X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>();
uint64_t MaxAlign = calculateMaxStackAlign(MF); // Desired stack alignment.
uint64_t StackSize = MFI->getStackSize(); // Number of bytes to allocate.
+ bool IsFunclet = MBB.isEHFuncletEntry();
bool HasFP = hasFP(MF);
bool IsWin64CC = STI.isCallingConvWin64(Fn->getCallingConv());
bool IsWin64Prologue = MF.getTarget().getMCAsmInfo()->usesWindowsCFI();
uint64_t NumBytes = 0;
int stackGrowth = -SlotSize;
- if (MBB.isEHFuncletEntry()) {
- assert(STI.isOSWindows() && "funclets only supported on Windows");
-
- // Set up the FramePtr and BasePtr physical registers using the address
- // passed as EBP or RDX by the MSVC EH runtime.
- if (STI.is32Bit()) {
- // PUSH32r %ebp
- BuildMI(MBB, MBBI, DL, TII.get(X86::PUSH32r))
- .addReg(MachineFramePtr, RegState::Kill)
- .setMIFlag(MachineInstr::FrameSetup);
- // Reset EBP / ESI to something good.
- MBBI = restoreWin32EHFrameAndBasePtr(MBB, MBBI, DL);
- } else {
- // FIXME: Add SEH directives.
- NeedsWinCFI = false;
- // Immediately spill RDX into the home slot. The runtime cares about this.
- unsigned RDX = Uses64BitFramePtr ? X86::RDX : X86::EDX;
- // MOV64mr %rdx, 16(%rsp)
- unsigned MOVmr = Uses64BitFramePtr ? X86::MOV64mr : X86::MOV32mr;
- addRegOffset(BuildMI(MBB, MBBI, DL, TII.get(MOVmr)),
- StackPtr, true, 16)
+ unsigned RDX = Uses64BitFramePtr ? X86::RDX : X86::EDX;
+ if (IsWin64Prologue && IsFunclet) {
+ // Immediately spill RDX into the home slot. The runtime cares about this.
+ // MOV64mr %rdx, 16(%rsp)
+ unsigned MOVmr = Uses64BitFramePtr ? X86::MOV64mr : X86::MOV32mr;
+ addRegOffset(BuildMI(MBB, MBBI, DL, TII.get(MOVmr)), StackPtr, true, 16)
.addReg(RDX)
.setMIFlag(MachineInstr::FrameSetup);
- // PUSH64r %rbp
- BuildMI(MBB, MBBI, DL, TII.get(X86::PUSH64r))
- .addReg(MachineFramePtr, RegState::Kill)
- .setMIFlag(MachineInstr::FrameSetup);
- // MOV64rr %rdx, %rbp
- unsigned MOVrr = Uses64BitFramePtr ? X86::MOV64rr : X86::MOV32rr;
- BuildMI(MBB, MBBI, DL, TII.get(MOVrr), FramePtr)
- .addReg(RDX)
- .setMIFlag(MachineInstr::FrameSetup);
- assert(!TRI->hasBasePointer(MF) &&
- "x64 funclets with base ptrs not yet implemented");
- }
+ }
- // For EH funclets, only allocate enough space for outgoing calls.
- NumBytes = MFI->getMaxCallFrameSize();
- } else if (HasFP) {
+ if (HasFP) {
// Calculate required stack adjustment.
uint64_t FrameSize = StackSize - SlotSize;
// If required, include space for extra hidden slot for stashing base pointer.
// Get the offset of the stack slot for the EBP register, which is
// guaranteed to be the last slot by processFunctionBeforeFrameFinalized.
// Update the frame offset adjustment.
- MFI->setOffsetAdjustment(-NumBytes);
+ if (!IsFunclet)
+ MFI->setOffsetAdjustment(-NumBytes);
+ else
+ assert(MFI->getOffsetAdjustment() == -(int)NumBytes &&
+ "should calculate same local variable offset for funclets");
// Save EBP/RBP into the appropriate stack slot.
BuildMI(MBB, MBBI, DL, TII.get(Is64Bit ? X86::PUSH64r : X86::PUSH32r))
.setMIFlag(MachineInstr::FrameSetup);
}
- if (!IsWin64Prologue) {
+ if (!IsWin64Prologue && !IsFunclet) {
// Update EBP with the new base value.
BuildMI(MBB, MBBI, DL,
TII.get(Uses64BitFramePtr ? X86::MOV64rr : X86::MOV32rr),
FramePtr)
.addReg(StackPtr)
.setMIFlag(MachineInstr::FrameSetup);
- }
- if (NeedsDwarfCFI) {
- // Mark effective beginning of when frame pointer becomes valid.
- // Define the current CFA to use the EBP/RBP register.
- unsigned DwarfFramePtr = TRI->getDwarfRegNum(MachineFramePtr, true);
- BuildCFI(MBB, MBBI, DL,
- MCCFIInstruction::createDefCfaRegister(nullptr, DwarfFramePtr));
+ if (NeedsDwarfCFI) {
+ // Mark effective beginning of when frame pointer becomes valid.
+ // Define the current CFA to use the EBP/RBP register.
+ unsigned DwarfFramePtr = TRI->getDwarfRegNum(MachineFramePtr, true);
+ BuildCFI(MBB, MBBI, DL, MCCFIInstruction::createDefCfaRegister(
+ nullptr, DwarfFramePtr));
+ }
}
// Mark the FramePtr as live-in in every block.
for (MachineFunction::iterator I = MF.begin(), E = MF.end(); I != E; ++I)
I->addLiveIn(MachineFramePtr);
} else {
+ assert(!IsFunclet && "funclets without FPs not yet implemented");
NumBytes = StackSize - X86FI->getCalleeSavedFrameSize();
}
+ // For EH funclets, only allocate enough space for outgoing calls. Save the
+ // NumBytes value that we would've used for the parent frame.
+ unsigned ParentFrameNumBytes = NumBytes;
+ if (IsFunclet)
+ NumBytes = MFI->getMaxCallFrameSize();
+
// Skip the callee-saved push instructions.
bool PushedRegs = false;
int StackOffset = 2 * stackGrowth;
// Realign stack after we pushed callee-saved registers (so that we'll be
// able to calculate their offsets from the frame pointer).
// Don't do this for Win64, it needs to realign the stack after the prologue.
- if (!IsWin64Prologue && TRI->needsStackRealignment(MF)) {
+ if (!IsWin64Prologue && !IsFunclet && TRI->needsStackRealignment(MF)) {
assert(HasFP && "There should be a frame pointer if stack is realigned.");
BuildStackAlignAND(MBB, MBBI, DL, MaxAlign);
}
// increments is necessary to ensure that the guard pages used by the OS
// virtual memory manager are allocated in correct sequence.
uint64_t AlignedNumBytes = NumBytes;
- if (IsWin64Prologue && TRI->needsStackRealignment(MF))
+ if (IsWin64Prologue && !IsFunclet && TRI->needsStackRealignment(MF))
AlignedNumBytes = RoundUpToAlignment(AlignedNumBytes, MaxAlign);
if (AlignedNumBytes >= StackProbeSize && UseStackProbe) {
// Check whether EAX is livein for this function.
int SEHFrameOffset = 0;
if (IsWin64Prologue && HasFP) {
- SEHFrameOffset = calculateSetFPREG(NumBytes);
+ // Set RBP to a small fixed offset from RSP. In the funclet case, we base
+ // this calculation on the incoming RDX, which holds the value of RSP from
+ // the parent frame at the end of the prologue.
+ unsigned SPOrRDX = !IsFunclet ? StackPtr : RDX;
+ SEHFrameOffset = calculateSetFPREG(ParentFrameNumBytes);
if (SEHFrameOffset)
addRegOffset(BuildMI(MBB, MBBI, DL, TII.get(X86::LEA64r), FramePtr),
- StackPtr, false, SEHFrameOffset);
+ SPOrRDX, false, SEHFrameOffset);
else
- BuildMI(MBB, MBBI, DL, TII.get(X86::MOV64rr), FramePtr).addReg(StackPtr);
+ BuildMI(MBB, MBBI, DL, TII.get(X86::MOV64rr), FramePtr)
+ .addReg(SPOrRDX);
- if (NeedsWinCFI)
+ // If this is not a funclet, emit the CFI describing our frame pointer.
+ if (NeedsWinCFI && !IsFunclet)
BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_SetFrame))
.addImm(FramePtr)
.addImm(SEHFrameOffset)
.setMIFlag(MachineInstr::FrameSetup);
+ } else if (IsFunclet && STI.is32Bit()) {
+ // Reset EBP / ESI to something good for funclets.
+ MBBI = restoreWin32EHStackPointers(MBB, MBBI, DL);
}
while (MBBI != MBB.end() && MBBI->getFlag(MachineInstr::FrameSetup)) {
static bool isFuncletReturnInstr(MachineInstr *MI) {
switch (MI->getOpcode()) {
case X86::CATCHRET:
- case X86::CATCHRET64:
case X86::CLEANUPRET:
- case X86::CLEANUPRET64:
return true;
default:
return false;
bool IsWin64Prologue = MF.getTarget().getMCAsmInfo()->usesWindowsCFI();
bool NeedsWinCFI =
IsWin64Prologue && MF.getFunction()->needsUnwindTableEntry();
+ bool IsFunclet = isFuncletReturnInstr(MBBI);
+ MachineBasicBlock *RestoreMBB = nullptr;
// Get the number of bytes to allocate from the FrameInfo.
uint64_t StackSize = MFI->getStackSize();
unsigned CSSize = X86FI->getCalleeSavedFrameSize();
uint64_t NumBytes = 0;
- if (isFuncletReturnInstr(MBBI)) {
+ if (MBBI->getOpcode() == X86::CATCHRET) {
NumBytes = MFI->getMaxCallFrameSize();
- assert(hasFP(MF) && "win64 EH funclets without FP not yet implemented");
+ assert(hasFP(MF) && "EH funclets without FP not yet implemented");
+ MachineBasicBlock *TargetMBB = MBBI->getOperand(0).getMBB();
+
+ // If this is SEH, this isn't really a funclet return.
+ bool IsSEH = isAsynchronousEHPersonality(
+ classifyEHPersonality(MF.getFunction()->getPersonalityFn()));
+ if (IsSEH) {
+ if (STI.is32Bit())
+ restoreWin32EHStackPointers(MBB, MBBI, DL, /*RestoreSP=*/true);
+ BuildMI(MBB, MBBI, DL, TII.get(X86::JMP_4)).addMBB(TargetMBB);
+ MBBI->eraseFromParent();
+ return;
+ }
+
+ // For 32-bit, create a new block for the restore code.
+ RestoreMBB = TargetMBB;
+ if (STI.is32Bit()) {
+ RestoreMBB = MF.CreateMachineBasicBlock(MBB.getBasicBlock());
+ MF.insert(TargetMBB, RestoreMBB);
+ MBB.removeSuccessor(TargetMBB);
+ MBB.addSuccessor(RestoreMBB);
+ RestoreMBB->addSuccessor(TargetMBB);
+ MBBI->getOperand(0).setMBB(RestoreMBB);
+ }
// Pop EBP.
BuildMI(MBB, MBBI, DL, TII.get(Is64Bit ? X86::POP64r : X86::POP32r),
- MachineFramePtr).setMIFlag(MachineInstr::FrameDestroy);
+ MachineFramePtr)
+ .setMIFlag(MachineInstr::FrameDestroy);
+
+ // Insert frame restoration code in a new block.
+ if (STI.is32Bit()) {
+ auto RestoreMBBI = RestoreMBB->begin();
+ restoreWin32EHStackPointers(*RestoreMBB, RestoreMBBI, DL,
+ /*RestoreSP=*/true);
+ BuildMI(*RestoreMBB, RestoreMBBI, DL, TII.get(X86::JMP_4))
+ .addMBB(TargetMBB);
+ }
+ } else if (MBBI->getOpcode() == X86::CLEANUPRET) {
+ NumBytes = MFI->getMaxCallFrameSize();
+ assert(hasFP(MF) && "EH funclets without FP not yet implemented");
+ BuildMI(MBB, MBBI, DL, TII.get(Is64Bit ? X86::POP64r : X86::POP32r),
+ MachineFramePtr)
+ .setMIFlag(MachineInstr::FrameDestroy);
} else if (hasFP(MF)) {
// Calculate required stack adjustment.
uint64_t FrameSize = StackSize - SlotSize;
}
MachineBasicBlock::iterator FirstCSPop = MBBI;
+ if (RestoreMBB) {
+ // Fill EAX/RAX with the address of the target block.
+ unsigned ReturnReg = STI.is64Bit() ? X86::RAX : X86::EAX;
+ if (STI.is64Bit()) {
+ // LEA64r RestoreMBB(%rip), %rax
+ BuildMI(MBB, FirstCSPop, DL, TII.get(X86::LEA64r), ReturnReg)
+ .addReg(X86::RIP)
+ .addImm(0)
+ .addReg(0)
+ .addMBB(RestoreMBB)
+ .addReg(0);
+ } else {
+ // MOV32ri $RestoreMBB, %eax
+ BuildMI(MBB, FirstCSPop, DL, TII.get(X86::MOV32ri))
+ .addReg(ReturnReg)
+ .addMBB(RestoreMBB);
+ }
+ }
+
if (MBBI != MBB.end())
DL = MBBI->getDebugLoc();
// If dynamic alloca is used, then reset esp to point to the last callee-saved
// slot before popping them off! Same applies for the case, when stack was
- // realigned.
- if (TRI->needsStackRealignment(MF) || MFI->hasVarSizedObjects()) {
+ // realigned. Don't do this if this was a funclet epilogue, since the funclets
+ // will not do realignment or dynamic stack allocation.
+ if ((TRI->needsStackRealignment(MF) || MFI->hasVarSizedObjects()) &&
+ !IsFunclet) {
if (TRI->needsStackRealignment(MF))
MBBI = FirstCSPop;
unsigned SEHFrameOffset = calculateSetFPREG(SEHStackAllocAmt);
if (CSI.empty())
return false;
- // Don't restore CSRs in 32-bit EH funclets. Matches
- // spillCalleeSavedRegisters.
- if (isFuncletReturnInstr(MI) && STI.is32Bit() && STI.isOSWindows())
- return true;
+ if (isFuncletReturnInstr(MI) && STI.isOSWindows()) {
+ // Don't restore CSRs in 32-bit EH funclets. Matches
+ // spillCalleeSavedRegisters.
+ if (STI.is32Bit())
+ return true;
+ // Don't restore CSRs before an SEH catchret. SEH except blocks do not form
+ // funclets. emitEpilogue transforms these to normal jumps.
+ if (MI->getOpcode() == X86::CATCHRET) {
+ const Function *Func = MBB.getParent()->getFunction();
+ bool IsSEH = isAsynchronousEHPersonality(
+ classifyEHPersonality(Func->getPersonalityFn()));
+ if (IsSEH)
+ return true;
+ }
+ }
DebugLoc DL = MBB.findDebugLoc(MI);
bool X86FrameLowering::adjustStackWithPops(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MBBI, DebugLoc DL, int Offset) const {
+ if (Offset <= 0)
+ return false;
+
if (Offset % SlotSize)
return false;
unsigned FoundRegs = 0;
auto RegMask = Prev->getOperand(1);
-
- // Try to find up to NumPops free registers.
- for (auto Candidate : X86::GR32_NOREX_NOSPRegClass) {
+
+ auto &RegClass =
+ Is64Bit ? X86::GR64_NOREX_NOSPRegClass : X86::GR32_NOREX_NOSPRegClass;
+ // Try to find up to NumPops free registers.
+ for (auto Candidate : RegClass) {
// Poor man's liveness:
// Since we're immediately after a call, any register that is clobbered
// If the stack pointer can be changed after prologue, turn the
// adjcallstackup instruction into a 'sub ESP, <amt>' and the
// adjcallstackdown instruction into 'add ESP, <amt>'
- if (Amount == 0)
- return;
// We need to keep the stack aligned properly. To do this, we round the
// amount of space needed for the outgoing arguments up to the next
unsigned StackAlign = getStackAlignment();
Amount = RoundUpToAlignment(Amount, StackAlign);
+ // If we have any exception handlers in this function, and we adjust
+ // the SP before calls, we may need to indicate this to the unwinder,
+ // using GNU_ARGS_SIZE. Note that this may be necessary
+ // even when Amount == 0, because the preceding function may have
+ // set a non-0 GNU_ARGS_SIZE.
+ // TODO: We don't need to reset this between subsequent functions,
+ // if it didn't change.
+ bool HasDwarfEHHandlers =
+ !MF.getTarget().getMCAsmInfo()->usesWindowsCFI() &&
+ !MF.getMMI().getLandingPads().empty();
+
+ if (HasDwarfEHHandlers && !isDestroy &&
+ MF.getInfo<X86MachineFunctionInfo>()->getHasPushSequences())
+ BuildCFI(MBB, I, DL,\r
+ MCCFIInstruction::createGnuArgsSize(nullptr, Amount));\r
+
+ if (Amount == 0)
+ return;
+
// Factor out the amount that gets handled inside the sequence
// (Pushes of argument for frame setup, callee pops for frame destroy)
Amount -= InternalAmt;
return !terminatorsNeedFlagsAsInput(MBB);
}
-MachineBasicBlock::iterator X86FrameLowering::restoreWin32EHFrameAndBasePtr(
+MachineBasicBlock::iterator X86FrameLowering::restoreWin32EHStackPointers(
MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI,
- DebugLoc DL) const {
+ DebugLoc DL, bool RestoreSP) const {
assert(STI.isTargetWindowsMSVC() && "funclets only supported in MSVC env");
assert(STI.isTargetWin32() && "EBP/ESI restoration only required on win32");
assert(STI.is32Bit() && !Uses64BitFramePtr &&
// FIXME: Don't set FrameSetup flag in catchret case.
int FI = FuncInfo.EHRegNodeFrameIndex;
+ int EHRegSize = MFI->getObjectSize(FI);
+
+ if (RestoreSP) {
+ // MOV32rm -EHRegSize(%ebp), %esp
+ addRegOffset(BuildMI(MBB, MBBI, DL, TII.get(X86::MOV32rm), X86::ESP),
+ X86::EBP, true, -EHRegSize)
+ .setMIFlag(MachineInstr::FrameSetup);
+ }
+
unsigned UsedReg;
int EHRegOffset = getFrameIndexReference(MF, FI, UsedReg);
- int EHRegSize = MFI->getObjectSize(FI);
int EndOffset = -EHRegOffset - EHRegSize;
- assert(EndOffset >= 0 &&
- "end of registration object above normal EBP position!");
+ FuncInfo.EHRegNodeEndOffset = EndOffset;
+
if (UsedReg == FramePtr) {
// ADD $offset, %ebp
- assert(UsedReg == FramePtr);
unsigned ADDri = getADDriOpcode(false, EndOffset);
BuildMI(MBB, MBBI, DL, TII.get(ADDri), FramePtr)
.addReg(FramePtr)
.setMIFlag(MachineInstr::FrameSetup)
->getOperand(3)
.setIsDead();
- } else {
- assert(UsedReg == BasePtr);
+ assert(EndOffset >= 0 &&
+ "end of registration object above normal EBP position!");
+ } else if (UsedReg == BasePtr) {
// LEA offset(%ebp), %esi
addRegOffset(BuildMI(MBB, MBBI, DL, TII.get(X86::LEA32r), BasePtr),
FramePtr, false, EndOffset)
.setMIFlag(MachineInstr::FrameSetup);
- // MOV32mr SavedEBPOffset(%esi), %ebp
+ // MOV32rm SavedEBPOffset(%esi), %ebp
assert(X86FI->getHasSEHFramePtrSave());
int Offset =
getFrameIndexReference(MF, X86FI->getSEHFramePtrSaveIndex(), UsedReg);
assert(UsedReg == BasePtr);
- addRegOffset(BuildMI(MBB, MBBI, DL, TII.get(X86::MOV32mr)), UsedReg, true,
- Offset)
- .addReg(FramePtr)
+ addRegOffset(BuildMI(MBB, MBBI, DL, TII.get(X86::MOV32rm), FramePtr),
+ UsedReg, true, Offset)
.setMIFlag(MachineInstr::FrameSetup);
+ } else {
+ llvm_unreachable("32-bit frames with WinEH must use FramePtr or BasePtr");
}
return MBBI;
}
projects / oota-llvm.git / blobdiff