304 } else { 305 if (Optional<std::string> Path = findFromSearchPaths(S)) 306 Driver->addFile(Saver.save(*Path), /*WithLOption=*/true); 307 else 308 setError("unable to find " + S); 309 } 310 } 311 312 void ScriptParser::readAsNeeded() { 313 expect("("); 314 bool Orig = Config->AsNeeded; 315 Config->AsNeeded = true; 316 while (!errorCount() && !consume(")")) 317 addFile(unquote(next())); 318 Config->AsNeeded = Orig; 319 } 320 321 void ScriptParser::readEntry() { 322 // -e <symbol> takes predecence over ENTRY(<symbol>). 323 expect("("); 324 StringRef Tok = next(); 325 if (Config->Entry.empty()) 326 Config->Entry = Tok; 327 expect(")"); 328 } 329 330 void ScriptParser::readExtern() { 331 expect("("); 332 while (!errorCount() && !consume(")")) 333 Config->Undefined.push_back(next()); 334 } 335 336 void ScriptParser::readGroup() { 337 bool Orig = InputFile::IsInGroup; 338 InputFile::IsInGroup = true; 339 readInput(); 340 InputFile::IsInGroup = Orig; 341 if (!Orig) 342 ++InputFile::NextGroupId; 343 } 344 345 void ScriptParser::readInclude() { 346 StringRef Tok = unquote(next()); 347 348 if (!Seen.insert(Tok).second) { 349 setError("there is a cycle in linker script INCLUDEs"); 350 return; 351 } 352 353 if (Optional<std::string> Path = searchScript(Tok)) { 354 if (Optional<MemoryBufferRef> MB = readFile(*Path)) 355 tokenize(*MB); 356 return; 357 } 358 setError("cannot find linker script " + Tok); 359 } 360 361 void ScriptParser::readInput() { 362 expect("("); 363 while (!errorCount() && !consume(")")) { 364 if (consume("AS_NEEDED")) 365 readAsNeeded(); 366 else 367 addFile(unquote(next())); 368 } 369 } 370 371 void ScriptParser::readOutput() { 372 // -o <file> takes predecence over OUTPUT(<file>). 373 expect("("); 374 StringRef Tok = next(); 375 if (Config->OutputFile.empty()) 376 Config->OutputFile = unquote(Tok); 377 expect(")"); 378 } 379 380 void ScriptParser::readOutputArch() { 381 // OUTPUT_ARCH is ignored for now. 382 expect("("); 383 while (!errorCount() && !consume(")")) 384 skip(); 385 } 386 387 std::tuple<ELFKind, uint16_t, bool> ScriptParser::readBfdName() { 388 StringRef S = unquote(next()); 389 if (S == "elf32-i386") 390 return std::make_tuple(ELF32LEKind, EM_386, false); 391 if (S == "elf32-iamcu") 392 return std::make_tuple(ELF32LEKind, EM_IAMCU, false); 393 if (S == "elf32-littlearm") 394 return std::make_tuple(ELF32LEKind, EM_ARM, false); 395 if (S == "elf32-x86-64") 396 return std::make_tuple(ELF32LEKind, EM_X86_64, false); 397 if (S == "elf64-littleaarch64") 398 return std::make_tuple(ELF64LEKind, EM_AARCH64, false); 399 if (S == "elf64-powerpc") 400 return std::make_tuple(ELF64BEKind, EM_PPC64, false); 401 if (S == "elf64-powerpcle") 402 return std::make_tuple(ELF64LEKind, EM_PPC64, false); 403 if (S == "elf64-x86-64") 404 return std::make_tuple(ELF64LEKind, EM_X86_64, false); 405 if (S == "elf32-tradbigmips") 406 return std::make_tuple(ELF32BEKind, EM_MIPS, false); 407 if (S == "elf32-ntradbigmips") 408 return std::make_tuple(ELF32BEKind, EM_MIPS, true); 409 if (S == "elf32-tradlittlemips") 410 return std::make_tuple(ELF32LEKind, EM_MIPS, false); 411 if (S == "elf32-ntradlittlemips") 412 return std::make_tuple(ELF32LEKind, EM_MIPS, true); 413 if (S == "elf64-tradbigmips") 414 return std::make_tuple(ELF64BEKind, EM_MIPS, false); 415 if (S == "elf64-tradlittlemips") 416 return std::make_tuple(ELF64LEKind, EM_MIPS, false); 417 418 setError("unknown output format name: " + S); 419 return std::make_tuple(ELFNoneKind, EM_NONE, false); 420 } 421 422 // Parse OUTPUT_FORMAT(bfdname) or OUTPUT_FORMAT(bfdname, big, little). 423 // Currently we ignore big and little parameters. 424 void ScriptParser::readOutputFormat() { 425 expect("("); 426 427 std::tuple<ELFKind, uint16_t, bool> BfdTuple = readBfdName(); 428 if (Config->EKind == ELFNoneKind) 429 std::tie(Config->EKind, Config->EMachine, Config->MipsN32Abi) = BfdTuple; 430 431 if (consume(")")) 432 return; 433 expect(","); 434 skip(); 435 expect(","); 436 skip(); 437 expect(")"); 438 } 439 440 void ScriptParser::readPhdrs() { 441 expect("{"); 442 443 while (!errorCount() && !consume("}")) { 444 PhdrsCommand Cmd; 445 Cmd.Name = next(); 446 Cmd.Type = readPhdrType(); 447 448 while (!errorCount() && !consume(";")) { 449 if (consume("FILEHDR")) 450 Cmd.HasFilehdr = true; 451 else if (consume("PHDRS")) 452 Cmd.HasPhdrs = true; 453 else if (consume("AT")) 454 Cmd.LMAExpr = readParenExpr(); 455 else if (consume("FLAGS")) 456 Cmd.Flags = readParenExpr()().getValue(); 457 else 458 setError("unexpected header attribute: " + next()); 459 } 460 461 Script->PhdrsCommands.push_back(Cmd); 462 } 463 } 464 465 void ScriptParser::readRegionAlias() { 466 expect("("); 467 StringRef Alias = unquote(next()); 468 expect(","); 469 StringRef Name = next(); 470 expect(")"); 471 472 if (Script->MemoryRegions.count(Alias)) 473 setError("redefinition of memory region '" + Alias + "'"); 474 if (!Script->MemoryRegions.count(Name)) 475 setError("memory region '" + Name + "' is not defined"); 476 Script->MemoryRegions.insert({Alias, Script->MemoryRegions[Name]}); 477 } 478 479 void ScriptParser::readSearchDir() { 480 expect("("); 481 StringRef Tok = next(); 482 if (!Config->Nostdlib) 483 Config->SearchPaths.push_back(unquote(Tok)); 484 expect(")"); 485 } 486 487 // This reads an overlay description. Overlays are used to describe output 488 // sections that use the same virtual memory range and normally would trigger 489 // linker's sections sanity check failures. 490 // https://sourceware.org/binutils/docs/ld/Overlay-Description.html#Overlay-Description 491 std::vector<BaseCommand *> ScriptParser::readOverlay() { 492 // VA and LMA expressions are optional, though for simplicity of 493 // implementation we assume they are not. That is what OVERLAY was designed 494 // for first of all: to allow sections with overlapping VAs at different LMAs. 495 Expr AddrExpr = readExpr(); 496 expect(":"); 497 expect("AT"); 498 Expr LMAExpr = readParenExpr(); 499 expect("{"); 500 501 std::vector<BaseCommand *> V; 502 OutputSection *Prev = nullptr; 503 while (!errorCount() && !consume("}")) { 504 // VA is the same for all sections. The LMAs are consecutive in memory 505 // starting from the base load address specified. 506 OutputSection *OS = readOverlaySectionDescription(); 507 OS->AddrExpr = AddrExpr; 508 if (Prev) 509 OS->LMAExpr = [=] { return Prev->getLMA() + Prev->Size; }; 510 else 511 OS->LMAExpr = LMAExpr; 512 V.push_back(OS); 513 Prev = OS; 514 } 515 516 // According to the specification, at the end of the overlay, the location 517 // counter should be equal to the overlay base address plus size of the 518 // largest section seen in the overlay. 519 // Here we want to create the Dot assignment command to achieve that. 520 Expr MoveDot = [=] { 521 uint64_t Max = 0; 522 for (BaseCommand *Cmd : V) 523 Max = std::max(Max, cast<OutputSection>(Cmd)->Size); 524 return AddrExpr().getValue() + Max; 525 }; 526 V.push_back(make<SymbolAssignment>(".", MoveDot, getCurrentLocation())); 527 return V; 528 } 529 530 void ScriptParser::readSections() { 531 Script->HasSectionsCommand = true; 532 533 // -no-rosegment is used to avoid placing read only non-executable sections in 534 // their own segment. We do the same if SECTIONS command is present in linker 535 // script. See comment for computeFlags(). 536 Config->SingleRoRx = true; 537 538 expect("{"); 539 std::vector<BaseCommand *> V; 540 while (!errorCount() && !consume("}")) { 541 StringRef Tok = next(); 542 if (Tok == "OVERLAY") { 543 for (BaseCommand *Cmd : readOverlay()) 544 V.push_back(Cmd); 545 continue; 546 } else if (Tok == "INCLUDE") { 547 readInclude(); 548 continue; 549 } 550 551 if (BaseCommand *Cmd = readAssignment(Tok)) 552 V.push_back(Cmd); 553 else 554 V.push_back(readOutputSectionDescription(Tok)); 555 } 556 557 if (!atEOF() && consume("INSERT")) { 558 std::vector<BaseCommand *> *Dest = nullptr; 559 if (consume("AFTER")) 560 Dest = &Script->InsertAfterCommands[next()]; 561 else if (consume("BEFORE")) 562 Dest = &Script->InsertBeforeCommands[next()]; 563 else 564 setError("expected AFTER/BEFORE, but got '" + next() + "'"); 565 if (Dest) 566 Dest->insert(Dest->end(), V.begin(), V.end()); 567 return; 568 } 569 570 Script->SectionCommands.insert(Script->SectionCommands.end(), V.begin(), 571 V.end()); 572 } 573 574 void ScriptParser::readTarget() { 575 // TARGET(foo) is an alias for "--format foo". Unlike GNU linkers, 576 // we accept only a limited set of BFD names (i.e. "elf" or "binary") 577 // for --format. We recognize only /^elf/ and "binary" in the linker 578 // script as well. 579 expect("("); 580 StringRef Tok = next(); 581 expect(")"); 582 583 if (Tok.startswith("elf")) 584 Config->FormatBinary = false; 585 else if (Tok == "binary") 586 Config->FormatBinary = true; 587 else 588 setError("unknown target: " + Tok); 589 } 590 591 static int precedence(StringRef Op) { 592 return StringSwitch<int>(Op) 593 .Cases("*", "/", "%", 8) 594 .Cases("+", "-", 7) 595 .Cases("<<", ">>", 6) 596 .Cases("<", "<=", ">", ">=", "==", "!=", 5) 597 .Case("&", 4) 598 .Case("|", 3) 599 .Case("&&", 2) 600 .Case("||", 1) 601 .Default(-1); 602 } 603 604 StringMatcher ScriptParser::readFilePatterns() { 605 std::vector<StringRef> V; 606 while (!errorCount() && !consume(")")) 607 V.push_back(next()); 608 return StringMatcher(V); 609 } 610 611 SortSectionPolicy ScriptParser::readSortKind() { 612 if (consume("SORT") || consume("SORT_BY_NAME")) 613 return SortSectionPolicy::Name; 614 if (consume("SORT_BY_ALIGNMENT")) 615 return SortSectionPolicy::Alignment; 616 if (consume("SORT_BY_INIT_PRIORITY")) 617 return SortSectionPolicy::Priority; 618 if (consume("SORT_NONE")) 619 return SortSectionPolicy::None; 620 return SortSectionPolicy::Default; 621 } 622 623 // Reads SECTIONS command contents in the following form: 624 // 625 // <contents> ::= <elem>* 626 // <elem> ::= <exclude>? <glob-pattern> 627 // <exclude> ::= "EXCLUDE_FILE" "(" <glob-pattern>+ ")" 628 // 629 // For example, 630 // 631 // *(.foo EXCLUDE_FILE (a.o) .bar EXCLUDE_FILE (b.o) .baz) 632 // 633 // is parsed as ".foo", ".bar" with "a.o", and ".baz" with "b.o". 634 // The semantics of that is section .foo in any file, section .bar in 635 // any file but a.o, and section .baz in any file but b.o. 636 std::vector<SectionPattern> ScriptParser::readInputSectionsList() { 637 std::vector<SectionPattern> Ret; 638 while (!errorCount() && peek() != ")") { 639 StringMatcher ExcludeFilePat; 640 if (consume("EXCLUDE_FILE")) { 641 expect("("); 642 ExcludeFilePat = readFilePatterns(); 643 } 644 645 std::vector<StringRef> V; 646 while (!errorCount() && peek() != ")" && peek() != "EXCLUDE_FILE") 647 V.push_back(next()); 648 649 if (!V.empty()) 650 Ret.push_back({std::move(ExcludeFilePat), StringMatcher(V)}); 651 else 652 setError("section pattern is expected"); 653 } 654 return Ret; 655 } 656 657 // Reads contents of "SECTIONS" directive. That directive contains a 658 // list of glob patterns for input sections. The grammar is as follows. 659 // 660 // <patterns> ::= <section-list> 661 // | <sort> "(" <section-list> ")" 662 // | <sort> "(" <sort> "(" <section-list> ")" ")" 663 // 664 // <sort> ::= "SORT" | "SORT_BY_NAME" | "SORT_BY_ALIGNMENT" 665 // | "SORT_BY_INIT_PRIORITY" | "SORT_NONE" 666 // 667 // <section-list> is parsed by readInputSectionsList(). 668 InputSectionDescription * 669 ScriptParser::readInputSectionRules(StringRef FilePattern) { 670 auto *Cmd = make<InputSectionDescription>(FilePattern); 671 expect("("); 672 673 while (!errorCount() && !consume(")")) { 674 SortSectionPolicy Outer = readSortKind(); 675 SortSectionPolicy Inner = SortSectionPolicy::Default; 676 std::vector<SectionPattern> V; 677 if (Outer != SortSectionPolicy::Default) { 678 expect("("); 679 Inner = readSortKind(); 680 if (Inner != SortSectionPolicy::Default) { 681 expect("("); 682 V = readInputSectionsList(); 683 expect(")"); 684 } else { 685 V = readInputSectionsList(); 686 } 687 expect(")"); 688 } else { 689 V = readInputSectionsList(); 690 } 691 692 for (SectionPattern &Pat : V) { 693 Pat.SortInner = Inner; 694 Pat.SortOuter = Outer; 695 } 696 697 std::move(V.begin(), V.end(), std::back_inserter(Cmd->SectionPatterns)); 698 } 699 return Cmd; 700 } 701 702 InputSectionDescription * 703 ScriptParser::readInputSectionDescription(StringRef Tok) { 704 // Input section wildcard can be surrounded by KEEP. 705 // https://sourceware.org/binutils/docs/ld/Input-Section-Keep.html#Input-Section-Keep 706 if (Tok == "KEEP") { 707 expect("("); 708 StringRef FilePattern = next(); 709 InputSectionDescription *Cmd = readInputSectionRules(FilePattern); 710 expect(")"); 711 Script->KeptSections.push_back(Cmd); 712 return Cmd; 713 } 714 return readInputSectionRules(Tok); 715 } 716 717 void ScriptParser::readSort() { 718 expect("("); 719 expect("CONSTRUCTORS"); 720 expect(")"); 721 } 722 723 Expr ScriptParser::readAssert() { 724 expect("("); 725 Expr E = readExpr(); 726 expect(","); 727 StringRef Msg = unquote(next()); 728 expect(")"); 729 730 return [=] { 731 if (!E().getValue()) 732 error(Msg); 733 return Script->getDot(); 734 }; 735 } 736 737 // Reads a FILL(expr) command. We handle the FILL command as an 738 // alias for =fillexp section attribute, which is different from 739 // what GNU linkers do. 740 // https://sourceware.org/binutils/docs/ld/Output-Section-Data.html 741 std::array<uint8_t, 4> ScriptParser::readFill() { 742 expect("("); 743 std::array<uint8_t, 4> V = parseFill(next()); 744 expect(")"); 745 return V; 746 } 747 748 // Tries to read the special directive for an output section definition which 749 // can be one of following: "(NOLOAD)", "(COPY)", "(INFO)" or "(OVERLAY)". 750 // Tok1 and Tok2 are next 2 tokens peeked. See comment for readSectionAddressType below. 751 bool ScriptParser::readSectionDirective(OutputSection *Cmd, StringRef Tok1, StringRef Tok2) { 752 if (Tok1 != "(") 753 return false; 754 if (Tok2 != "NOLOAD" && Tok2 != "COPY" && Tok2 != "INFO" && Tok2 != "OVERLAY") 755 return false; 756 757 expect("("); 758 if (consume("NOLOAD")) { 759 Cmd->Noload = true; 760 } else { 761 skip(); // This is "COPY", "INFO" or "OVERLAY". 762 Cmd->NonAlloc = true; 763 } 764 expect(")"); 765 return true; 766 } 767 768 // Reads an expression and/or the special directive for an output 769 // section definition. Directive is one of following: "(NOLOAD)", 770 // "(COPY)", "(INFO)" or "(OVERLAY)". 771 // 772 // An output section name can be followed by an address expression 773 // and/or directive. This grammar is not LL(1) because "(" can be 774 // interpreted as either the beginning of some expression or beginning 775 // of directive. 776 // 777 // https://sourceware.org/binutils/docs/ld/Output-Section-Address.html 778 // https://sourceware.org/binutils/docs/ld/Output-Section-Type.html 779 void ScriptParser::readSectionAddressType(OutputSection *Cmd) { 780 if (readSectionDirective(Cmd, peek(), peek2())) 781 return; 782 783 Cmd->AddrExpr = readExpr(); 784 if (peek() == "(" && !readSectionDirective(Cmd, "(", peek2())) 785 setError("unknown section directive: " + peek2()); 786 } 787 788 static Expr checkAlignment(Expr E, std::string &Loc) { 789 return [=] { 790 uint64_t Alignment = std::max((uint64_t)1, E().getValue()); 791 if (!isPowerOf2_64(Alignment)) { 792 error(Loc + ": alignment must be power of 2"); 793 return (uint64_t)1; // Return a dummy value. 794 } 795 return Alignment; 796 }; 797 } 798 799 OutputSection *ScriptParser::readOverlaySectionDescription() { 800 OutputSection *Cmd = 801 Script->createOutputSection(next(), getCurrentLocation()); 802 Cmd->InOverlay = true; 803 expect("{"); 804 while (!errorCount() && !consume("}")) 805 Cmd->SectionCommands.push_back(readInputSectionRules(next())); 806 Cmd->Phdrs = readOutputSectionPhdrs(); 807 return Cmd; 808 } 809 810 OutputSection *ScriptParser::readOutputSectionDescription(StringRef OutSec) { 811 OutputSection *Cmd = 812 Script->createOutputSection(OutSec, getCurrentLocation()); 813 814 size_t SymbolsReferenced = Script->ReferencedSymbols.size(); 815 816 if (peek() != ":") 817 readSectionAddressType(Cmd); 818 expect(":"); 819 820 std::string Location = getCurrentLocation(); 821 if (consume("AT")) 822 Cmd->LMAExpr = readParenExpr(); 823 if (consume("ALIGN")) 824 Cmd->AlignExpr = checkAlignment(readParenExpr(), Location); 825 if (consume("SUBALIGN")) 826 Cmd->SubalignExpr = checkAlignment(readParenExpr(), Location); 827 828 // Parse constraints. 829 if (consume("ONLY_IF_RO")) 830 Cmd->Constraint = ConstraintKind::ReadOnly; 831 if (consume("ONLY_IF_RW")) 832 Cmd->Constraint = ConstraintKind::ReadWrite; 833 expect("{"); 834 835 while (!errorCount() && !consume("}")) { 836 StringRef Tok = next(); 837 if (Tok == ";") { 838 // Empty commands are allowed. Do nothing here. 839 } else if (SymbolAssignment *Assign = readAssignment(Tok)) { 840 Cmd->SectionCommands.push_back(Assign); 841 } else if (ByteCommand *Data = readByteCommand(Tok)) { 842 Cmd->SectionCommands.push_back(Data); 843 } else if (Tok == "CONSTRUCTORS") { 844 // CONSTRUCTORS is a keyword to make the linker recognize C++ ctors/dtors 845 // by name. This is for very old file formats such as ECOFF/XCOFF. 846 // For ELF, we should ignore. 847 } else if (Tok == "FILL") { 848 Cmd->Filler = readFill(); 849 } else if (Tok == "SORT") { 850 readSort(); 851 } else if (Tok == "INCLUDE") { 852 readInclude(); 853 } else if (peek() == "(") { 854 Cmd->SectionCommands.push_back(readInputSectionDescription(Tok)); 855 } else { 856 // We have a file name and no input sections description. It is not a 857 // commonly used syntax, but still acceptable. In that case, all sections 858 // from the file will be included. 859 auto *ISD = make<InputSectionDescription>(Tok); 860 ISD->SectionPatterns.push_back({{}, StringMatcher({"*"})}); 861 Cmd->SectionCommands.push_back(ISD); 862 } 863 } 864 865 if (consume(">")) 866 Cmd->MemoryRegionName = next(); 867 868 if (consume("AT")) { 869 expect(">"); 870 Cmd->LMARegionName = next(); 871 } 872 873 if (Cmd->LMAExpr && !Cmd->LMARegionName.empty()) 874 error("section can't have both LMA and a load region"); 875 876 Cmd->Phdrs = readOutputSectionPhdrs(); 877 878 if (consume("=")) 879 Cmd->Filler = parseFill(next()); 880 else if (peek().startswith("=")) 881 Cmd->Filler = parseFill(next().drop_front()); 882 883 // Consume optional comma following output section command. 884 consume(","); 885 886 if (Script->ReferencedSymbols.size() > SymbolsReferenced) 887 Cmd->ExpressionsUseSymbols = true; 888 return Cmd; 889 } 890 891 // Parses a given string as a octal/decimal/hexadecimal number and 892 // returns it as a big-endian number. Used for `=<fillexp>`. 893 // https://sourceware.org/binutils/docs/ld/Output-Section-Fill.html 894 // 895 // When reading a hexstring, ld.bfd handles it as a blob of arbitrary 896 // size, while ld.gold always handles it as a 32-bit big-endian number. 897 // We are compatible with ld.gold because it's easier to implement. 898 std::array<uint8_t, 4> ScriptParser::parseFill(StringRef Tok) { 899 uint32_t V = 0; 900 if (!to_integer(Tok, V)) 901 setError("invalid filler expression: " + Tok); 902 903 std::array<uint8_t, 4> Buf; 904 write32be(Buf.data(), V); 905 return Buf; 906 } 907 908 SymbolAssignment *ScriptParser::readProvideHidden(bool Provide, bool Hidden) { 909 expect("("); 910 SymbolAssignment *Cmd = readSymbolAssignment(next()); 911 Cmd->Provide = Provide; 912 Cmd->Hidden = Hidden; 913 expect(")"); 914 return Cmd; 915 } 916 917 SymbolAssignment *ScriptParser::readAssignment(StringRef Tok) { 918 // Assert expression returns Dot, so this is equal to ".=." 919 if (Tok == "ASSERT") 920 return make<SymbolAssignment>(".", readAssert(), getCurrentLocation()); 921 922 size_t OldPos = Pos; 923 SymbolAssignment *Cmd = nullptr; 924 if (peek() == "=" || peek() == "+=") 925 Cmd = readSymbolAssignment(Tok); 926 else if (Tok == "PROVIDE") 927 Cmd = readProvideHidden(true, false); 928 else if (Tok == "HIDDEN") 929 Cmd = readProvideHidden(false, true); 930 else if (Tok == "PROVIDE_HIDDEN") 931 Cmd = readProvideHidden(true, true); 932 933 if (Cmd) { 934 Cmd->CommandString = 935 Tok.str() + " " + 936 llvm::join(Tokens.begin() + OldPos, Tokens.begin() + Pos, " "); 937 expect(";"); 938 } 939 return Cmd; 940 } 941 942 SymbolAssignment *ScriptParser::readSymbolAssignment(StringRef Name) { 943 StringRef Op = next(); 944 assert(Op == "=" || Op == "+="); 945 Expr E = readExpr(); 946 if (Op == "+=") { 947 std::string Loc = getCurrentLocation(); 948 E = [=] { return add(Script->getSymbolValue(Name, Loc), E()); }; 949 } 950 return make<SymbolAssignment>(Name, E, getCurrentLocation()); 951 } 952 953 // This is an operator-precedence parser to parse a linker 954 // script expression. 955 Expr ScriptParser::readExpr() { 956 // Our lexer is context-aware. Set the in-expression bit so that 957 // they apply different tokenization rules. 958 bool Orig = InExpr; 959 InExpr = true; 960 Expr E = readExpr1(readPrimary(), 0); 961 InExpr = Orig; 962 return E; 963 } 964 965 Expr ScriptParser::combine(StringRef Op, Expr L, Expr R) { 966 if (Op == "+") 967 return [=] { return add(L(), R()); }; 968 if (Op == "-") 969 return [=] { return sub(L(), R()); }; 970 if (Op == "*") 971 return [=] { return L().getValue() * R().getValue(); }; 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 + ": division by zero"); 978 return 0; 979 }; 980 } 981 if (Op == "%") { 982 std::string Loc = getCurrentLocation(); 983 return [=]() -> uint64_t { 984 if (uint64_t RV = R().getValue()) 985 return L().getValue() % RV; 986 error(Loc + ": modulo by zero"); 987 return 0; 988 }; 989 } 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 L().getValue() == R().getValue(); }; 1004 if (Op == "!=") 1005 return [=] { return L().getValue() != R().getValue(); }; 1006 if (Op == "||") 1007 return [=] { return L().getValue() || R().getValue(); }; 1008 if (Op == "&&") 1009 return [=] { return L().getValue() && R().getValue(); }; 1010 if (Op == "&") 1011 return [=] { return bitAnd(L(), R()); }; 1012 if (Op == "|") 1013 return [=] { return bitOr(L(), R()); }; 1014 llvm_unreachable("invalid operator"); 1015 } 1016 1017 // This is a part of the operator-precedence parser. This function 1018 // assumes that the remaining token stream starts with an operator. 1019 Expr ScriptParser::readExpr1(Expr Lhs, int MinPrec) { 1020 while (!atEOF() && !errorCount()) { 1021 // Read an operator and an expression. 1022 if (consume("?")) 1023 return readTernary(Lhs); 1024 StringRef Op1 = peek(); 1025 if (precedence(Op1) < MinPrec) 1026 break; 1027 skip(); 1028 Expr Rhs = readPrimary(); 1029 1030 // Evaluate the remaining part of the expression first if the 1031 // next operator has greater precedence than the previous one. 1032 // For example, if we have read "+" and "3", and if the next 1033 // operator is "*", then we'll evaluate 3 * ... part first. 1034 while (!atEOF()) { 1035 StringRef Op2 = peek(); 1036 if (precedence(Op2) <= precedence(Op1)) 1037 break; 1038 Rhs = readExpr1(Rhs, precedence(Op2)); 1039 } 1040 1041 Lhs = combine(Op1, Lhs, Rhs); 1042 } 1043 return Lhs; 1044 } 1045 1046 Expr ScriptParser::getPageSize() { 1047 std::string Location = getCurrentLocation(); 1048 return [=]() -> uint64_t { 1049 if (Target) 1050 return Target->PageSize; 1051 error(Location + ": unable to calculate page size"); 1052 return 4096; // Return a dummy value. 1053 }; 1054 } 1055 1056 Expr ScriptParser::readConstant() { 1057 StringRef S = readParenLiteral(); 1058 if (S == "COMMONPAGESIZE") 1059 return getPageSize(); 1060 if (S == "MAXPAGESIZE") 1061 return [] { return Config->MaxPageSize; }; 1062 setError("unknown constant: " + S); 1063 return [] { return 0; }; 1064 } 1065 1066 // Parses Tok as an integer. It recognizes hexadecimal (prefixed with 1067 // "0x" or suffixed with "H") and decimal numbers. Decimal numbers may 1068 // have "K" (Ki) or "M" (Mi) suffixes. 1069 static Optional<uint64_t> parseInt(StringRef Tok) { 1070 // Hexadecimal 1071 uint64_t Val; 1072 if (Tok.startswith_lower("0x")) { 1073 if (!to_integer(Tok.substr(2), Val, 16)) 1074 return None; 1075 return Val; 1076 } 1077 if (Tok.endswith_lower("H")) { 1078 if (!to_integer(Tok.drop_back(), Val, 16)) 1079 return None; 1080 return Val; 1081 } 1082 1083 // Decimal 1084 if (Tok.endswith_lower("K")) { 1085 if (!to_integer(Tok.drop_back(), Val, 10)) 1086 return None; 1087 return Val * 1024; 1088 } 1089 if (Tok.endswith_lower("M")) { 1090 if (!to_integer(Tok.drop_back(), Val, 10)) 1091 return None; 1092 return Val * 1024 * 1024; 1093 } 1094 if (!to_integer(Tok, Val, 10)) 1095 return None; 1096 return Val; 1097 } 1098 1099 ByteCommand *ScriptParser::readByteCommand(StringRef Tok) { 1100 int Size = StringSwitch<int>(Tok) 1101 .Case("BYTE", 1) 1102 .Case("SHORT", 2) 1103 .Case("LONG", 4) 1104 .Case("QUAD", 8) 1105 .Default(-1); 1106 if (Size == -1) 1107 return nullptr; 1108 1109 size_t OldPos = Pos; 1110 Expr E = readParenExpr(); 1111 std::string CommandString = 1112 Tok.str() + " " + 1113 llvm::join(Tokens.begin() + OldPos, Tokens.begin() + Pos, " "); 1114 return make<ByteCommand>(E, Size, CommandString); 1115 } 1116 1117 StringRef ScriptParser::readParenLiteral() { 1118 expect("("); 1119 bool Orig = InExpr; 1120 InExpr = false; 1121 StringRef Tok = next(); 1122 InExpr = Orig; 1123 expect(")"); 1124 return Tok; 1125 } 1126 1127 static void checkIfExists(OutputSection *Cmd, StringRef Location) { 1128 if (Cmd->Location.empty() && Script->ErrorOnMissingSection) 1129 error(Location + ": undefined section " + Cmd->Name); 1130 } 1131 1132 Expr ScriptParser::readPrimary() { 1133 if (peek() == "(") 1134 return readParenExpr(); 1135 1136 if (consume("~")) { 1137 Expr E = readPrimary(); 1138 return [=] { return ~E().getValue(); }; 1139 } 1140 if (consume("!")) { 1141 Expr E = readPrimary(); 1142 return [=] { return !E().getValue(); }; 1143 } 1144 if (consume("-")) { 1145 Expr E = readPrimary(); 1146 return [=] { return -E().getValue(); }; 1147 } 1148 1149 StringRef Tok = next(); 1150 std::string Location = getCurrentLocation(); 1151 1152 // Built-in functions are parsed here. 1153 // https://sourceware.org/binutils/docs/ld/Builtin-Functions.html. 1154 if (Tok == "ABSOLUTE") { 1155 Expr Inner = readParenExpr(); 1156 return [=] { 1157 ExprValue I = Inner(); 1158 I.ForceAbsolute = true; 1159 return I; 1160 }; 1161 } 1162 if (Tok == "ADDR") { 1163 StringRef Name = readParenLiteral(); 1164 OutputSection *Sec = Script->getOrCreateOutputSection(Name); 1165 return [=]() -> ExprValue { 1166 checkIfExists(Sec, Location); 1167 return {Sec, false, 0, Location}; 1168 }; 1169 } 1170 if (Tok == "ALIGN") { 1171 expect("("); 1172 Expr E = readExpr(); 1173 if (consume(")")) { 1174 E = checkAlignment(E, Location); 1175 return [=] { return alignTo(Script->getDot(), E().getValue()); }; 1176 } 1177 expect(","); 1178 Expr E2 = checkAlignment(readExpr(), Location); 1179 expect(")"); 1180 return [=] { 1181 ExprValue V = E(); 1182 V.Alignment = E2().getValue(); 1183 return V; 1184 }; 1185 } 1186 if (Tok == "ALIGNOF") { 1187 StringRef Name = readParenLiteral(); 1188 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1189 return [=] { 1190 checkIfExists(Cmd, Location); 1191 return Cmd->Alignment; 1192 }; 1193 } 1194 if (Tok == "ASSERT") 1195 return readAssert(); 1196 if (Tok == "CONSTANT") 1197 return readConstant(); 1198 if (Tok == "DATA_SEGMENT_ALIGN") { 1199 expect("("); 1200 Expr E = readExpr(); 1201 expect(","); 1202 readExpr(); 1203 expect(")"); 1204 return [=] { 1205 return alignTo(Script->getDot(), std::max((uint64_t)1, E().getValue())); 1206 }; 1207 } 1208 if (Tok == "DATA_SEGMENT_END") { 1209 expect("("); 1210 expect("."); 1211 expect(")"); 1212 return [] { return Script->getDot(); }; 1213 } 1214 if (Tok == "DATA_SEGMENT_RELRO_END") { 1215 // GNU linkers implements more complicated logic to handle 1216 // DATA_SEGMENT_RELRO_END. We instead ignore the arguments and 1217 // just align to the next page boundary for simplicity. 1218 expect("("); 1219 readExpr(); 1220 expect(","); 1221 readExpr(); 1222 expect(")"); 1223 Expr E = getPageSize(); 1224 return [=] { return alignTo(Script->getDot(), E().getValue()); }; 1225 } 1226 if (Tok == "DEFINED") { 1227 StringRef Name = readParenLiteral(); 1228 return [=] { return Symtab->find(Name) ? 1 : 0; }; 1229 } 1230 if (Tok == "LENGTH") { 1231 StringRef Name = readParenLiteral(); 1232 if (Script->MemoryRegions.count(Name) == 0) { 1233 setError("memory region not defined: " + Name); 1234 return [] { return 0; }; 1235 } 1236 return [=] { return Script->MemoryRegions[Name]->Length; }; 1237 } 1238 if (Tok == "LOADADDR") { 1239 StringRef Name = readParenLiteral(); 1240 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1241 return [=] { 1242 checkIfExists(Cmd, Location); 1243 return Cmd->getLMA(); 1244 }; 1245 } 1246 if (Tok == "MAX" || Tok == "MIN") { 1247 expect("("); 1248 Expr A = readExpr(); 1249 expect(","); 1250 Expr B = readExpr(); 1251 expect(")"); 1252 if (Tok == "MIN") 1253 return [=] { return std::min(A().getValue(), B().getValue()); }; 1254 return [=] { return std::max(A().getValue(), B().getValue()); }; 1255 } 1256 if (Tok == "ORIGIN") { 1257 StringRef Name = readParenLiteral(); 1258 if (Script->MemoryRegions.count(Name) == 0) { 1259 setError("memory region not defined: " + Name); 1260 return [] { return 0; }; 1261 } 1262 return [=] { return Script->MemoryRegions[Name]->Origin; }; 1263 } 1264 if (Tok == "SEGMENT_START") { 1265 expect("("); 1266 skip(); 1267 expect(","); 1268 Expr E = readExpr(); 1269 expect(")"); 1270 return [=] { return E(); }; 1271 } 1272 if (Tok == "SIZEOF") { 1273 StringRef Name = readParenLiteral(); 1274 OutputSection *Cmd = Script->getOrCreateOutputSection(Name); 1275 // Linker script does not create an output section if its content is empty. 1276 // We want to allow SIZEOF(.foo) where .foo is a section which happened to 1277 // be empty. 1278 return [=] { return Cmd->Size; }; 1279 } 1280 if (Tok == "SIZEOF_HEADERS") 1281 return [=] { return elf::getHeaderSize(); }; 1282 1283 // Tok is the dot. 1284 if (Tok == ".") 1285 return [=] { return Script->getSymbolValue(Tok, Location); }; 1286 1287 // Tok is a literal number. 1288 if (Optional<uint64_t> Val = parseInt(Tok)) 1289 return [=] { return *Val; }; 1290 1291 // Tok is a symbol name. 1292 if (!isValidCIdentifier(Tok)) 1293 setError("malformed number: " + Tok); 1294 Script->ReferencedSymbols.push_back(Tok); 1295 return [=] { return Script->getSymbolValue(Tok, Location); }; 1296 } 1297 1298 Expr ScriptParser::readTernary(Expr Cond) { 1299 Expr L = readExpr(); 1300 expect(":"); 1301 Expr R = readExpr(); 1302 return [=] { return Cond().getValue() ? L() : R(); }; 1303 } 1304 1305 Expr ScriptParser::readParenExpr() { 1306 expect("("); 1307 Expr E = readExpr(); 1308 expect(")"); 1309 return E; 1310 } 1311 1312 std::vector<StringRef> ScriptParser::readOutputSectionPhdrs() { 1313 std::vector<StringRef> Phdrs; 1314 while (!errorCount() && peek().startswith(":")) { 1315 StringRef Tok = next(); 1316 Phdrs.push_back((Tok.size() == 1) ? next() : Tok.substr(1)); 1317 } 1318 return Phdrs; 1319 } 1320 1321 // Read a program header type name. The next token must be a 1322 // name of a program header type or a constant (e.g. "0x3"). 1323 unsigned ScriptParser::readPhdrType() { 1324 StringRef Tok = next(); 1325 if (Optional<uint64_t> Val = parseInt(Tok)) 1326 return *Val; 1327 1328 unsigned Ret = StringSwitch<unsigned>(Tok) 1329 .Case("PT_NULL", PT_NULL) 1330 .Case("PT_LOAD", PT_LOAD) 1331 .Case("PT_DYNAMIC", PT_DYNAMIC) 1332 .Case("PT_INTERP", PT_INTERP) 1333 .Case("PT_NOTE", PT_NOTE) 1334 .Case("PT_SHLIB", PT_SHLIB) 1335 .Case("PT_PHDR", PT_PHDR) 1336 .Case("PT_TLS", PT_TLS) 1337 .Case("PT_GNU_EH_FRAME", PT_GNU_EH_FRAME) 1338 .Case("PT_GNU_STACK", PT_GNU_STACK) 1339 .Case("PT_GNU_RELRO", PT_GNU_RELRO) 1340 .Case("PT_OPENBSD_RANDOMIZE", PT_OPENBSD_RANDOMIZE) 1341 .Case("PT_OPENBSD_WXNEEDED", PT_OPENBSD_WXNEEDED) 1342 .Case("PT_OPENBSD_BOOTDATA", PT_OPENBSD_BOOTDATA) 1343 .Default(-1); 1344 1345 if (Ret == (unsigned)-1) { 1346 setError("invalid program header type: " + Tok); 1347 return PT_NULL; 1348 } 1349 return Ret; 1350 } 1351 1352 // Reads an anonymous version declaration. 1353 void ScriptParser::readAnonymousDeclaration() { 1354 std::vector<SymbolVersion> Locals; 1355 std::vector<SymbolVersion> Globals; 1356 std::tie(Locals, Globals) = readSymbols(); 1357 1358 for (SymbolVersion V : Locals) { 1359 if (V.Name == "*") 1360 Config->DefaultSymbolVersion = VER_NDX_LOCAL; 1361 else 1362 Config->VersionScriptLocals.push_back(V); 1363 } 1364 1365 for (SymbolVersion V : Globals) 1366 Config->VersionScriptGlobals.push_back(V); 1367 1368 expect(";"); 1369 } 1370 1371 // Reads a non-anonymous version definition, 1372 // e.g. "VerStr { global: foo; bar; local: *; };". 1373 void ScriptParser::readVersionDeclaration(StringRef VerStr) { 1374 // Read a symbol list. 1375 std::vector<SymbolVersion> Locals; 1376 std::vector<SymbolVersion> Globals; 1377 std::tie(Locals, Globals) = readSymbols(); 1378 1379 for (SymbolVersion V : Locals) { 1380 if (V.Name == "*") 1381 Config->DefaultSymbolVersion = VER_NDX_LOCAL; 1382 else 1383 Config->VersionScriptLocals.push_back(V); 1384 } 1385 1386 // Create a new version definition and add that to the global symbols. 1387 VersionDefinition Ver; 1388 Ver.Name = VerStr; 1389 Ver.Globals = Globals; 1390 1391 // User-defined version number starts from 2 because 0 and 1 are 1392 // reserved for VER_NDX_LOCAL and VER_NDX_GLOBAL, respectively. 1393 Ver.Id = Config->VersionDefinitions.size() + 2; 1394 Config->VersionDefinitions.push_back(Ver); 1395 1396 // Each version may have a parent version. For example, "Ver2" 1397 // defined as "Ver2 { global: foo; local: *; } Ver1;" has "Ver1" 1398 // as a parent. This version hierarchy is, probably against your 1399 // instinct, purely for hint; the runtime doesn't care about it 1400 // at all. In LLD, we simply ignore it. 1401 if (peek() != ";") 1402 skip(); 1403 expect(";"); 1404 } 1405 1406 static bool hasWildcard(StringRef S) { 1407 return S.find_first_of("?*[") != StringRef::npos; 1408 } 1409 1410 // Reads a list of symbols, e.g. "{ global: foo; bar; local: *; };". 1411 std::pair<std::vector<SymbolVersion>, std::vector<SymbolVersion>> 1412 ScriptParser::readSymbols() { 1413 std::vector<SymbolVersion> Locals; 1414 std::vector<SymbolVersion> Globals; 1415 std::vector<SymbolVersion> *V = &Globals; 1416 1417 while (!errorCount()) { 1418 if (consume("}")) 1419 break; 1420 if (consumeLabel("local")) { 1421 V = &Locals; 1422 continue; 1423 } 1424 if (consumeLabel("global")) { 1425 V = &Globals; 1426 continue; 1427 } 1428 1429 if (consume("extern")) { 1430 std::vector<SymbolVersion> Ext = readVersionExtern(); 1431 V->insert(V->end(), Ext.begin(), Ext.end()); 1432 } else { 1433 StringRef Tok = next(); 1434 V->push_back({unquote(Tok), false, hasWildcard(Tok)}); 1435 } 1436 expect(";"); 1437 } 1438 return {Locals, Globals}; 1439 } 1440 1441 // Reads an "extern C++" directive, e.g., 1442 // "extern "C++" { ns::*; "f(int, double)"; };" 1443 // 1444 // The last semicolon is optional. E.g. this is OK: 1445 // "extern "C++" { ns::*; "f(int, double)" };" 1446 std::vector<SymbolVersion> ScriptParser::readVersionExtern() { 1447 StringRef Tok = next(); 1448 bool IsCXX = Tok == "\"C++\""; 1449 if (!IsCXX && Tok != "\"C\"") 1450 setError("Unknown language"); 1451 expect("{"); 1452 1453 std::vector<SymbolVersion> Ret; 1454 while (!errorCount() && peek() != "}") { 1455 StringRef Tok = next(); 1456 bool HasWildcard = !Tok.startswith("\"") && hasWildcard(Tok); 1457 Ret.push_back({unquote(Tok), IsCXX, HasWildcard}); 1458 if (consume("}")) 1459 return Ret; 1460 expect(";"); 1461 } 1462 1463 expect("}"); 1464 return Ret; 1465 } 1466 1467 uint64_t ScriptParser::readMemoryAssignment(StringRef S1, StringRef S2, 1468 StringRef S3) { 1469 if (!consume(S1) && !consume(S2) && !consume(S3)) { 1470 setError("expected one of: " + S1 + ", " + S2 + ", or " + S3); 1471 return 0; 1472 } 1473 expect("="); 1474 return readExpr()().getValue(); 1475 } 1476 1477 // Parse the MEMORY command as specified in: 1478 // https://sourceware.org/binutils/docs/ld/MEMORY.html 1479 // 1480 // MEMORY { name [(attr)] : ORIGIN = origin, LENGTH = len ... } 1481 void ScriptParser::readMemory() { 1482 expect("{"); 1483 while (!errorCount() && !consume("}")) { 1484 StringRef Tok = next(); 1485 if (Tok == "INCLUDE") { 1486 readInclude(); 1487 continue; 1488 } 1489 1490 uint32_t Flags = 0; 1491 uint32_t NegFlags = 0; 1492 if (consume("(")) { 1493 std::tie(Flags, NegFlags) = readMemoryAttributes(); 1494 expect(")"); 1495 } 1496 expect(":"); 1497 1498 uint64_t Origin = readMemoryAssignment("ORIGIN", "org", "o"); 1499 expect(","); 1500 uint64_t Length = readMemoryAssignment("LENGTH", "len", "l"); 1501 1502 // Add the memory region to the region map. 1503 MemoryRegion *MR = make<MemoryRegion>(Tok, Origin, Length, Flags, NegFlags); 1504 if (!Script->MemoryRegions.insert({Tok, MR}).second) 1505 setError("region '" + Tok + "' already defined"); 1506 } 1507 } 1508 1509 // This function parses the attributes used to match against section 1510 // flags when placing output sections in a memory region. These flags 1511 // are only used when an explicit memory region name is not used. 1512 std::pair<uint32_t, uint32_t> ScriptParser::readMemoryAttributes() { 1513 uint32_t Flags = 0; 1514 uint32_t NegFlags = 0; 1515 bool Invert = false; 1516 1517 for (char C : next().lower()) { 1518 uint32_t Flag = 0; 1519 if (C == '!') 1520 Invert = !Invert; 1521 else if (C == 'w') 1522 Flag = SHF_WRITE; 1523 else if (C == 'x') 1524 Flag = SHF_EXECINSTR; 1525 else if (C == 'a') 1526 Flag = SHF_ALLOC; 1527 else if (C != 'r') 1528 setError("invalid memory region attribute"); 1529 1530 if (Invert) 1531 NegFlags |= Flag; 1532 else 1533 Flags |= Flag; 1534 } 1535 return {Flags, NegFlags}; 1536 } 1537 1538 void elf::readLinkerScript(MemoryBufferRef MB) { 1539 ScriptParser(MB).readLinkerScript(); 1540 } 1541 1542 void elf::readVersionScript(MemoryBufferRef MB) { 1543 ScriptParser(MB).readVersionScript(); 1544 } 1545 1546 void elf::readDynamicList(MemoryBufferRef MB) { 1547 ScriptParser(MB).readDynamicList(); 1548 } 1549 1550 void elf::readDefsym(StringRef Name, MemoryBufferRef MB) { 1551 ScriptParser(MB).readDefsym(Name); 1552 } 1553