From 2c94d0faa0e1c268893d5e04dc77e8a35889db00 Mon Sep 17 00:00:00 2001 From: Ahmed Bougacha Date: Fri, 24 May 2013 00:39:57 +0000 Subject: Add MCSymbolizer for symbolic/annotated disassembly. This is a basic first step towards symbolization of disassembled instructions. This used to be done using externally provided (C API) callbacks. This patch introduces: - the MCSymbolizer class, that mimics the same functions that were used in the X86 and ARM disassemblers to symbolize immediate operands and to annotate loads based off PC (for things like c string literals). - the MCExternalSymbolizer class, which implements the old C API. - the MCRelocationInfo class, which provides a way for targets to translate relocations (either object::RelocationRef, or disassembler C API VariantKinds) to MCExprs. - the MCObjectSymbolizer class, which does symbolization using what it finds in an object::ObjectFile. This makes simple symbolization (with no fancy relocation stuff) work for all object formats! - x86-64 Mach-O and ELF MCRelocationInfos. - A basic ARM Mach-O MCRelocationInfo, that provides just enough to support the C API VariantKinds. Most of what works in otool (the only user of the old symbolization API that I know of) for x86-64 symbolic disassembly (-tvV) works, namely: - symbol references: call _foo; jmp 15 <_foo+50> - relocations: call _foo-_bar; call _foo-4 - __cf?string: leaq 193(%rip), %rax ## literal pool for "hello" Stub support is the main missing part (because libObject doesn't know, among other things, about mach-o indirect symbols). As for the MCSymbolizer API, instead of relying on the disassemblers to call the tryAdding* methods, maybe this could be done automagically using InstrInfo? For instance, even though PC-relative LEAs are used to get the address of string literals in a typical Mach-O file, a MOV would be used in an ELF file. And right now, the explicit symbolization only recognizes PC-relative LEAs. InstrInfo should have already have most of what is needed to know what to symbolize, so this can definitely be improved. I'd also like to remove object::RelocationRef::getValueString (it seems only used by relocation printing in objdump), as simply printing the created MCExpr is definitely enough (and cleaner than string concats). git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182625 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/MC/MCObjectSymbolizer.cpp | 175 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 175 insertions(+) create mode 100644 lib/MC/MCObjectSymbolizer.cpp (limited to 'lib/MC/MCObjectSymbolizer.cpp') diff --git a/lib/MC/MCObjectSymbolizer.cpp b/lib/MC/MCObjectSymbolizer.cpp new file mode 100644 index 0000000000..e1d504ec8e --- /dev/null +++ b/lib/MC/MCObjectSymbolizer.cpp @@ -0,0 +1,175 @@ +//===-- lib/MC/MCObjectSymbolizer.cpp -------------------------------------===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// + +#include "llvm/MC/MCObjectSymbolizer.h" +#include "llvm/ADT/SmallString.h" +#include "llvm/MC/MCContext.h" +#include "llvm/MC/MCExpr.h" +#include "llvm/MC/MCInst.h" +#include "llvm/MC/MCRelocationInfo.h" +#include "llvm/MC/MCSymbol.h" +#include "llvm/Object/MachO.h" +#include "llvm/Object/ELF.h" +#include "llvm/Support/raw_ostream.h" + +using namespace llvm; +using namespace object; + +//===- MCMachObjectSymbolizer ---------------------------------------------===// + +namespace { +class MCMachObjectSymbolizer : public MCObjectSymbolizer { +public: + MCMachObjectSymbolizer(MCContext &Ctx, OwningPtr &RelInfo, + const object::MachOObjectFile *MachOOF) + : MCObjectSymbolizer(Ctx, RelInfo, MachOOF) + {} + + void tryAddingPcLoadReferenceComment(raw_ostream &cStream, + int64_t Value, uint64_t Address) { + AddrToRelocMap::iterator RI = AddrToReloc.find(Address); + if (RI != AddrToReloc.end()) { + const MCExpr *RelExpr = RelInfo->createExprForRelocation(RI->second); + if (!RelExpr || RelExpr->EvaluateAsAbsolute(Value) == false) + return; + } + uint64_t Addr = Value; + AddrToSectionMap::const_iterator SI = AddrToSection.find(Addr); + if (SI.valid()) { + DataRefImpl DRI; DRI.p = *SI; + SectionRef S(DRI, Obj); + StringRef Name; S.getName(Name); + if (Name == "__cstring") { + StringRef Contents; + S.getContents(Contents); + Contents = Contents.substr(Addr - SI.start()); + cStream << " ## literal pool for: " + << Contents.substr(0, Contents.find_first_of(0)); + } + } + } +}; +} // End unnamed namespace + +//===- MCObjectSymbolizer -------------------------------------------------===// + +MCObjectSymbolizer::MCObjectSymbolizer(MCContext &Ctx, + OwningPtr &RelInfo, + const ObjectFile *Obj) + : MCSymbolizer(Ctx, RelInfo), Obj(Obj), + AddrToSectionAllocator(), AddrToSection(AddrToSectionAllocator), + AddrToReloc() { + error_code ec; + for (section_iterator SI = Obj->begin_sections(), + SE = Obj->end_sections(); + SI != SE; + SI.increment(ec)) { + if (ec) break; + uint64_t StartAddr; SI->getAddress(StartAddr); + uint64_t Size; SI->getSize(Size); + StringRef SecName; SI->getName(SecName); + bool RequiredForExec; SI->isRequiredForExecution(RequiredForExec); + if (RequiredForExec == false || Size == 0) + continue; + AddrToSection.insert(StartAddr, StartAddr + Size - 1, + SI->getRawDataRefImpl().p); + for (relocation_iterator RI = SI->begin_relocations(), + RE = SI->end_relocations(); + RI != RE; + RI.increment(ec)) { + if (ec) break; + // FIXME: libObject is inconsistent regarding error handling. The + // overwhelming majority of methods always return object_error::success, + // and assert for simple errors.. Here, ELFObjectFile::getRelocationOffset + // asserts when the file type isn't ET_REL. + // This workaround handles x86-64 elf, the only one that has a relocinfo. + uint64_t Offset; + if (Obj->isELF()) { + const ELF64LEObjectFile *ELFObj = dyn_cast(Obj); + if (ELFObj == 0) + break; + if (ELFObj->getElfHeader()->e_type == ELF::ET_REL) { + RI->getOffset(Offset); + Offset += StartAddr; + } else { + RI->getAddress(Offset); + } + } else { + RI->getOffset(Offset); + Offset += StartAddr; + } + // At a specific address, only keep the first relocation. + if (AddrToReloc.find(Offset) == AddrToReloc.end()) + AddrToReloc[Offset] = *RI; + } + } +} + +bool MCObjectSymbolizer:: +tryAddingSymbolicOperand(MCInst &MI, raw_ostream &cStream, + int64_t Value, uint64_t Address, bool IsBranch, + uint64_t Offset, uint64_t InstSize) { + AddrToRelocMap::iterator RI = AddrToReloc.find(Address + Offset); + if (RI != AddrToReloc.end()) { + if (const MCExpr *RelExpr = RelInfo->createExprForRelocation(RI->second)) { + MI.addOperand(MCOperand::CreateExpr(RelExpr)); + return true; + } + // Only try to create a symbol+offset expression if there is no relocation. + return false; + } + + // Interpret Value as a branch target. + if (IsBranch == false) + return false; + uint64_t UValue = Value; + // FIXME: map instead of looping each time? + error_code ec; + for (symbol_iterator SI = Obj->begin_symbols(), + SE = Obj->end_symbols(); + SI != SE; + SI.increment(ec)) { + if (ec) break; + uint64_t SymAddr; SI->getAddress(SymAddr); + uint64_t SymSize; SI->getSize(SymSize); + StringRef SymName; SI->getName(SymName); + SymbolRef::Type SymType; SI->getType(SymType); + if (SymAddr == UnknownAddressOrSize || SymSize == UnknownAddressOrSize + || SymName.empty() || SymType != SymbolRef::ST_Function) + continue; + + if ( SymAddr == UValue || + (SymAddr <= UValue && SymAddr + SymSize > UValue)) { + MCSymbol *Sym = Ctx.GetOrCreateSymbol(SymName); + const MCExpr *Expr = MCSymbolRefExpr::Create(Sym, Ctx); + if (SymAddr != UValue) { + const MCExpr *Off = MCConstantExpr::Create(UValue - SymAddr, Ctx); + Expr = MCBinaryExpr::CreateAdd(Expr, Off, Ctx); + } + MI.addOperand(MCOperand::CreateExpr(Expr)); + return true; + } + } + return false; +} + +void MCObjectSymbolizer:: +tryAddingPcLoadReferenceComment(raw_ostream &cStream, + int64_t Value, uint64_t Address) { +} + +MCObjectSymbolizer * +MCObjectSymbolizer::createObjectSymbolizer(MCContext &Ctx, + OwningPtr &RelInfo, + const ObjectFile *Obj) { + if (const MachOObjectFile *MachOOF = dyn_cast(Obj)) { + return new MCMachObjectSymbolizer(Ctx, RelInfo, MachOOF); + } + return new MCObjectSymbolizer(Ctx, RelInfo, Obj); +} -- cgit v1.2.3 From 7486d92a6c949a193bb75c0ffa0170eeb2fabb80 Mon Sep 17 00:00:00 2001 From: Rafael Espindola Date: Thu, 30 May 2013 03:05:14 +0000 Subject: Change how we iterate over relocations on ELF. For COFF and MachO, sections semantically have relocations that apply to them. That is not the case on ELF. In relocatable objects (.o), a section with relocations in ELF has offsets to another section where the relocations should be applied. In dynamic objects and executables, relocations don't have an offset, they have a virtual address. The section sh_info may or may not point to another section, but that is not actually used for resolving the relocations. This patch exposes that in the ObjectFile API. It has the following advantages: * Most (all?) clients can handle this more efficiently. They will normally walk all relocations, so doing an effort to iterate in a particular order doesn't save time. * llvm-readobj now prints relocations in the same way the native readelf does. * probably most important, relocations that don't point to any section are now visible. This is the case of relocations in the rela.dyn section. See the updated relocation-executable.test for example. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182908 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/MC/MCObjectSymbolizer.cpp | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'lib/MC/MCObjectSymbolizer.cpp') diff --git a/lib/MC/MCObjectSymbolizer.cpp b/lib/MC/MCObjectSymbolizer.cpp index e1d504ec8e..1803498991 100644 --- a/lib/MC/MCObjectSymbolizer.cpp +++ b/lib/MC/MCObjectSymbolizer.cpp @@ -71,10 +71,14 @@ MCObjectSymbolizer::MCObjectSymbolizer(MCContext &Ctx, SI != SE; SI.increment(ec)) { if (ec) break; - uint64_t StartAddr; SI->getAddress(StartAddr); - uint64_t Size; SI->getSize(Size); - StringRef SecName; SI->getName(SecName); - bool RequiredForExec; SI->isRequiredForExecution(RequiredForExec); + + section_iterator RelSecI = SI->getRelocatedSection(); + if (RelSecI == Obj->end_sections()) + continue; + + uint64_t StartAddr; RelSecI->getAddress(StartAddr); + uint64_t Size; RelSecI->getSize(Size); + bool RequiredForExec; RelSecI->isRequiredForExecution(RequiredForExec); if (RequiredForExec == false || Size == 0) continue; AddrToSection.insert(StartAddr, StartAddr + Size - 1, -- cgit v1.2.3 From b54d29735af9ddabce268c817ccc2ab3eb54d719 Mon Sep 17 00:00:00 2001 From: Ahmed Bougacha Date: Thu, 30 May 2013 18:18:36 +0000 Subject: MCObjectSymbolizer: Switch from IntervalMap to sorted vector, following r182625. This removes the need for the missing SectionRef operator< workaround, and fixes an IntervalMap assert about alignment on MSVC. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182949 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/MC/MCObjectSymbolizer.cpp | 56 +++++++++++++++++++++++++++++++++++-------- 1 file changed, 46 insertions(+), 10 deletions(-) (limited to 'lib/MC/MCObjectSymbolizer.cpp') diff --git a/lib/MC/MCObjectSymbolizer.cpp b/lib/MC/MCObjectSymbolizer.cpp index 1803498991..923c9a0d5f 100644 --- a/lib/MC/MCObjectSymbolizer.cpp +++ b/lib/MC/MCObjectSymbolizer.cpp @@ -17,6 +17,7 @@ #include "llvm/Object/MachO.h" #include "llvm/Object/ELF.h" #include "llvm/Support/raw_ostream.h" +#include using namespace llvm; using namespace object; @@ -40,15 +41,16 @@ public: return; } uint64_t Addr = Value; - AddrToSectionMap::const_iterator SI = AddrToSection.find(Addr); - if (SI.valid()) { - DataRefImpl DRI; DRI.p = *SI; - SectionRef S(DRI, Obj); + SortedSectionList::const_iterator SI = findSectionContaining(Addr); + errs() << " looking for sec " << Addr << "\n"; + if (SI != SortedSections.end()) { + const SectionRef &S = *SI; StringRef Name; S.getName(Name); + uint64_t SAddr; S.getAddress(SAddr); if (Name == "__cstring") { StringRef Contents; S.getContents(Contents); - Contents = Contents.substr(Addr - SI.start()); + Contents = Contents.substr(Addr - SAddr); cStream << " ## literal pool for: " << Contents.substr(0, Contents.find_first_of(0)); } @@ -62,9 +64,7 @@ public: MCObjectSymbolizer::MCObjectSymbolizer(MCContext &Ctx, OwningPtr &RelInfo, const ObjectFile *Obj) - : MCSymbolizer(Ctx, RelInfo), Obj(Obj), - AddrToSectionAllocator(), AddrToSection(AddrToSectionAllocator), - AddrToReloc() { + : MCSymbolizer(Ctx, RelInfo), Obj(Obj), SortedSections(), AddrToReloc() { error_code ec; for (section_iterator SI = Obj->begin_sections(), SE = Obj->end_sections(); @@ -81,8 +81,7 @@ MCObjectSymbolizer::MCObjectSymbolizer(MCContext &Ctx, bool RequiredForExec; RelSecI->isRequiredForExecution(RequiredForExec); if (RequiredForExec == false || Size == 0) continue; - AddrToSection.insert(StartAddr, StartAddr + Size - 1, - SI->getRawDataRefImpl().p); + insertSection(*SI); for (relocation_iterator RI = SI->begin_relocations(), RE = SI->end_relocations(); RI != RE; @@ -177,3 +176,40 @@ MCObjectSymbolizer::createObjectSymbolizer(MCContext &Ctx, } return new MCObjectSymbolizer(Ctx, RelInfo, Obj); } + +// SortedSections implementation. + +static bool SectionStartsBefore(const SectionRef &S, uint64_t Addr) { + uint64_t SAddr; S.getAddress(SAddr); + return SAddr < Addr; +} + +MCObjectSymbolizer::SortedSectionList::const_iterator +MCObjectSymbolizer::findSectionContaining(uint64_t Addr) const { + SortedSectionList::const_iterator + EndIt = SortedSections.end(), + It = std::lower_bound(SortedSections.begin(), EndIt, + Addr, SectionStartsBefore); + if (It == EndIt) + return It; + uint64_t SAddr; It->getAddress(SAddr); + uint64_t SSize; It->getSize(SSize); + if (Addr >= SAddr + SSize) + return EndIt; + return It; +} + +void MCObjectSymbolizer::insertSection(SectionRef Sec) { + uint64_t SAddr; Sec.getAddress(SAddr); + uint64_t SSize; Sec.getSize(SSize); + SortedSectionList::iterator It = std::lower_bound(SortedSections.begin(), + SortedSections.end(), + SAddr, + SectionStartsBefore); + if (It != SortedSections.end()) { + uint64_t FoundSAddr; It->getAddress(FoundSAddr); + if (FoundSAddr < SAddr + SSize) + llvm_unreachable("Inserting overlapping sections"); + } + SortedSections.insert(It, Sec); +} -- cgit v1.2.3 From 6f6ec56b12aff45977b5ea1433739179198d4b81 Mon Sep 17 00:00:00 2001 From: Ahmed Bougacha Date: Thu, 30 May 2013 23:46:47 +0000 Subject: Remove debug print added in r182949. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182973 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/MC/MCObjectSymbolizer.cpp | 1 - 1 file changed, 1 deletion(-) (limited to 'lib/MC/MCObjectSymbolizer.cpp') diff --git a/lib/MC/MCObjectSymbolizer.cpp b/lib/MC/MCObjectSymbolizer.cpp index 923c9a0d5f..aa7648e879 100644 --- a/lib/MC/MCObjectSymbolizer.cpp +++ b/lib/MC/MCObjectSymbolizer.cpp @@ -42,7 +42,6 @@ public: } uint64_t Addr = Value; SortedSectionList::const_iterator SI = findSectionContaining(Addr); - errs() << " looking for sec " << Addr << "\n"; if (SI != SortedSections.end()) { const SectionRef &S = *SI; StringRef Name; S.getName(Name); -- cgit v1.2.3