#include "llvm/ADT/StringMap.h"
#include "llvm/ADT/OwningPtr.h"
+#include "llvm/ExecutionEngine/RuntimeDyld.h"
#include "llvm/Object/MachOObject.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/ManagedStatic.h"
using namespace llvm;
using namespace llvm::object;
-static cl::opt<std::string>
-InputFile(cl::Positional, cl::desc("<input file>"), cl::init("-"));
+static cl::list<std::string>
+InputFileList(cl::Positional, cl::ZeroOrMore,
+ cl::desc("<input file>"));
enum ActionType {
AC_Execute
"Load, link, and execute the inputs."),
clEnumValEnd));
+static cl::opt<std::string>
+EntryPoint("entry",
+ cl::desc("Function to call as entry point."),
+ cl::init("_main"));
+
/* *** */
+// A trivial memory manager that doesn't do anything fancy, just uses the
+// support library allocation routines directly.
+class TrivialMemoryManager : public RTDyldMemoryManager {
+public:
+ SmallVector<sys::MemoryBlock, 16> FunctionMemory;
+ SmallVector<sys::MemoryBlock, 16> DataMemory;
+
+ uint8_t *allocateCodeSection(uintptr_t Size, unsigned Alignment,
+ unsigned SectionID);
+ uint8_t *allocateDataSection(uintptr_t Size, unsigned Alignment,
+ unsigned SectionID);
+
+ uint8_t *startFunctionBody(const char *Name, uintptr_t &Size);
+ void endFunctionBody(const char *Name, uint8_t *FunctionStart,
+ uint8_t *FunctionEnd);
+};
+
+uint8_t *TrivialMemoryManager::allocateCodeSection(uintptr_t Size,
+ unsigned Alignment,
+ unsigned SectionID) {
+ return (uint8_t*)sys::Memory::AllocateRWX(Size, 0, 0).base();
+}
+
+uint8_t *TrivialMemoryManager::allocateDataSection(uintptr_t Size,
+ unsigned Alignment,
+ unsigned SectionID) {
+ return (uint8_t*)sys::Memory::AllocateRWX(Size, 0, 0).base();
+}
+
+uint8_t *TrivialMemoryManager::startFunctionBody(const char *Name,
+ uintptr_t &Size) {
+ return (uint8_t*)sys::Memory::AllocateRWX(Size, 0, 0).base();
+}
+
+void TrivialMemoryManager::endFunctionBody(const char *Name,
+ uint8_t *FunctionStart,
+ uint8_t *FunctionEnd) {
+ uintptr_t Size = FunctionEnd - FunctionStart + 1;
+ FunctionMemory.push_back(sys::MemoryBlock(FunctionStart, Size));
+}
+
static const char *ProgramName;
static void Message(const char *Type, const Twine &Msg) {
/* *** */
-static int ExecuteInput() {
- // Load the input memory buffer.
- OwningPtr<MemoryBuffer> InputBuffer;
- if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFile, InputBuffer))
- return Error("unable to read input: '" + ec.message() + "'");
-
- // Load the Mach-O wrapper object.
- std::string ErrorStr;
- OwningPtr<MachOObject> Obj(
- MachOObject::LoadFromBuffer(InputBuffer.take(), &ErrorStr));
- if (!Obj)
- return Error("unable to load object: '" + ErrorStr + "'");
-
- // Validate that the load commands match what we expect.
- const MachOObject::LoadCommandInfo *SegmentLCI = 0, *SymtabLCI = 0,
- *DysymtabLCI = 0;
- for (unsigned i = 0; i != Obj->getHeader().NumLoadCommands; ++i) {
- const MachOObject::LoadCommandInfo &LCI = Obj->getLoadCommandInfo(i);
- switch (LCI.Command.Type) {
- case macho::LCT_Segment:
- case macho::LCT_Segment64:
- if (SegmentLCI)
- return Error("unexpected input object (multiple segments)");
- SegmentLCI = &LCI;
- break;
- case macho::LCT_Symtab:
- if (SymtabLCI)
- return Error("unexpected input object (multiple symbol tables)");
- SymtabLCI = &LCI;
- break;
- case macho::LCT_Dysymtab:
- if (DysymtabLCI)
- return Error("unexpected input object (multiple symbol tables)");
- DysymtabLCI = &LCI;
- break;
- default:
- return Error("unexpected input object (unexpected load command");
+static int executeInput() {
+ // Instantiate a dynamic linker.
+ TrivialMemoryManager *MemMgr = new TrivialMemoryManager;
+ RuntimeDyld Dyld(MemMgr);
+
+ // If we don't have any input files, read from stdin.
+ if (!InputFileList.size())
+ InputFileList.push_back("-");
+ for(unsigned i = 0, e = InputFileList.size(); i != e; ++i) {
+ // Load the input memory buffer.
+ OwningPtr<MemoryBuffer> InputBuffer;
+ if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFileList[i],
+ InputBuffer))
+ return Error("unable to read input: '" + ec.message() + "'");
+
+ // Load the object file into it.
+ if (Dyld.loadObject(InputBuffer.take())) {
+ return Error(Dyld.getErrorString());
}
}
- if (!SymtabLCI)
- return Error("no symbol table found in object");
- if (!SegmentLCI)
- return Error("no symbol table found in object");
-
- // Read and register the symbol table data.
- InMemoryStruct<macho::SymtabLoadCommand> SymtabLC;
- Obj->ReadSymtabLoadCommand(*SymtabLCI, SymtabLC);
- if (!SymtabLC)
- return Error("unable to load symbol table load command");
- Obj->RegisterStringTable(*SymtabLC);
-
- // Read the dynamic link-edit information, if present (not present in static
- // objects).
- if (DysymtabLCI) {
- InMemoryStruct<macho::DysymtabLoadCommand> DysymtabLC;
- Obj->ReadDysymtabLoadCommand(*DysymtabLCI, DysymtabLC);
- if (!DysymtabLC)
- return Error("unable to load dynamic link-exit load command");
-
- // FIXME: We don't support anything interesting yet.
- if (DysymtabLC->LocalSymbolsIndex != 0)
- return Error("NOT YET IMPLEMENTED: local symbol entries");
- if (DysymtabLC->ExternalSymbolsIndex != 0)
- return Error("NOT YET IMPLEMENTED: non-external symbol entries");
- if (DysymtabLC->UndefinedSymbolsIndex != SymtabLC->NumSymbolTableEntries)
- return Error("NOT YET IMPLEMENTED: undefined symbol entries");
- }
-
- // Load the segment load command.
- if (SegmentLCI->Command.Type != macho::LCT_Segment64)
- return Error("Segment32 not yet implemented!");
- InMemoryStruct<macho::Segment64LoadCommand> Segment64LC;
- Obj->ReadSegment64LoadCommand(*SegmentLCI, Segment64LC);
- if (!Segment64LC)
- return Error("unable to load segment load command");
-
- // Map the segment into memory.
- sys::MemoryBlock Data = sys::Memory::AllocateRWX(Segment64LC->VMSize,
- 0, &ErrorStr);
- if (!Data.base())
- return Error("unable to allocate memory block: '" + ErrorStr + "'");
- memcpy(Data.base(), Obj->getData(Segment64LC->FileOffset,
- Segment64LC->FileSize).data(),
- Segment64LC->FileSize);
- memset((char*)Data.base() + Segment64LC->FileSize, 0,
- Segment64LC->VMSize - Segment64LC->FileSize);
-
- // Bind the section indices to address.
- void **SectionBases = new void*[Segment64LC->NumSections];
- for (unsigned i = 0; i != Segment64LC->NumSections; ++i) {
- InMemoryStruct<macho::Section64> Sect;
- Obj->ReadSection64(*SegmentLCI, i, Sect);
- if (!Sect)
- return Error("unable to load section: '" + Twine(i) + "'");
-
- // FIXME: We don't support relocations yet.
- if (Sect->NumRelocationTableEntries != 0)
- return Error("not yet implemented: relocations!");
-
- // FIXME: Improve check.
- if (Sect->Flags != 0x80000400)
- return Error("unsupported section type!");
-
- SectionBases[i] = (char*) Data.base() + Sect->Address;
- }
-
- // Bind all the symbols to address.
- StringMap<void*> SymbolTable;
- for (unsigned i = 0; i != SymtabLC->NumSymbolTableEntries; ++i) {
- InMemoryStruct<macho::Symbol64TableEntry> STE;
- Obj->ReadSymbol64TableEntry(SymtabLC->SymbolTableOffset, i, STE);
- if (!STE)
- return Error("unable to read symbol: '" + Twine(i) + "'");
- if (STE->SectionIndex == 0)
- return Error("unexpected undefined symbol!");
-
- unsigned Index = STE->SectionIndex - 1;
- if (Index >= Segment64LC->NumSections)
- return Error("invalid section index for symbol: '" + Twine() + "'");
+ // Resolve all the relocations we can.
+ Dyld.resolveRelocations();
- // Get the symbol name.
- StringRef Name = Obj->getStringAtIndex(STE->StringIndex);
+ // FIXME: Error out if there are unresolved relocations.
- // Get the section base address.
- void *SectionBase = SectionBases[Index];
+ // Get the address of the entry point (_main by default).
+ void *MainAddress = Dyld.getSymbolAddress(EntryPoint);
+ if (MainAddress == 0)
+ return Error("no definition for '" + EntryPoint + "'");
- // Get the symbol address.
- void *Address = (char*) SectionBase + STE->Value;
-
- // FIXME: Check the symbol type and flags.
- if (STE->Type != 0xF)
- return Error("unexpected symbol type!");
- if (STE->Flags != 0x0)
- return Error("unexpected symbol type!");
-
- SymbolTable[Name] = Address;
+ // Invalidate the instruction cache for each loaded function.
+ for (unsigned i = 0, e = MemMgr->FunctionMemory.size(); i != e; ++i) {
+ sys::MemoryBlock &Data = MemMgr->FunctionMemory[i];
+ // Make sure the memory is executable.
+ std::string ErrorStr;
+ sys::Memory::InvalidateInstructionCache(Data.base(), Data.size());
+ if (!sys::Memory::setExecutable(Data, &ErrorStr))
+ return Error("unable to mark function executable: '" + ErrorStr + "'");
}
- // Get the address of "_main".
- StringMap<void*>::iterator it = SymbolTable.find("_main");
- if (it == SymbolTable.end())
- return Error("no definition for '_main'");
-
- // Invalidate the instruction cache.
- sys::Memory::InvalidateInstructionCache(Data.base(), Data.size());
-
- // Make sure the memory is executable.
- if (!sys::Memory::setExecutable(Data, &ErrorStr))
- return Error("unable to mark function executable: '" + ErrorStr + "'");
-
// Dispatch to _main().
- void *MainAddress = it->second;
- errs() << "loaded '_main' at: " << MainAddress << "\n";
+ errs() << "loaded '" << EntryPoint << "' at: " << (void*)MainAddress << "\n";
int (*Main)(int, const char**) =
(int(*)(int,const char**)) uintptr_t(MainAddress);
const char **Argv = new const char*[2];
- Argv[0] = InputFile.c_str();
+ // Use the name of the first input object module as argv[0] for the target.
+ Argv[0] = InputFileList[0].c_str();
Argv[1] = 0;
return Main(1, Argv);
}
cl::ParseCommandLineOptions(argc, argv, "llvm MC-JIT tool\n");
switch (Action) {
- default:
case AC_Execute:
- return ExecuteInput();
+ return executeInput();
}
-
- return 0;
}