Correct dwarf unwind information in function epilogue
This patch aims to provide correct dwarf unwind information in function
epilogue for X86.
It consists of two parts. The first part inserts CFI instructions that set
appropriate cfa offset and cfa register in emitEpilogue() in
X86FrameLowering. This part is X86 specific.
The second part is platform independent and ensures that:
* CFI instructions do not affect code generation (they are not counted as
instructions when tail duplicating or tail merging)
* Unwind information remains correct when a function is modified by
different passes. This is done in a late pass by analyzing information
about cfa offset and cfa register in BBs and inserting additional CFI
directives where necessary.
Added CFIInstrInserter pass:
* analyzes each basic block to determine cfa offset and register are valid
at its entry and exit
* verifies that outgoing cfa offset and register of predecessor blocks match
incoming values of their successors
* inserts additional CFI directives at basic block beginning to correct the
rule for calculating CFA
Having CFI instructions in function epilogue can cause incorrect CFA
calculation rule for some basic blocks. This can happen if, due to basic
block reordering, or the existence of multiple epilogue blocks, some of the
blocks have wrong cfa offset and register values set by the epilogue block
above them.
CFIInstrInserter is currently run only on X86, but can be used by any target
that implements support for adding CFI instructions in epilogue.
Patch by Violeta Vukobrat.
Differential Revision: https://reviews.llvm.org/D42848
llvm-svn: 330706
diff --git a/llvm/lib/Target/X86/X86FrameLowering.cpp b/llvm/lib/Target/X86/X86FrameLowering.cpp
index d052998..42f9676 100644
--- a/llvm/lib/Target/X86/X86FrameLowering.cpp
+++ b/llvm/lib/Target/X86/X86FrameLowering.cpp
@@ -399,28 +399,30 @@
return 0;
MachineBasicBlock::iterator PI = doMergeWithPrevious ? std::prev(MBBI) : MBBI;
- MachineBasicBlock::iterator NI = doMergeWithPrevious ? nullptr
- : std::next(MBBI);
+
PI = skipDebugInstructionsBackward(PI, MBB.begin());
- if (NI != nullptr)
- NI = skipDebugInstructionsForward(NI, MBB.end());
+ // It is assumed that ADD/SUB/LEA instruction is succeded by one CFI
+ // instruction, and that there are no DBG_VALUE or other instructions between
+ // ADD/SUB/LEA and its corresponding CFI instruction.
+ /* TODO: Add support for the case where there are multiple CFI instructions
+ below the ADD/SUB/LEA, e.g.:
+ ...
+ add
+ cfi_def_cfa_offset
+ cfi_offset
+ ...
+ */
+ if (doMergeWithPrevious && PI != MBB.begin() && PI->isCFIInstruction())
+ PI = std::prev(PI);
unsigned Opc = PI->getOpcode();
int Offset = 0;
- if (!doMergeWithPrevious && NI != MBB.end() &&
- NI->getOpcode() == TargetOpcode::CFI_INSTRUCTION) {
- // Don't merge with the next instruction if it has CFI.
- return Offset;
- }
-
if ((Opc == X86::ADD64ri32 || Opc == X86::ADD64ri8 ||
Opc == X86::ADD32ri || Opc == X86::ADD32ri8) &&
PI->getOperand(0).getReg() == StackPtr){
assert(PI->getOperand(1).getReg() == StackPtr);
- Offset += PI->getOperand(2).getImm();
- MBB.erase(PI);
- if (!doMergeWithPrevious) MBBI = NI;
+ Offset = PI->getOperand(2).getImm();
} else if ((Opc == X86::LEA32r || Opc == X86::LEA64_32r) &&
PI->getOperand(0).getReg() == StackPtr &&
PI->getOperand(1).getReg() == StackPtr &&
@@ -428,17 +430,19 @@
PI->getOperand(3).getReg() == X86::NoRegister &&
PI->getOperand(5).getReg() == X86::NoRegister) {
// For LEAs we have: def = lea SP, FI, noreg, Offset, noreg.
- Offset += PI->getOperand(4).getImm();
- MBB.erase(PI);
- if (!doMergeWithPrevious) MBBI = NI;
+ Offset = PI->getOperand(4).getImm();
} else if ((Opc == X86::SUB64ri32 || Opc == X86::SUB64ri8 ||
Opc == X86::SUB32ri || Opc == X86::SUB32ri8) &&
PI->getOperand(0).getReg() == StackPtr) {
assert(PI->getOperand(1).getReg() == StackPtr);
- Offset -= PI->getOperand(2).getImm();
- MBB.erase(PI);
- if (!doMergeWithPrevious) MBBI = NI;
- }
+ Offset = -PI->getOperand(2).getImm();
+ } else
+ return 0;
+
+ PI = MBB.erase(PI);
+ if (PI != MBB.end() && PI->isCFIInstruction()) PI = MBB.erase(PI);
+ if (!doMergeWithPrevious)
+ MBBI = skipDebugInstructionsForward(PI, MBB.end());
return Offset;
}
@@ -1573,6 +1577,11 @@
bool HasFP = hasFP(MF);
uint64_t NumBytes = 0;
+ bool NeedsDwarfCFI =
+ (!MF.getTarget().getTargetTriple().isOSDarwin() &&
+ !MF.getTarget().getTargetTriple().isOSWindows()) &&
+ (MF.getMMI().hasDebugInfo() || MF.getFunction().needsUnwindTableEntry());
+
if (IsFunclet) {
assert(HasFP && "EH funclets without FP not yet implemented");
NumBytes = getWinEHFuncletFrameSize(MF);
@@ -1595,6 +1604,13 @@
BuildMI(MBB, MBBI, DL, TII.get(Is64Bit ? X86::POP64r : X86::POP32r),
MachineFramePtr)
.setMIFlag(MachineInstr::FrameDestroy);
+ if (NeedsDwarfCFI) {
+ unsigned DwarfStackPtr =
+ TRI->getDwarfRegNum(Is64Bit ? X86::RSP : X86::ESP, true);
+ BuildCFI(MBB, MBBI, DL, MCCFIInstruction::createDefCfa(
+ nullptr, DwarfStackPtr, -SlotSize));
+ --MBBI;
+ }
}
MachineBasicBlock::iterator FirstCSPop = MBBI;
@@ -1658,6 +1674,11 @@
} else if (NumBytes) {
// Adjust stack pointer back: ESP += numbytes.
emitSPUpdate(MBB, MBBI, DL, NumBytes, /*InEpilogue=*/true);
+ if (!hasFP(MF) && NeedsDwarfCFI) {
+ // Define the current CFA rule to use the provided offset.
+ BuildCFI(MBB, MBBI, DL, MCCFIInstruction::createDefCfaOffset(
+ nullptr, -CSSize - SlotSize));
+ }
--MBBI;
}
@@ -1670,6 +1691,23 @@
if (NeedsWin64CFI && MF.hasWinCFI())
BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_Epilogue));
+ if (!hasFP(MF) && NeedsDwarfCFI) {
+ MBBI = FirstCSPop;
+ int64_t Offset = -CSSize - SlotSize;
+ // Mark callee-saved pop instruction.
+ // Define the current CFA rule to use the provided offset.
+ while (MBBI != MBB.end()) {
+ MachineBasicBlock::iterator PI = MBBI;
+ unsigned Opc = PI->getOpcode();
+ ++MBBI;
+ if (Opc == X86::POP32r || Opc == X86::POP64r) {
+ Offset += SlotSize;
+ BuildCFI(MBB, MBBI, DL,
+ MCCFIInstruction::createDefCfaOffset(nullptr, Offset));
+ }
+ }
+ }
+
if (Terminator == MBB.end() || !isTailCallOpcode(Terminator->getOpcode())) {
// Add the return addr area delta back since we are not tail calling.
int Offset = -1 * X86FI->getTCReturnAddrDelta();
@@ -2719,7 +2757,6 @@
// Add Amount to SP to destroy a frame, or subtract to setup.
int64_t StackAdjustment = isDestroy ? Amount : -Amount;
- int64_t CfaAdjustment = -StackAdjustment;
if (StackAdjustment) {
// Merge with any previous or following adjustment instruction. Note: the
@@ -2744,6 +2781,7 @@
// offset to be correct at each call site, while for debugging we want
// it to be more precise.
+ int64_t CfaAdjustment = -StackAdjustment;
// TODO: When not using precise CFA, we also need to adjust for the
// InternalAmt here.
if (CfaAdjustment) {
@@ -2874,6 +2912,15 @@
return MBBI;
}
+int X86FrameLowering::getInitialCFAOffset(const MachineFunction &MF) const {
+ return TRI->getSlotSize();
+}
+
+unsigned X86FrameLowering::getInitialCFARegister(const MachineFunction &MF)
+ const {
+ return TRI->getDwarfRegNum(StackPtr, true);
+}
+
namespace {
// Struct used by orderFrameObjects to help sort the stack objects.
struct X86FrameSortingObject {