From ef99356dfebb96f6f90efb912c2877214bad060e Mon Sep 17 00:00:00 2001 From: Ahmed Bougacha Date: Fri, 24 May 2013 01:07:04 +0000 Subject: MC: Disassembled CFG reconstruction. This patch builds on some existing code to do CFG reconstruction from a disassembled binary: - MCModule represents the binary, and has a list of MCAtoms. - MCAtom represents either disassembled instructions (MCTextAtom), or contiguous data (MCDataAtom), and covers a specific range of addresses. - MCBasicBlock and MCFunction form the reconstructed CFG. An MCBB is backed by an MCTextAtom, and has the usual successors/predecessors. - MCObjectDisassembler creates a module from an ObjectFile using a disassembler. It first builds an atom for each section. It can also construct the CFG, and this splits the text atoms into basic blocks. MCModule and MCAtom were only sketched out; MCFunction and MCBB were implemented under the experimental "-cfg" llvm-objdump -macho option. This cleans them up for further use; llvm-objdump -d -cfg now generates graphviz files for each function found in the binary. In the future, MCObjectDisassembler may be the right place to do "intelligent" disassembly: for example, handling constant islands is just a matter of splitting the atom, using information that may be available in the ObjectFile. Also, better initial atom formation than just using sections is possible using symbols (and things like Mach-O's function_starts load command). This brings two minor regressions in llvm-objdump -macho -cfg: - The printing of a relocation's referenced symbol. - An annotation on loop BBs, i.e., which are their own successor. Relocation printing is replaced by the MCSymbolizer; the basic CFG annotation will be superseded by more related functionality. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182628 91177308-0d34-0410-b5e6-96231b3b80d8 --- tools/llvm-objdump/CMakeLists.txt | 1 - tools/llvm-objdump/MCFunction.cpp | 138 --------------- tools/llvm-objdump/MCFunction.h | 100 ----------- tools/llvm-objdump/MachODump.cpp | 334 +++++------------------------------- tools/llvm-objdump/llvm-objdump.cpp | 89 +++++++++- tools/llvm-objdump/llvm-objdump.h | 21 +-- 6 files changed, 130 insertions(+), 553 deletions(-) delete mode 100644 tools/llvm-objdump/MCFunction.cpp delete mode 100644 tools/llvm-objdump/MCFunction.h (limited to 'tools/llvm-objdump') diff --git a/tools/llvm-objdump/CMakeLists.txt b/tools/llvm-objdump/CMakeLists.txt index 0c49d0b457..e983ec92fb 100644 --- a/tools/llvm-objdump/CMakeLists.txt +++ b/tools/llvm-objdump/CMakeLists.txt @@ -12,5 +12,4 @@ add_llvm_tool(llvm-objdump COFFDump.cpp ELFDump.cpp MachODump.cpp - MCFunction.cpp ) diff --git a/tools/llvm-objdump/MCFunction.cpp b/tools/llvm-objdump/MCFunction.cpp deleted file mode 100644 index 5c67f1b70a..0000000000 --- a/tools/llvm-objdump/MCFunction.cpp +++ /dev/null @@ -1,138 +0,0 @@ -//===-- MCFunction.cpp ----------------------------------------------------===// -// -// The LLVM Compiler Infrastructure -// -// This file is distributed under the University of Illinois Open Source -// License. See LICENSE.TXT for details. -// -//===----------------------------------------------------------------------===// -// -// This file defines the algorithm to break down a region of machine code -// into basic blocks and try to reconstruct a CFG from it. -// -//===----------------------------------------------------------------------===// - -#include "MCFunction.h" -#include "llvm/ADT/STLExtras.h" -#include "llvm/MC/MCDisassembler.h" -#include "llvm/MC/MCInst.h" -#include "llvm/MC/MCInstPrinter.h" -#include "llvm/MC/MCInstrAnalysis.h" -#include "llvm/MC/MCInstrDesc.h" -#include "llvm/MC/MCInstrInfo.h" -#include "llvm/Support/MemoryObject.h" -#include "llvm/Support/raw_ostream.h" -#include "llvm/Support/system_error.h" -#include -using namespace llvm; - -MCFunction -MCFunction::createFunctionFromMC(StringRef Name, const MCDisassembler *DisAsm, - const MemoryObject &Region, uint64_t Start, - uint64_t End, const MCInstrAnalysis *Ana, - raw_ostream &DebugOut, - SmallVectorImpl &Calls) { - std::vector Instructions; - std::set Splits; - Splits.insert(Start); - uint64_t Size; - - MCFunction f(Name); - - { - DenseSet VisitedInsts; - SmallVector WorkList; - WorkList.push_back(Start); - // Disassemble code and gather basic block split points. - while (!WorkList.empty()) { - uint64_t Index = WorkList.pop_back_val(); - if (VisitedInsts.find(Index) != VisitedInsts.end()) - continue; // Already visited this location. - - for (;Index < End; Index += Size) { - VisitedInsts.insert(Index); - - MCInst Inst; - if (DisAsm->getInstruction(Inst, Size, Region, Index, DebugOut, nulls())){ - Instructions.push_back(MCDecodedInst(Index, Size, Inst)); - if (Ana->isBranch(Inst)) { - uint64_t targ = Ana->evaluateBranch(Inst, Index, Size); - if (targ != -1ULL && targ == Index+Size) - continue; // Skip nop jumps. - - // If we could determine the branch target, make a note to start a - // new basic block there and add the target to the worklist. - if (targ != -1ULL) { - Splits.insert(targ); - WorkList.push_back(targ); - WorkList.push_back(Index+Size); - } - Splits.insert(Index+Size); - break; - } else if (Ana->isReturn(Inst)) { - // Return instruction. This basic block ends here. - Splits.insert(Index+Size); - break; - } else if (Ana->isCall(Inst)) { - uint64_t targ = Ana->evaluateBranch(Inst, Index, Size); - // Add the call to the call list if the destination is known. - if (targ != -1ULL && targ != Index+Size) - Calls.push_back(targ); - } - } else { - errs().write_hex(Index) << ": warning: invalid instruction encoding\n"; - if (Size == 0) - Size = 1; // skip illegible bytes - } - } - } - } - - // Make sure the instruction list is sorted. - std::sort(Instructions.begin(), Instructions.end()); - - // Create basic blocks. - unsigned ii = 0, ie = Instructions.size(); - for (std::set::iterator spi = Splits.begin(), - spe = llvm::prior(Splits.end()); spi != spe; ++spi) { - MCBasicBlock BB; - uint64_t BlockEnd = *llvm::next(spi); - // Add instructions to the BB. - for (; ii != ie; ++ii) { - if (Instructions[ii].Address < *spi || - Instructions[ii].Address >= BlockEnd) - break; - BB.addInst(Instructions[ii]); - } - f.addBlock(*spi, BB); - } - - std::sort(f.Blocks.begin(), f.Blocks.end()); - - // Calculate successors of each block. - for (MCFunction::iterator i = f.begin(), e = f.end(); i != e; ++i) { - MCBasicBlock &BB = const_cast(i->second); - if (BB.getInsts().empty()) continue; - const MCDecodedInst &Inst = BB.getInsts().back(); - - if (Ana->isBranch(Inst.Inst)) { - uint64_t targ = Ana->evaluateBranch(Inst.Inst, Inst.Address, Inst.Size); - if (targ == -1ULL) { - // Indirect branch. Bail and add all blocks of the function as a - // successor. - for (MCFunction::iterator i = f.begin(), e = f.end(); i != e; ++i) - BB.addSucc(i->first); - } else if (targ != Inst.Address+Inst.Size) - BB.addSucc(targ); - // Conditional branches can also fall through to the next block. - if (Ana->isConditionalBranch(Inst.Inst) && llvm::next(i) != e) - BB.addSucc(llvm::next(i)->first); - } else { - // No branch. Fall through to the next block. - if (!Ana->isReturn(Inst.Inst) && llvm::next(i) != e) - BB.addSucc(llvm::next(i)->first); - } - } - - return f; -} diff --git a/tools/llvm-objdump/MCFunction.h b/tools/llvm-objdump/MCFunction.h deleted file mode 100644 index 6d3a548d48..0000000000 --- a/tools/llvm-objdump/MCFunction.h +++ /dev/null @@ -1,100 +0,0 @@ -//===-- MCFunction.h ------------------------------------------------------===// -// -// The LLVM Compiler Infrastructure -// -// This file is distributed under the University of Illinois Open Source -// License. See LICENSE.TXT for details. -// -//===----------------------------------------------------------------------===// -// -// This file defines the data structures to hold a CFG reconstructed from -// machine code. -// -//===----------------------------------------------------------------------===// - -#ifndef LLVM_OBJECTDUMP_MCFUNCTION_H -#define LLVM_OBJECTDUMP_MCFUNCTION_H - -#include "llvm/ADT/ArrayRef.h" -#include "llvm/ADT/DenseSet.h" -#include "llvm/MC/MCInst.h" -#include - -namespace llvm { - -class MCDisassembler; -class MCInstrAnalysis; -class MemoryObject; -class raw_ostream; - -/// MCDecodedInst - Small container to hold an MCInst and associated info like -/// address and size. -struct MCDecodedInst { - uint64_t Address; - uint64_t Size; - MCInst Inst; - - MCDecodedInst() {} - MCDecodedInst(uint64_t Address, uint64_t Size, MCInst Inst) - : Address(Address), Size(Size), Inst(Inst) {} - - bool operator<(const MCDecodedInst &RHS) const { - return Address < RHS.Address; - } -}; - -/// MCBasicBlock - Consists of multiple MCDecodedInsts and a list of successing -/// MCBasicBlocks. -class MCBasicBlock { - std::vector Insts; - typedef DenseSet SetTy; - SetTy Succs; -public: - ArrayRef getInsts() const { return Insts; } - - typedef SetTy::const_iterator succ_iterator; - succ_iterator succ_begin() const { return Succs.begin(); } - succ_iterator succ_end() const { return Succs.end(); } - - bool contains(uint64_t Addr) const { return Succs.count(Addr); } - - void addInst(const MCDecodedInst &Inst) { Insts.push_back(Inst); } - void addSucc(uint64_t Addr) { Succs.insert(Addr); } - - bool operator<(const MCBasicBlock &RHS) const { - return Insts.size() < RHS.Insts.size(); - } -}; - -/// MCFunction - Represents a named function in machine code, containing -/// multiple MCBasicBlocks. -class MCFunction { - const StringRef Name; - // Keep BBs sorted by address. - typedef std::vector > MapTy; - MapTy Blocks; -public: - MCFunction(StringRef Name) : Name(Name) {} - - // Create an MCFunction from a region of binary machine code. - static MCFunction - createFunctionFromMC(StringRef Name, const MCDisassembler *DisAsm, - const MemoryObject &Region, uint64_t Start, uint64_t End, - const MCInstrAnalysis *Ana, raw_ostream &DebugOut, - SmallVectorImpl &Calls); - - typedef MapTy::const_iterator iterator; - iterator begin() const { return Blocks.begin(); } - iterator end() const { return Blocks.end(); } - - StringRef getName() const { return Name; } - - MCBasicBlock &addBlock(uint64_t Address, const MCBasicBlock &BB) { - Blocks.push_back(std::make_pair(Address, BB)); - return Blocks.back().second; - } -}; - -} - -#endif diff --git a/tools/llvm-objdump/MachODump.cpp b/tools/llvm-objdump/MachODump.cpp index c7e5cc1ede..03a383eb12 100644 --- a/tools/llvm-objdump/MachODump.cpp +++ b/tools/llvm-objdump/MachODump.cpp @@ -12,9 +12,9 @@ //===----------------------------------------------------------------------===// #include "llvm-objdump.h" -#include "MCFunction.h" #include "llvm/ADT/OwningPtr.h" #include "llvm/ADT/STLExtras.h" +#include "llvm/ADT/StringExtras.h" #include "llvm/ADT/Triple.h" #include "llvm/DebugInfo/DIContext.h" #include "llvm/MC/MCAsmInfo.h" @@ -43,10 +43,6 @@ using namespace llvm; using namespace object; -static cl::opt - CFG("cfg", cl::desc("Create a CFG for every symbol in the object file and" - " write it to a graphviz file (MachO-only)")); - static cl::opt UseDbg("g", cl::desc("Print line information from debug info if available")); @@ -91,99 +87,6 @@ struct SymbolSorter { } }; -// Print additional information about an address, if available. -static void DumpAddress(uint64_t Address, ArrayRef Sections, - const MachOObjectFile *MachOObj, raw_ostream &OS) { - for (unsigned i = 0; i != Sections.size(); ++i) { - uint64_t SectAddr = 0, SectSize = 0; - Sections[i].getAddress(SectAddr); - Sections[i].getSize(SectSize); - uint64_t addr = SectAddr; - if (SectAddr <= Address && - SectAddr + SectSize > Address) { - StringRef bytes, name; - Sections[i].getContents(bytes); - Sections[i].getName(name); - // Print constant strings. - if (!name.compare("__cstring")) - OS << '"' << bytes.substr(addr, bytes.find('\0', addr)) << '"'; - // Print constant CFStrings. - if (!name.compare("__cfstring")) - OS << "@\"" << bytes.substr(addr, bytes.find('\0', addr)) << '"'; - } - } -} - -typedef std::map FunctionMapTy; -typedef SmallVector FunctionListTy; -static void createMCFunctionAndSaveCalls(StringRef Name, - const MCDisassembler *DisAsm, - MemoryObject &Object, uint64_t Start, - uint64_t End, - MCInstrAnalysis *InstrAnalysis, - uint64_t Address, - raw_ostream &DebugOut, - FunctionMapTy &FunctionMap, - FunctionListTy &Functions) { - SmallVector Calls; - MCFunction f = - MCFunction::createFunctionFromMC(Name, DisAsm, Object, Start, End, - InstrAnalysis, DebugOut, Calls); - Functions.push_back(f); - FunctionMap[Address] = &Functions.back(); - - // Add the gathered callees to the map. - for (unsigned i = 0, e = Calls.size(); i != e; ++i) - FunctionMap.insert(std::make_pair(Calls[i], (MCFunction*)0)); -} - -// Write a graphviz file for the CFG inside an MCFunction. -static void emitDOTFile(const char *FileName, const MCFunction &f, - MCInstPrinter *IP) { - // Start a new dot file. - std::string Error; - raw_fd_ostream Out(FileName, Error); - if (!Error.empty()) { - errs() << "llvm-objdump: warning: " << Error << '\n'; - return; - } - - Out << "digraph " << f.getName() << " {\n"; - Out << "graph [ rankdir = \"LR\" ];\n"; - for (MCFunction::iterator i = f.begin(), e = f.end(); i != e; ++i) { - bool hasPreds = false; - // Only print blocks that have predecessors. - // FIXME: Slow. - for (MCFunction::iterator pi = f.begin(), pe = f.end(); pi != pe; - ++pi) - if (pi->second.contains(i->first)) { - hasPreds = true; - break; - } - - if (!hasPreds && i != f.begin()) - continue; - - Out << '"' << i->first << "\" [ label=\""; - // Print instructions. - for (unsigned ii = 0, ie = i->second.getInsts().size(); ii != ie; - ++ii) { - // Escape special chars and print the instruction in mnemonic form. - std::string Str; - raw_string_ostream OS(Str); - IP->printInst(&i->second.getInsts()[ii].Inst, OS, ""); - Out << DOT::EscapeString(OS.str()) << '|'; - } - Out << "\" shape=\"record\" ];\n"; - - // Add edges. - for (MCBasicBlock::succ_iterator si = i->second.succ_begin(), - se = i->second.succ_end(); si != se; ++si) - Out << i->first << ":o -> " << *si <<":a\n"; - } - Out << "}\n"; -} - static void getSectionsAndSymbols(const macho::Header Header, MachOObjectFile *MachOObj, @@ -272,6 +175,12 @@ static void DisassembleInputMachO2(StringRef Filename, macho::Header Header = MachOOF->getHeader(); + // FIXME: FoundFns isn't used anymore. Using symbols/LC_FUNCTION_STARTS to + // determine function locations will eventually go in MCObjectDisassembler. + // FIXME: Using the -cfg command line option, this code used to be able to + // annotate relocations with the referenced symbol's name, and if this was + // inside a __[cf]string section, the data it points to. This is now replaced + // by the upcoming MCSymbolizer, which needs the appropriate setup done above. std::vector Sections; std::vector Symbols; SmallVector FoundFns; @@ -308,31 +217,24 @@ static void DisassembleInputMachO2(StringRef Filename, diContext.reset(DIContext::getDWARFContext(DbgObj)); } - FunctionMapTy FunctionMap; - FunctionListTy Functions; - for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) { + + bool SectIsText = false; + Sections[SectIdx].isText(SectIsText); + if (SectIsText == false) + continue; + StringRef SectName; if (Sections[SectIdx].getName(SectName) || SectName != "__text") continue; // Skip non-text sections DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl(); + StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR); if (SegmentName != "__TEXT") continue; - // Insert the functions from the function starts segment into our map. - uint64_t VMAddr; - Sections[SectIdx].getAddress(VMAddr); - for (unsigned i = 0, e = FoundFns.size(); i != e; ++i) { - StringRef SectBegin; - Sections[SectIdx].getContents(SectBegin); - uint64_t Offset = (uint64_t)SectBegin.data(); - FunctionMap.insert(std::make_pair(VMAddr + FoundFns[i]-Offset, - (MCFunction*)0)); - } - StringRef Bytes; Sections[SectIdx].getContents(Bytes); StringRefMemoryObject memoryObject(Bytes); @@ -403,52 +305,39 @@ static void DisassembleInputMachO2(StringRef Filename, symbolTableWorked = true; - if (!CFG) { - // Normal disassembly, print addresses, bytes and mnemonic form. - StringRef SymName; - Symbols[SymIdx].getName(SymName); - - outs() << SymName << ":\n"; - DILineInfo lastLine; - for (uint64_t Index = Start; Index < End; Index += Size) { - MCInst Inst; - - if (DisAsm->getInstruction(Inst, Size, memoryObject, Index, - DebugOut, nulls())) { - uint64_t SectAddress = 0; - Sections[SectIdx].getAddress(SectAddress); - outs() << format("%8" PRIx64 ":\t", SectAddress + Index); - - DumpBytes(StringRef(Bytes.data() + Index, Size)); - IP->printInst(&Inst, outs(), ""); - - // Print debug info. - if (diContext) { - DILineInfo dli = - diContext->getLineInfoForAddress(SectAddress + Index); - // Print valid line info if it changed. - if (dli != lastLine && dli.getLine() != 0) - outs() << "\t## " << dli.getFileName() << ':' - << dli.getLine() << ':' << dli.getColumn(); - lastLine = dli; - } - outs() << "\n"; - } else { - errs() << "llvm-objdump: warning: invalid instruction encoding\n"; - if (Size == 0) - Size = 1; // skip illegible bytes + outs() << SymName << ":\n"; + DILineInfo lastLine; + for (uint64_t Index = Start; Index < End; Index += Size) { + MCInst Inst; + + if (DisAsm->getInstruction(Inst, Size, memoryObject, Index, + DebugOut, nulls())) { + uint64_t SectAddress = 0; + Sections[SectIdx].getAddress(SectAddress); + outs() << format("%8" PRIx64 ":\t", SectAddress + Index); + + DumpBytes(StringRef(Bytes.data() + Index, Size)); + IP->printInst(&Inst, outs(), ""); + + // Print debug info. + if (diContext) { + DILineInfo dli = + diContext->getLineInfoForAddress(SectAddress + Index); + // Print valid line info if it changed. + if (dli != lastLine && dli.getLine() != 0) + outs() << "\t## " << dli.getFileName() << ':' + << dli.getLine() << ':' << dli.getColumn(); + lastLine = dli; } + outs() << "\n"; + } else { + errs() << "llvm-objdump: warning: invalid instruction encoding\n"; + if (Size == 0) + Size = 1; // skip illegible bytes } - } else { - // Create CFG and use it for disassembly. - StringRef SymName; - Symbols[SymIdx].getName(SymName); - createMCFunctionAndSaveCalls( - SymName, DisAsm.get(), memoryObject, Start, End, - InstrAnalysis.get(), Start, DebugOut, FunctionMap, Functions); } } - if (!CFG && !symbolTableWorked) { + if (!symbolTableWorked) { // Reading the symbol table didn't work, disassemble the whole section. uint64_t SectAddress; Sections[SectIdx].getAddress(SectAddress); @@ -471,142 +360,5 @@ static void DisassembleInputMachO2(StringRef Filename, } } } - - if (CFG) { - if (!symbolTableWorked) { - // Reading the symbol table didn't work, create a big __TEXT symbol. - uint64_t SectSize = 0, SectAddress = 0; - Sections[SectIdx].getSize(SectSize); - Sections[SectIdx].getAddress(SectAddress); - createMCFunctionAndSaveCalls("__TEXT", DisAsm.get(), memoryObject, - 0, SectSize, - InstrAnalysis.get(), - SectAddress, DebugOut, - FunctionMap, Functions); - } - for (std::map::iterator mi = FunctionMap.begin(), - me = FunctionMap.end(); mi != me; ++mi) - if (mi->second == 0) { - // Create functions for the remaining callees we have gathered, - // but we didn't find a name for them. - uint64_t SectSize = 0; - Sections[SectIdx].getSize(SectSize); - - SmallVector Calls; - MCFunction f = - MCFunction::createFunctionFromMC("unknown", DisAsm.get(), - memoryObject, mi->first, - SectSize, - InstrAnalysis.get(), DebugOut, - Calls); - Functions.push_back(f); - mi->second = &Functions.back(); - for (unsigned i = 0, e = Calls.size(); i != e; ++i) { - std::pair p(Calls[i], (MCFunction*)0); - if (FunctionMap.insert(p).second) - mi = FunctionMap.begin(); - } - } - - DenseSet PrintedBlocks; - for (unsigned ffi = 0, ffe = Functions.size(); ffi != ffe; ++ffi) { - MCFunction &f = Functions[ffi]; - for (MCFunction::iterator fi = f.begin(), fe = f.end(); fi != fe; ++fi){ - if (!PrintedBlocks.insert(fi->first).second) - continue; // We already printed this block. - - // We assume a block has predecessors when it's the first block after - // a symbol. - bool hasPreds = FunctionMap.find(fi->first) != FunctionMap.end(); - - // See if this block has predecessors. - // FIXME: Slow. - for (MCFunction::iterator pi = f.begin(), pe = f.end(); pi != pe; - ++pi) - if (pi->second.contains(fi->first)) { - hasPreds = true; - break; - } - - uint64_t SectSize = 0, SectAddress; - Sections[SectIdx].getSize(SectSize); - Sections[SectIdx].getAddress(SectAddress); - - // No predecessors, this is a data block. Print as .byte directives. - if (!hasPreds) { - uint64_t End = llvm::next(fi) == fe ? SectSize : - llvm::next(fi)->first; - outs() << "# " << End-fi->first << " bytes of data:\n"; - for (unsigned pos = fi->first; pos != End; ++pos) { - outs() << format("%8x:\t", SectAddress + pos); - DumpBytes(StringRef(Bytes.data() + pos, 1)); - outs() << format("\t.byte 0x%02x\n", (uint8_t)Bytes[pos]); - } - continue; - } - - if (fi->second.contains(fi->first)) // Print a header for simple loops - outs() << "# Loop begin:\n"; - - DILineInfo lastLine; - // Walk over the instructions and print them. - for (unsigned ii = 0, ie = fi->second.getInsts().size(); ii != ie; - ++ii) { - const MCDecodedInst &Inst = fi->second.getInsts()[ii]; - - // If there's a symbol at this address, print its name. - if (FunctionMap.find(SectAddress + Inst.Address) != - FunctionMap.end()) - outs() << FunctionMap[SectAddress + Inst.Address]-> getName() - << ":\n"; - - outs() << format("%8" PRIx64 ":\t", SectAddress + Inst.Address); - DumpBytes(StringRef(Bytes.data() + Inst.Address, Inst.Size)); - - if (fi->second.contains(fi->first)) // Indent simple loops. - outs() << '\t'; - - IP->printInst(&Inst.Inst, outs(), ""); - - // Look for relocations inside this instructions, if there is one - // print its target and additional information if available. - for (unsigned j = 0; j != Relocs.size(); ++j) - if (Relocs[j].first >= SectAddress + Inst.Address && - Relocs[j].first < SectAddress + Inst.Address + Inst.Size) { - StringRef SymName; - uint64_t Addr; - Relocs[j].second.getAddress(Addr); - Relocs[j].second.getName(SymName); - - outs() << "\t# " << SymName << ' '; - DumpAddress(Addr, Sections, MachOOF, outs()); - } - - // If this instructions contains an address, see if we can evaluate - // it and print additional information. - uint64_t targ = InstrAnalysis->evaluateBranch(Inst.Inst, - Inst.Address, - Inst.Size); - if (targ != -1ULL) - DumpAddress(targ, Sections, MachOOF, outs()); - - // Print debug info. - if (diContext) { - DILineInfo dli = - diContext->getLineInfoForAddress(SectAddress + Inst.Address); - // Print valid line info if it changed. - if (dli != lastLine && dli.getLine() != 0) - outs() << "\t## " << dli.getFileName() << ':' - << dli.getLine() << ':' << dli.getColumn(); - lastLine = dli; - } - - outs() << '\n'; - } - } - - emitDOTFile((f.getName().str() + ".dot").c_str(), f, IP.get()); - } - } } } diff --git a/tools/llvm-objdump/llvm-objdump.cpp b/tools/llvm-objdump/llvm-objdump.cpp index 570ec7ed6f..d8611d8b3d 100644 --- a/tools/llvm-objdump/llvm-objdump.cpp +++ b/tools/llvm-objdump/llvm-objdump.cpp @@ -17,22 +17,26 @@ //===----------------------------------------------------------------------===// #include "llvm-objdump.h" -#include "MCFunction.h" #include "llvm/ADT/OwningPtr.h" #include "llvm/ADT/STLExtras.h" #include "llvm/ADT/StringExtras.h" #include "llvm/ADT/Triple.h" #include "llvm/MC/MCAsmInfo.h" +#include "llvm/MC/MCAtom.h" #include "llvm/MC/MCContext.h" #include "llvm/MC/MCDisassembler.h" +#include "llvm/MC/MCFunction.h" #include "llvm/MC/MCInst.h" #include "llvm/MC/MCInstPrinter.h" +#include "llvm/MC/MCInstrAnalysis.h" #include "llvm/MC/MCInstrInfo.h" +#include "llvm/MC/MCModule.h" +#include "llvm/MC/MCObjectDisassembler.h" #include "llvm/MC/MCObjectFileInfo.h" #include "llvm/MC/MCObjectSymbolizer.h" #include "llvm/MC/MCRegisterInfo.h" -#include "llvm/MC/MCSubtargetInfo.h" #include "llvm/MC/MCRelocationInfo.h" +#include "llvm/MC/MCSubtargetInfo.h" #include "llvm/Object/Archive.h" #include "llvm/Object/COFF.h" #include "llvm/Object/MachO.h" @@ -131,6 +135,10 @@ static cl::opt Symbolize("symbolize", cl::desc("When disassembling instructions, " "try to symbolize operands.")); +static cl::opt +CFG("cfg", cl::desc("Create a CFG for every function found in the object" + " and write it to a graphviz file")); + static StringRef ToolName; bool llvm::error(error_code ec) { @@ -169,7 +177,51 @@ static const Target *getTarget(const ObjectFile *Obj = NULL) { return TheTarget; } -void llvm::StringRefMemoryObject::anchor() { } +// Write a graphviz file for the CFG inside an MCFunction. +static void emitDOTFile(const char *FileName, const MCFunction &f, + MCInstPrinter *IP) { + // Start a new dot file. + std::string Error; + raw_fd_ostream Out(FileName, Error); + if (!Error.empty()) { + errs() << "llvm-objdump: warning: " << Error << '\n'; + return; + } + + Out << "digraph \"" << f.getName() << "\" {\n"; + Out << "graph [ rankdir = \"LR\" ];\n"; + for (MCFunction::const_iterator i = f.begin(), e = f.end(); i != e; ++i) { + // Only print blocks that have predecessors. + bool hasPreds = (*i)->pred_begin() != (*i)->pred_end(); + + if (!hasPreds && i != f.begin()) + continue; + + Out << '"' << (*i)->getInsts()->getBeginAddr() << "\" [ label=\""; + // Print instructions. + for (unsigned ii = 0, ie = (*i)->getInsts()->size(); ii != ie; + ++ii) { + if (ii != 0) // Not the first line, start a new row. + Out << '|'; + if (ii + 1 == ie) // Last line, add an end id. + Out << ""; + + // Escape special chars and print the instruction in mnemonic form. + std::string Str; + raw_string_ostream OS(Str); + IP->printInst(&(*i)->getInsts()->at(ii).Inst, OS, ""); + Out << DOT::EscapeString(OS.str()); + } + Out << "\" shape=\"record\" ];\n"; + + // Add edges. + for (MCBasicBlock::succ_const_iterator si = (*i)->succ_begin(), + se = (*i)->succ_end(); si != se; ++si) + Out << (*i)->getInsts()->getBeginAddr() << ":o -> " + << (*si)->getInsts()->getBeginAddr() << ":a\n"; + } + Out << "}\n"; +} void llvm::DumpBytes(StringRef bytes) { static const char hex_rep[] = "0123456789abcdef"; @@ -269,6 +321,9 @@ static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { } } + OwningPtr + MIA(TheTarget->createMCInstrAnalysis(MII.get())); + int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); OwningPtr IP(TheTarget->createMCInstPrinter( AsmPrinterVariant, *AsmInfo, *MII, *MRI, *STI)); @@ -278,6 +333,34 @@ static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { return; } + if (CFG) { + OwningPtr OD( + new MCObjectDisassembler(*Obj, *DisAsm, *MIA)); + OwningPtr Mod(OD->buildModule(/* withCFG */ true)); + for (MCModule::const_atom_iterator AI = Mod->atom_begin(), + AE = Mod->atom_end(); + AI != AE; ++AI) { + outs() << "Atom " << (*AI)->getName() << ": \n"; + if (const MCTextAtom *TA = dyn_cast(*AI)) { + for (MCTextAtom::const_iterator II = TA->begin(), IE = TA->end(); + II != IE; + ++II) { + IP->printInst(&II->Inst, outs(), ""); + outs() << "\n"; + } + } + } + for (MCModule::const_func_iterator FI = Mod->func_begin(), + FE = Mod->func_end(); + FI != FE; ++FI) { + static int filenum = 0; + emitDOTFile((Twine((*FI)->getName()) + "_" + + utostr(filenum++) + ".dot").str().c_str(), + **FI, IP.get()); + } + } + + error_code ec; for (section_iterator i = Obj->begin_sections(), e = Obj->end_sections(); diff --git a/tools/llvm-objdump/llvm-objdump.h b/tools/llvm-objdump/llvm-objdump.h index 3c62240f8f..87f19ba257 100644 --- a/tools/llvm-objdump/llvm-objdump.h +++ b/tools/llvm-objdump/llvm-objdump.h @@ -13,7 +13,7 @@ #include "llvm/ADT/StringRef.h" #include "llvm/Support/CommandLine.h" #include "llvm/Support/DataTypes.h" -#include "llvm/Support/MemoryObject.h" +#include "llvm/Support/StringRefMemoryObject.h" namespace llvm { @@ -35,25 +35,6 @@ void DisassembleInputMachO(StringRef Filename); void printCOFFUnwindInfo(const object::COFFObjectFile* o); void printELFFileHeader(const object::ObjectFile *o); -class StringRefMemoryObject : public MemoryObject { - virtual void anchor(); - StringRef Bytes; - uint64_t Base; -public: - StringRefMemoryObject(StringRef bytes, uint64_t Base = 0) - : Bytes(bytes), Base(Base) {} - - uint64_t getBase() const { return Base; } - uint64_t getExtent() const { return Bytes.size(); } - - int readByte(uint64_t Addr, uint8_t *Byte) const { - if (Addr >= Base + getExtent() || Addr < Base) - return -1; - *Byte = Bytes[Addr - Base]; - return 0; - } -}; - } #endif -- cgit v1.2.3