#define DEBUG_TYPE "jit"
#include "X86JITInfo.h"
#include "X86Relocations.h"
#include "X86Subtarget.h"
#include "X86TargetMachine.h"
#include "llvm/Function.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/System/Valgrind.h"
#include <cstdlib>
#include <cstring>
using namespace llvm;
#if defined (__x86_64__) || defined (_M_AMD64) || defined (_M_X64)
# define X86_64_JIT
#elif defined(__i386__) || defined(i386) || defined(_M_IX86)
# define X86_32_JIT
#endif
void X86JITInfo::replaceMachineCodeForFunction(void *Old, void *New) {
unsigned char *OldByte = (unsigned char *)Old;
*OldByte++ = 0xE9; unsigned *OldWord = (unsigned *)OldByte;
unsigned NewAddr = (intptr_t)New;
unsigned OldAddr = (intptr_t)OldWord;
*OldWord = NewAddr - OldAddr - 4;
sys::ValgrindDiscardTranslations(Old, 5);
}
static TargetJITInfo::JITCompilerFn JITCompilerFunction;
#ifndef __USER_LABEL_PREFIX__
#define __USER_LABEL_PREFIX__
#endif
#define GETASMPREFIX2(X) #X
#define GETASMPREFIX(X) GETASMPREFIX2(X)
#define ASMPREFIX GETASMPREFIX(__USER_LABEL_PREFIX__)
#if defined(__ELF__)
# define SIZE(sym) ".size " #sym ", . - " #sym "\n"
# define TYPE_FUNCTION(sym) ".type " #sym ", @function\n"
#else
# define SIZE(sym)
# define TYPE_FUNCTION(sym)
#endif
#if defined(__APPLE__)
# define CFI(x)
#else
# define CFI(x)
#endif
extern "C" {
#if defined(X86_64_JIT)
# ifndef _MSC_VER
void X86CompilationCallback(void);
asm(
".text\n"
".align 8\n"
".globl " ASMPREFIX "X86CompilationCallback\n"
TYPE_FUNCTION(X86CompilationCallback)
ASMPREFIX "X86CompilationCallback:\n"
CFI(".cfi_startproc\n")
"pushq %rbp\n"
CFI(".cfi_def_cfa_offset 16\n")
CFI(".cfi_offset %rbp, -16\n")
"movq %rsp, %rbp\n"
CFI(".cfi_def_cfa_register %rbp\n")
"pushq %rdi\n"
CFI(".cfi_rel_offset %rdi, 0\n")
"pushq %rsi\n"
CFI(".cfi_rel_offset %rsi, 8\n")
"pushq %rdx\n"
CFI(".cfi_rel_offset %rdx, 16\n")
"pushq %rcx\n"
CFI(".cfi_rel_offset %rcx, 24\n")
"pushq %r8\n"
CFI(".cfi_rel_offset %r8, 32\n")
"pushq %r9\n"
CFI(".cfi_rel_offset %r9, 40\n")
"andq $-16, %rsp\n"
"subq $128, %rsp\n"
"movaps %xmm0, (%rsp)\n"
"movaps %xmm1, 16(%rsp)\n"
"movaps %xmm2, 32(%rsp)\n"
"movaps %xmm3, 48(%rsp)\n"
"movaps %xmm4, 64(%rsp)\n"
"movaps %xmm5, 80(%rsp)\n"
"movaps %xmm6, 96(%rsp)\n"
"movaps %xmm7, 112(%rsp)\n"
"movq %rbp, %rdi\n" "movq 8(%rbp), %rsi\n"
"call " ASMPREFIX "X86CompilationCallback2\n"
"movaps 112(%rsp), %xmm7\n"
"movaps 96(%rsp), %xmm6\n"
"movaps 80(%rsp), %xmm5\n"
"movaps 64(%rsp), %xmm4\n"
"movaps 48(%rsp), %xmm3\n"
"movaps 32(%rsp), %xmm2\n"
"movaps 16(%rsp), %xmm1\n"
"movaps (%rsp), %xmm0\n"
"movq %rbp, %rsp\n"
CFI(".cfi_def_cfa_register %rsp\n")
"subq $48, %rsp\n"
CFI(".cfi_adjust_cfa_offset 48\n")
"popq %r9\n"
CFI(".cfi_adjust_cfa_offset -8\n")
CFI(".cfi_restore %r9\n")
"popq %r8\n"
CFI(".cfi_adjust_cfa_offset -8\n")
CFI(".cfi_restore %r8\n")
"popq %rcx\n"
CFI(".cfi_adjust_cfa_offset -8\n")
CFI(".cfi_restore %rcx\n")
"popq %rdx\n"
CFI(".cfi_adjust_cfa_offset -8\n")
CFI(".cfi_restore %rdx\n")
"popq %rsi\n"
CFI(".cfi_adjust_cfa_offset -8\n")
CFI(".cfi_restore %rsi\n")
"popq %rdi\n"
CFI(".cfi_adjust_cfa_offset -8\n")
CFI(".cfi_restore %rdi\n")
"popq %rbp\n"
CFI(".cfi_adjust_cfa_offset -8\n")
CFI(".cfi_restore %rbp\n")
"ret\n"
CFI(".cfi_endproc\n")
SIZE(X86CompilationCallback)
);
# else
void X86CompilationCallback();
# endif
#elif defined (X86_32_JIT)
# ifndef _MSC_VER
void X86CompilationCallback(void);
asm(
".text\n"
".align 8\n"
".globl " ASMPREFIX "X86CompilationCallback\n"
TYPE_FUNCTION(X86CompilationCallback)
ASMPREFIX "X86CompilationCallback:\n"
CFI(".cfi_startproc\n")
"pushl %ebp\n"
CFI(".cfi_def_cfa_offset 8\n")
CFI(".cfi_offset %ebp, -8\n")
"movl %esp, %ebp\n" CFI(".cfi_def_cfa_register %ebp\n")
"pushl %eax\n"
CFI(".cfi_rel_offset %eax, 0\n")
"pushl %edx\n" CFI(".cfi_rel_offset %edx, 4\n")
"pushl %ecx\n"
CFI(".cfi_rel_offset %ecx, 8\n")
# if defined(__APPLE__)
"andl $-16, %esp\n" # endif
"subl $16, %esp\n"
"movl 4(%ebp), %eax\n" "movl %eax, 4(%esp)\n"
"movl %ebp, (%esp)\n"
"call " ASMPREFIX "X86CompilationCallback2\n"
"movl %ebp, %esp\n" CFI(".cfi_def_cfa_register %esp\n")
"subl $12, %esp\n"
CFI(".cfi_adjust_cfa_offset 12\n")
"popl %ecx\n"
CFI(".cfi_adjust_cfa_offset -4\n")
CFI(".cfi_restore %ecx\n")
"popl %edx\n"
CFI(".cfi_adjust_cfa_offset -4\n")
CFI(".cfi_restore %edx\n")
"popl %eax\n"
CFI(".cfi_adjust_cfa_offset -4\n")
CFI(".cfi_restore %eax\n")
"popl %ebp\n"
CFI(".cfi_adjust_cfa_offset -4\n")
CFI(".cfi_restore %ebp\n")
"ret\n"
CFI(".cfi_endproc\n")
SIZE(X86CompilationCallback)
);
void X86CompilationCallback_SSE(void);
asm(
".text\n"
".align 8\n"
".globl " ASMPREFIX "X86CompilationCallback_SSE\n"
TYPE_FUNCTION(X86CompilationCallback_SSE)
ASMPREFIX "X86CompilationCallback_SSE:\n"
CFI(".cfi_startproc\n")
"pushl %ebp\n"
CFI(".cfi_def_cfa_offset 8\n")
CFI(".cfi_offset %ebp, -8\n")
"movl %esp, %ebp\n" CFI(".cfi_def_cfa_register %ebp\n")
"pushl %eax\n"
CFI(".cfi_rel_offset %eax, 0\n")
"pushl %edx\n" CFI(".cfi_rel_offset %edx, 4\n")
"pushl %ecx\n"
CFI(".cfi_rel_offset %ecx, 8\n")
"andl $-16, %esp\n" "subl $64, %esp\n"
"movaps %xmm0, (%esp)\n"
"movaps %xmm1, 16(%esp)\n"
"movaps %xmm2, 32(%esp)\n"
"movaps %xmm3, 48(%esp)\n"
"subl $16, %esp\n"
"movl 4(%ebp), %eax\n" "movl %eax, 4(%esp)\n"
"movl %ebp, (%esp)\n"
"call " ASMPREFIX "X86CompilationCallback2\n"
"addl $16, %esp\n"
"movaps 48(%esp), %xmm3\n"
CFI(".cfi_restore %xmm3\n")
"movaps 32(%esp), %xmm2\n"
CFI(".cfi_restore %xmm2\n")
"movaps 16(%esp), %xmm1\n"
CFI(".cfi_restore %xmm1\n")
"movaps (%esp), %xmm0\n"
CFI(".cfi_restore %xmm0\n")
"movl %ebp, %esp\n" CFI(".cfi_def_cfa_register esp\n")
"subl $12, %esp\n"
CFI(".cfi_adjust_cfa_offset 12\n")
"popl %ecx\n"
CFI(".cfi_adjust_cfa_offset -4\n")
CFI(".cfi_restore %ecx\n")
"popl %edx\n"
CFI(".cfi_adjust_cfa_offset -4\n")
CFI(".cfi_restore %edx\n")
"popl %eax\n"
CFI(".cfi_adjust_cfa_offset -4\n")
CFI(".cfi_restore %eax\n")
"popl %ebp\n"
CFI(".cfi_adjust_cfa_offset -4\n")
CFI(".cfi_restore %ebp\n")
"ret\n"
CFI(".cfi_endproc\n")
SIZE(X86CompilationCallback_SSE)
);
# else
void X86CompilationCallback2(intptr_t *StackPtr, intptr_t RetAddr);
_declspec(naked) void X86CompilationCallback(void) {
__asm {
push ebp
mov ebp, esp
push eax
push edx
push ecx
and esp, -16
sub esp, 16
mov eax, dword ptr [ebp+4]
mov dword ptr [esp+4], eax
mov dword ptr [esp], ebp
call X86CompilationCallback2
mov esp, ebp
sub esp, 12
pop ecx
pop edx
pop eax
pop ebp
ret
}
}
# endif // _MSC_VER
#else // Not an i386 host
void X86CompilationCallback() {
llvm_unreachable("Cannot call X86CompilationCallback() on a non-x86 arch!");
}
#endif
}
extern "C" {
#if !(defined (X86_64_JIT) && defined(_MSC_VER))
static
#endif
void ATTRIBUTE_USED
X86CompilationCallback2(intptr_t *StackPtr, intptr_t RetAddr) {
intptr_t *RetAddrLoc = &StackPtr[1];
assert(*RetAddrLoc == RetAddr &&
"Could not find return address on the stack!");
bool isStub = ((unsigned char*)RetAddr)[0] == 0xCE;
#if defined (X86_64_JIT)
RetAddr--; #else
RetAddr -= 4; #endif
#if 0
DEBUG(dbgs() << "In callback! Addr=" << (void*)RetAddr
<< " ESP=" << (void*)StackPtr
<< ": Resolving call to function: "
<< TheVM->getFunctionReferencedName((void*)RetAddr) << "\n");
#endif
#if defined (X86_64_JIT)
assert(((unsigned char*)RetAddr)[-2] == 0x41 &&"Not a call instr!");
assert(((unsigned char*)RetAddr)[-1] == 0xFF &&"Not a call instr!");
#else
assert(((unsigned char*)RetAddr)[-1] == 0xE8 &&"Not a call instr!");
#endif
intptr_t NewVal = (intptr_t)JITCompilerFunction((void*)RetAddr);
#if defined (X86_64_JIT)
assert(isStub &&
"X86-64 doesn't support rewriting non-stub lazy compilation calls:"
" the call instruction varies too much.");
#else
*(intptr_t *)RetAddr = (intptr_t)(NewVal-RetAddr-4);
#endif
if (isStub) {
#if defined (X86_64_JIT)
intptr_t diff = NewVal-RetAddr+7;
if (diff >= -2147483648LL && diff <= 2147483647LL) {
*(unsigned char*)(RetAddr-0xc) = 0xE9;
*(intptr_t *)(RetAddr-0xb) = diff & 0xffffffff;
} else {
*(intptr_t *)(RetAddr - 0xa) = NewVal;
((unsigned char*)RetAddr)[0] = (2 | (4 << 3) | (3 << 6));
}
sys::ValgrindDiscardTranslations((void*)(RetAddr-0xc), 0xd);
#else
((unsigned char*)RetAddr)[-1] = 0xE9;
sys::ValgrindDiscardTranslations((void*)(RetAddr-1), 5);
#endif
}
#if defined (X86_64_JIT)
*RetAddrLoc -= 0xd;
#else
*RetAddrLoc -= 5;
#endif
}
}
TargetJITInfo::LazyResolverFn
X86JITInfo::getLazyResolverFunction(JITCompilerFn F) {
JITCompilerFunction = F;
#if defined (X86_32_JIT) && !defined (_MSC_VER)
if (Subtarget->hasSSE1())
return X86CompilationCallback_SSE;
#endif
return X86CompilationCallback;
}
X86JITInfo::X86JITInfo(X86TargetMachine &tm) : TM(tm) {
Subtarget = &TM.getSubtarget<X86Subtarget>();
useGOT = 0;
TLSOffset = 0;
}
void *X86JITInfo::emitGlobalValueIndirectSym(const GlobalValue* GV, void *ptr,
JITCodeEmitter &JCE) {
#if defined (X86_64_JIT)
const unsigned Alignment = 8;
uint8_t Buffer[8];
uint8_t *Cur = Buffer;
MachineCodeEmitter::emitWordLEInto(Cur, (unsigned)(intptr_t)ptr);
MachineCodeEmitter::emitWordLEInto(Cur, (unsigned)(((intptr_t)ptr) >> 32));
#else
const unsigned Alignment = 4;
uint8_t Buffer[4];
uint8_t *Cur = Buffer;
MachineCodeEmitter::emitWordLEInto(Cur, (intptr_t)ptr);
#endif
return JCE.allocIndirectGV(GV, Buffer, sizeof(Buffer), Alignment);
}
TargetJITInfo::StubLayout X86JITInfo::getStubLayout() {
StubLayout Result = {14, 4};
return Result;
}
void *X86JITInfo::emitFunctionStub(const Function* F, void *Target,
JITCodeEmitter &JCE) {
#if defined (X86_32_JIT) && !defined (_MSC_VER)
bool NotCC = (Target != (void*)(intptr_t)X86CompilationCallback &&
Target != (void*)(intptr_t)X86CompilationCallback_SSE);
#else
bool NotCC = Target != (void*)(intptr_t)X86CompilationCallback;
#endif
JCE.emitAlignment(4);
void *Result = (void*)JCE.getCurrentPCValue();
if (NotCC) {
#if defined (X86_64_JIT)
JCE.emitByte(0x49); JCE.emitByte(0xB8+2); JCE.emitWordLE((unsigned)(intptr_t)Target);
JCE.emitWordLE((unsigned)(((intptr_t)Target) >> 32));
JCE.emitByte(0x41); JCE.emitByte(0xFF); JCE.emitByte(2 | (4 << 3) | (3 << 6));
#else
JCE.emitByte(0xE9);
JCE.emitWordLE((intptr_t)Target-JCE.getCurrentPCValue()-4);
#endif
return Result;
}
#if defined (X86_64_JIT)
JCE.emitByte(0x49); JCE.emitByte(0xB8+2); JCE.emitWordLE((unsigned)(intptr_t)Target);
JCE.emitWordLE((unsigned)(((intptr_t)Target) >> 32));
JCE.emitByte(0x41); JCE.emitByte(0xFF); JCE.emitByte(2 | (2 << 3) | (3 << 6));
#else
JCE.emitByte(0xE8);
JCE.emitWordLE((intptr_t)Target-JCE.getCurrentPCValue()-4);
#endif
JCE.emitByte(0xCE); return Result;
}
uintptr_t X86JITInfo::getPICJumpTableEntry(uintptr_t BB, uintptr_t Entry) {
#if defined(X86_64_JIT)
return BB - Entry;
#else
return BB - PICBase;
#endif
}
void X86JITInfo::relocate(void *Function, MachineRelocation *MR,
unsigned NumRelocs, unsigned char* GOTBase) {
for (unsigned i = 0; i != NumRelocs; ++i, ++MR) {
void *RelocPos = (char*)Function + MR->getMachineCodeOffset();
intptr_t ResultPtr = (intptr_t)MR->getResultPointer();
switch ((X86::RelocationType)MR->getRelocationType()) {
case X86::reloc_pcrel_word: {
ResultPtr = ResultPtr -(intptr_t)RelocPos - 4 - MR->getConstantVal();
*((unsigned*)RelocPos) += (unsigned)ResultPtr;
break;
}
case X86::reloc_picrel_word: {
ResultPtr = ResultPtr - ((intptr_t)Function + MR->getConstantVal());
*((unsigned*)RelocPos) += (unsigned)ResultPtr;
break;
}
case X86::reloc_absolute_word:
case X86::reloc_absolute_word_sext:
*((unsigned*)RelocPos) += (unsigned)ResultPtr;
break;
case X86::reloc_absolute_dword:
*((intptr_t*)RelocPos) += ResultPtr;
break;
}
}
}
char* X86JITInfo::allocateThreadLocalMemory(size_t size) {
#if defined(X86_32_JIT) && !defined(__APPLE__) && !defined(_MSC_VER)
TLSOffset -= size;
return TLSOffset;
#else
llvm_unreachable("Cannot allocate thread local storage on this arch!");
return 0;
#endif
}