AArch64: Implement support for the shadowcallstack attribute.

The implementation of shadow call stack on aarch64 is quite different to
the implementation on x86_64. Instead of reserving a segment register for
the shadow call stack, we reserve the platform register, x18. Any function
that spills lr to sp also spills it to the shadow call stack, a pointer to
which is stored in x18.

Differential Revision: https://reviews.llvm.org/D45239

llvm-svn: 329236

include/llvm/Support/TargetParser.h

@@ -212,6 +212,8 @@ ARM::EndianKind parseArchEndian(StringRef Arch);
 ARM::ProfileKind parseArchProfile(StringRef Arch);
 unsigned parseArchVersion(StringRef Arch);
 
+bool isX18ReservedByDefault(const Triple &TT);
+
 } // namespace AArch64
 
 namespace X86 {

lib/Support/TargetParser.cpp

@@ -917,3 +917,7 @@ ARM::ProfileKind AArch64::parseArchProfile(StringRef Arch) {
 unsigned llvm::AArch64::parseArchVersion(StringRef Arch) {
   return ARM::parseArchVersion(Arch);
 }
+
+bool llvm::AArch64::isX18ReservedByDefault(const Triple &TT) {
+  return TT.isOSDarwin() || TT.isOSFuchsia() || TT.isOSWindows();
+}

lib/Target/AArch64/AArch64CallingConvention.td

@@ -349,3 +349,18 @@ def CSR_AArch64_StackProbe_Windows
     : CalleeSavedRegs<(add (sequence "X%u", 0, 15),
                            (sequence "X%u", 18, 28), FP, SP,
                            (sequence "Q%u", 0, 31))>;
+
+// Variants of the standard calling conventions for shadow call stack.
+// These all preserve x18 in addition to any other registers.
+def CSR_AArch64_NoRegs_SCS
+    : CalleeSavedRegs<(add CSR_AArch64_NoRegs, X18)>;
+def CSR_AArch64_AllRegs_SCS
+    : CalleeSavedRegs<(add CSR_AArch64_AllRegs, X18)>;
+def CSR_AArch64_CXX_TLS_Darwin_SCS
+    : CalleeSavedRegs<(add CSR_AArch64_CXX_TLS_Darwin, X18)>;
+def CSR_AArch64_AAPCS_SwiftError_SCS
+    : CalleeSavedRegs<(add CSR_AArch64_AAPCS_SwiftError, X18)>;
+def CSR_AArch64_RT_MostRegs_SCS
+    : CalleeSavedRegs<(add CSR_AArch64_RT_MostRegs, X18)>;
+def CSR_AArch64_AAPCS_SCS
+    : CalleeSavedRegs<(add CSR_AArch64_AAPCS, X18)>;

lib/Target/AArch64/AArch64FrameLowering.cpp

@@ -414,6 +414,14 @@ bool AArch64FrameLowering::shouldCombineCSRLocalStackBump(
 static MachineBasicBlock::iterator convertCalleeSaveRestoreToSPPrePostIncDec(
     MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI,
     const DebugLoc &DL, const TargetInstrInfo *TII, int CSStackSizeInc) {
+  // Ignore instructions that do not operate on SP, i.e. shadow call stack
+  // instructions.
+  while (MBBI->getOpcode() == AArch64::STRXpost ||
+         MBBI->getOpcode() == AArch64::LDRXpre) {
+    assert(MBBI->getOperand(0).getReg() != AArch64::SP);
+    ++MBBI;
+  }
+
   unsigned NewOpc;
   bool NewIsUnscaled = false;
   switch (MBBI->getOpcode()) {
@@ -481,6 +489,14 @@ static MachineBasicBlock::iterator convertCalleeSaveRestoreToSPPrePostIncDec(
 static void fixupCalleeSaveRestoreStackOffset(MachineInstr &MI,
                                               unsigned LocalStackSize) {
   unsigned Opc = MI.getOpcode();
+
+  // Ignore instructions that do not operate on SP, i.e. shadow call stack
+  // instructions.
+  if (Opc == AArch64::STRXpost || Opc == AArch64::LDRXpre) {
+    assert(MI.getOperand(0).getReg() != AArch64::SP);
+    return;
+  }
+
   (void)Opc;
   assert((Opc == AArch64::STPXi || Opc == AArch64::STPDi ||
           Opc == AArch64::STRXui || Opc == AArch64::STRDui ||
@@ -935,6 +951,18 @@ void AArch64FrameLowering::emitEpilogue(MachineFunction &MF,
   // assumes the SP is at the same location as it was after the callee-save save
   // code in the prologue.
   if (AfterCSRPopSize) {
+    // Find an insertion point for the first ldp so that it goes before the
+    // shadow call stack epilog instruction. This ensures that the restore of
+    // lr from x18 is placed after the restore from sp.
+    auto FirstSPPopI = MBB.getFirstTerminator();
+    while (FirstSPPopI != Begin) {
+      auto Prev = std::prev(FirstSPPopI);
+      if (Prev->getOpcode() != AArch64::LDRXpre ||
+          Prev->getOperand(0).getReg() == AArch64::SP)
+        break;
+      FirstSPPopI = Prev;
+    }
+
     // Sometimes (when we restore in the same order as we save), we can end up
     // with code like this:
     //
@@ -949,7 +977,7 @@ void AArch64FrameLowering::emitEpilogue(MachineFunction &MF,
     // a post-index ldp.
     // If we managed to grab the first pop instruction, move it to the end.
     if (LastPopI != Begin)
-      MBB.splice(MBB.getFirstTerminator(), &MBB, LastPopI);
+      MBB.splice(FirstSPPopI, &MBB, LastPopI);
     // We should end up with something like this now:
     //
     // ldp      x24, x23, [sp, #16]
@@ -962,7 +990,7 @@ void AArch64FrameLowering::emitEpilogue(MachineFunction &MF,
     //
     // ldp      x26, x25, [sp], #64
     //
-    emitFrameOffset(MBB, MBB.getFirstTerminator(), DL, AArch64::SP, AArch64::SP,
+    emitFrameOffset(MBB, FirstSPPopI, DL, AArch64::SP, AArch64::SP,
                     AfterCSRPopSize, TII, MachineInstr::FrameDestroy);
   }
 }
@@ -1081,7 +1109,8 @@ struct RegPairInfo {
 
 static void computeCalleeSaveRegisterPairs(
     MachineFunction &MF, const std::vector<CalleeSavedInfo> &CSI,
-    const TargetRegisterInfo *TRI, SmallVectorImpl<RegPairInfo> &RegPairs) {
+    const TargetRegisterInfo *TRI, SmallVectorImpl<RegPairInfo> &RegPairs,
+    bool &NeedShadowCallStackProlog) {
 
   if (CSI.empty())
     return;
@@ -1115,6 +1144,15 @@ static void computeCalleeSaveRegisterPairs(
         RPI.Reg2 = NextReg;
     }
 
+    // If either of the registers to be saved is the lr register, it means that
+    // we also need to save lr in the shadow call stack.
+    if ((RPI.Reg1 == AArch64::LR || RPI.Reg2 == AArch64::LR) &&
+        MF.getFunction().hasFnAttribute(Attribute::ShadowCallStack)) {
+      if (!MF.getSubtarget<AArch64Subtarget>().isX18Reserved())
+        report_fatal_error("Must reserve x18 to use shadow call stack");
+      NeedShadowCallStackProlog = true;
+    }
+
     // GPRs and FPRs are saved in pairs of 64-bit regs. We expect the CSI
     // list to come in sorted by frame index so that we can issue the store
     // pair instructions directly. Assert if we see anything otherwise.
@@ -1165,9 +1203,24 @@ bool AArch64FrameLowering::spillCalleeSavedRegisters(
   DebugLoc DL;
   SmallVector<RegPairInfo, 8> RegPairs;
 
-  computeCalleeSaveRegisterPairs(MF, CSI, TRI, RegPairs);
+  bool NeedShadowCallStackProlog = false;
+  computeCalleeSaveRegisterPairs(MF, CSI, TRI, RegPairs,
+                                 NeedShadowCallStackProlog);
   const MachineRegisterInfo &MRI = MF.getRegInfo();
 
+  if (NeedShadowCallStackProlog) {
+    // Shadow call stack prolog: str x30, [x18], #8
+    BuildMI(MBB, MI, DL, TII.get(AArch64::STRXpost))
+        .addReg(AArch64::X18, RegState::Define)
+        .addReg(AArch64::LR)
+        .addReg(AArch64::X18)
+        .addImm(8)
+        .setMIFlag(MachineInstr::FrameSetup);
+
+    // This instruction also makes x18 live-in to the entry block.
+    MBB.addLiveIn(AArch64::X18);
+  }
+
   for (auto RPII = RegPairs.rbegin(), RPIE = RegPairs.rend(); RPII != RPIE;
        ++RPII) {
     RegPairInfo RPI = *RPII;
@@ -1231,7 +1284,9 @@ bool AArch64FrameLowering::restoreCalleeSavedRegisters(
   if (MI != MBB.end())
     DL = MI->getDebugLoc();
 
-  computeCalleeSaveRegisterPairs(MF, CSI, TRI, RegPairs);
+  bool NeedShadowCallStackProlog = false;
+  computeCalleeSaveRegisterPairs(MF, CSI, TRI, RegPairs,
+                                 NeedShadowCallStackProlog);
 
   auto EmitMI = [&](const RegPairInfo &RPI) {
     unsigned Reg1 = RPI.Reg1;
@@ -1280,6 +1335,17 @@ bool AArch64FrameLowering::restoreCalleeSavedRegisters(
   else
     for (const RegPairInfo &RPI : RegPairs)
       EmitMI(RPI);
+
+  if (NeedShadowCallStackProlog) {
+    // Shadow call stack epilog: ldr x30, [x18, #-8]!
+    BuildMI(MBB, MI, DL, TII.get(AArch64::LDRXpre))
+        .addReg(AArch64::X18, RegState::Define)
+        .addReg(AArch64::LR, RegState::Define)
+        .addReg(AArch64::X18)
+        .addImm(-8)
+        .setMIFlag(MachineInstr::FrameDestroy);
+  }
+
   return true;
 }
 

lib/Target/AArch64/AArch64RegisterInfo.cpp

@@ -75,21 +75,25 @@ const MCPhysReg *AArch64RegisterInfo::getCalleeSavedRegsViaCopy(
 const uint32_t *
 AArch64RegisterInfo::getCallPreservedMask(const MachineFunction &MF,
                                           CallingConv::ID CC) const {
+  bool SCS = MF.getFunction().hasFnAttribute(Attribute::ShadowCallStack);
   if (CC == CallingConv::GHC)
     // This is academic because all GHC calls are (supposed to be) tail calls
-    return CSR_AArch64_NoRegs_RegMask;
+    return SCS ? CSR_AArch64_NoRegs_SCS_RegMask : CSR_AArch64_NoRegs_RegMask;
   if (CC == CallingConv::AnyReg)
-    return CSR_AArch64_AllRegs_RegMask;
+    return SCS ? CSR_AArch64_AllRegs_SCS_RegMask : CSR_AArch64_AllRegs_RegMask;
   if (CC == CallingConv::CXX_FAST_TLS)
-    return CSR_AArch64_CXX_TLS_Darwin_RegMask;
+    return SCS ? CSR_AArch64_CXX_TLS_Darwin_SCS_RegMask
+               : CSR_AArch64_CXX_TLS_Darwin_RegMask;
   if (MF.getSubtarget<AArch64Subtarget>().getTargetLowering()
           ->supportSwiftError() &&
       MF.getFunction().getAttributes().hasAttrSomewhere(Attribute::SwiftError))
-    return CSR_AArch64_AAPCS_SwiftError_RegMask;
+    return SCS ? CSR_AArch64_AAPCS_SwiftError_SCS_RegMask
+               : CSR_AArch64_AAPCS_SwiftError_RegMask;
   if (CC == CallingConv::PreserveMost)
-    return CSR_AArch64_RT_MostRegs_RegMask;
+    return SCS ? CSR_AArch64_RT_MostRegs_SCS_RegMask
+               : CSR_AArch64_RT_MostRegs_RegMask;
   else
-    return CSR_AArch64_AAPCS_RegMask;
+    return SCS ? CSR_AArch64_AAPCS_SCS_RegMask : CSR_AArch64_AAPCS_RegMask;
 }
 
 const uint32_t *AArch64RegisterInfo::getTLSCallPreservedMask() const {

lib/Target/AArch64/AArch64Subtarget.cpp

@@ -24,6 +24,7 @@
 #include "llvm/CodeGen/GlobalISel/InstructionSelect.h"
 #include "llvm/CodeGen/MachineScheduler.h"
 #include "llvm/IR/GlobalValue.h"
+#include "llvm/Support/TargetParser.h"
 
 using namespace llvm;
 
@@ -151,8 +152,8 @@ AArch64Subtarget::AArch64Subtarget(const Triple &TT, const std::string &CPU,
                                    const std::string &FS,
                                    const TargetMachine &TM, bool LittleEndian)
     : AArch64GenSubtargetInfo(TT, CPU, FS),
-      ReserveX18(TT.isOSDarwin() || TT.isOSFuchsia() || TT.isOSWindows()),
+      ReserveX18(AArch64::isX18ReservedByDefault(TT)), IsLittle(LittleEndian),
-      IsLittle(LittleEndian), TargetTriple(TT), FrameLowering(),
+      TargetTriple(TT), FrameLowering(),
       InstrInfo(initializeSubtargetDependencies(FS, CPU)), TSInfo(),
       TLInfo(TM, *this) {
   CallLoweringInfo.reset(new AArch64CallLowering(*getTargetLowering()));

test/CodeGen/AArch64/shadow-call-stack.ll

@@ -0,0 +1,47 @@
+; RUN: llc -verify-machineinstrs -o - %s -mtriple=aarch64-linux-gnu -mattr=+reserve-x18 | FileCheck %s
+
+define void @f1() shadowcallstack {
+  ; CHECK: f1:
+  ; CHECK-NOT: x18
+  ; CHECK: ret
+  ret void
+}
+
+declare void @foo()
+
+define void @f2() shadowcallstack {
+  ; CHECK: f2:
+  ; CHECK-NOT: x18
+  ; CHECK: b foo
+  tail call void @foo()
+  ret void
+}
+
+declare i32 @bar()
+
+define i32 @f3() shadowcallstack {
+  ; CHECK: f3:
+  ; CHECK: str x30, [x18], #8
+  ; CHECK: str x30, [sp, #-16]!
+  %res = call i32 @bar()
+  %res1 = add i32 %res, 1
+  ; CHECK: ldr x30, [sp], #16
+  ; CHECK: ldr x30, [x18, #-8]!
+  ; CHECK: ret
+  ret i32 %res
+}
+
+define i32 @f4() shadowcallstack {
+  ; CHECK: f4:
+  %res1 = call i32 @bar()
+  %res2 = call i32 @bar()
+  %res3 = call i32 @bar()
+  %res4 = call i32 @bar()
+  %res12 = add i32 %res1, %res2
+  %res34 = add i32 %res3, %res4
+  %res1234 = add i32 %res12, %res34
+  ; CHECK: ldp {{.*}}x30, [sp
+  ; CHECK: ldr x30, [x18, #-8]!
+  ; CHECK: ret
+  ret i32 %res1234
+}