[AArch64] Stack probing for function prologues (#66524)

This adds code to AArch64 function prologues to protect against stack
clash attacks by probing (writing to) the stack at regular enough
intervals to ensure that the guard page cannot be skipped over.

The patch depends on and maintains the following invariants:

Upon function entry the caller guarantees that it has probed the stack
(e.g. performed a store) at some address [sp, #N], where`0 <= N <=
1024`. This invariant comes from a requirement for compatibility with
GCC. Any address range in the allocated stack, no smaller than
stack-probe-size bytes contains at least one probe At any time the stack
pointer is above or in the guard page Probes are performed in
descreasing address order
The stack-probe-size is a function attribute that can be set by a
platform to correspond to the guard page size.

By default, the stack probe size is 4KiB, which is a safe default as
this is the smallest possible page size for AArch64. Linux uses a 64KiB
guard for AArch64, so this can be overridden by the stack-probe-size
function attribute.

For small frames without a frame pointer (<= 240 bytes), no probes are
needed.

For larger frame sizes, LLVM always stores x29 to the stack. This serves
as an implicit stack probe. Thus, while allocating stack objects the
compiler assumes that the stack has been probed at [sp].

There are multiple probing sequences that can be emitted, depending on
the size of the stack allocation:

A straight-line sequence of subtracts and stores, used when the
allocation size is smaller than 5 guard pages. A loop allocating and
probing one page size per iteration, plus at most a single probe to deal
with the remainder, used when the allocation size is larger but still
known at compile time. A loop which moves the SP down to the target
value held in a register (or a loop, moving a scratch register to the
target value help in SP), used when the allocation size is not known at
compile-time, such as when allocating space for SVE values, or when
over-aligning the stack. This is emitted in AArch64InstrInfo because it
will also be used for dynamic allocas in a future patch. A single probe
where the amount of stack adjustment is unknown, but is known to be less
than or equal to a page size.

---------

Co-authored-by: Oliver Stannard <oliver.stannard@linaro.org>

Author: momchil-velikov

llvm/lib/Target/AArch64/AArch64FrameLowering.cpp

@@ -152,13 +152,26 @@ class AArch64FrameLowering : public TargetFrameLowering {
                                   MachineBasicBlock::iterator MBBI) const;
   void allocateStackSpace(MachineBasicBlock &MBB,
                           MachineBasicBlock::iterator MBBI,
-                          bool NeedsRealignment, StackOffset AllocSize,
+                          int64_t RealignmentPadding, StackOffset AllocSize,
                           bool NeedsWinCFI, bool *HasWinCFI, bool EmitCFI,
-                          StackOffset InitialOffset) const;
+                          StackOffset InitialOffset, bool FollowupAllocs) const;
 
   /// Emit target zero call-used regs.
   void emitZeroCallUsedRegs(BitVector RegsToZero,
                             MachineBasicBlock &MBB) const override;
+
+  /// Replace a StackProbe stub (if any) with the actual probe code inline
+  void inlineStackProbe(MachineFunction &MF,
+                        MachineBasicBlock &PrologueMBB) const override;
+
+  void inlineStackProbeFixed(MachineBasicBlock::iterator MBBI,
+                             Register ScratchReg, int64_t FrameSize,
+                             StackOffset CFAOffset) const;
+
+  MachineBasicBlock::iterator
+  inlineStackProbeLoopExactMultiple(MachineBasicBlock::iterator MBBI,
+                                    int64_t NegProbeSize,
+                                    Register TargetReg) const;
 };
 
 } // End llvm namespace

llvm/lib/Target/AArch64/AArch64FrameLowering.h

@@ -26815,3 +26815,9 @@ unsigned AArch64TargetLowering::getVectorTypeBreakdownForCallingConv(
 
   return NumRegs;
 }
+
+bool AArch64TargetLowering::hasInlineStackProbe(
+    const MachineFunction &MF) const {
+  return !Subtarget->isTargetWindows() &&
+         MF.getInfo<AArch64FunctionInfo>()->hasStackProbing();
+}

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -514,6 +514,13 @@ const unsigned RoundingBitsPos = 22;
 ArrayRef<MCPhysReg> getGPRArgRegs();
 ArrayRef<MCPhysReg> getFPRArgRegs();
 
+/// Maximum allowed number of unprobed bytes above SP at an ABI
+/// boundary.
+const unsigned StackProbeMaxUnprobedStack = 1024;
+
+/// Maximum number of iterations to unroll for a constant size probing loop.
+const unsigned StackProbeMaxLoopUnroll = 4;
+
 } // namespace AArch64
 
 class AArch64Subtarget;
@@ -966,6 +973,9 @@ class AArch64TargetLowering : public TargetLowering {
                                                 unsigned &NumIntermediates,
                                                 MVT &RegisterVT) const override;
 
+  /// True if stack clash protection is enabled for this functions.
+  bool hasInlineStackProbe(const MachineFunction &MF) const override;
+
 private:
   /// Keep a pointer to the AArch64Subtarget around so that we can
   /// make the right decision when generating code for different targets.

llvm/lib/Target/AArch64/AArch64ISelLowering.h

@@ -10,8 +10,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "AArch64ExpandImm.h"
 #include "AArch64InstrInfo.h"
+#include "AArch64ExpandImm.h"
 #include "AArch64FrameLowering.h"
 #include "AArch64MachineFunctionInfo.h"
 #include "AArch64PointerAuth.h"
@@ -21,6 +21,7 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
+#include "llvm/CodeGen/LivePhysRegs.h"
 #include "llvm/CodeGen/MachineBasicBlock.h"
 #include "llvm/CodeGen/MachineCombinerPattern.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
@@ -9467,6 +9468,94 @@ bool AArch64InstrInfo::isReallyTriviallyReMaterializable(
   return TargetInstrInfo::isReallyTriviallyReMaterializable(MI);
 }
 
+MachineBasicBlock::iterator
+AArch64InstrInfo::probedStackAlloc(MachineBasicBlock::iterator MBBI,
+                                   Register TargetReg, bool FrameSetup) const {
+  assert(TargetReg != AArch64::SP && "New top of stack cannot aleady be in SP");
+
+  MachineBasicBlock &MBB = *MBBI->getParent();
+  MachineFunction &MF = *MBB.getParent();
+  const AArch64InstrInfo *TII =
+      MF.getSubtarget<AArch64Subtarget>().getInstrInfo();
+  int64_t ProbeSize = MF.getInfo<AArch64FunctionInfo>()->getStackProbeSize();
+  DebugLoc DL = MBB.findDebugLoc(MBBI);
+
+  MachineFunction::iterator MBBInsertPoint = std::next(MBB.getIterator());
+  MachineBasicBlock *LoopTestMBB =
+      MF.CreateMachineBasicBlock(MBB.getBasicBlock());
+  MF.insert(MBBInsertPoint, LoopTestMBB);
+  MachineBasicBlock *LoopBodyMBB =
+      MF.CreateMachineBasicBlock(MBB.getBasicBlock());
+  MF.insert(MBBInsertPoint, LoopBodyMBB);
+  MachineBasicBlock *ExitMBB = MF.CreateMachineBasicBlock(MBB.getBasicBlock());
+  MF.insert(MBBInsertPoint, ExitMBB);
+  MachineInstr::MIFlag Flags =
+      FrameSetup ? MachineInstr::FrameSetup : MachineInstr::NoFlags;
+
+  // LoopTest:
+  //   SUB SP, SP, #ProbeSize
+  emitFrameOffset(*LoopTestMBB, LoopTestMBB->end(), DL, AArch64::SP,
+                  AArch64::SP, StackOffset::getFixed(-ProbeSize), TII, Flags);
+
+  //   CMP SP, TargetReg
+  BuildMI(*LoopTestMBB, LoopTestMBB->end(), DL, TII->get(AArch64::SUBSXrx64),
+          AArch64::XZR)
+      .addReg(AArch64::SP)
+      .addReg(TargetReg)
+      .addImm(AArch64_AM::getArithExtendImm(AArch64_AM::UXTX, 0))
+      .setMIFlags(Flags);
+
+  //   B.<Cond> LoopExit
+  BuildMI(*LoopTestMBB, LoopTestMBB->end(), DL, TII->get(AArch64::Bcc))
+      .addImm(AArch64CC::LE)
+      .addMBB(ExitMBB)
+      .setMIFlags(Flags);
+
+  //   STR XZR, [SP]
+  BuildMI(*LoopBodyMBB, LoopBodyMBB->end(), DL, TII->get(AArch64::STRXui))
+      .addReg(AArch64::XZR)
+      .addReg(AArch64::SP)
+      .addImm(0)
+      .setMIFlags(Flags);
+
+  //   B loop
+  BuildMI(*LoopBodyMBB, LoopBodyMBB->end(), DL, TII->get(AArch64::B))
+      .addMBB(LoopTestMBB)
+      .setMIFlags(Flags);
+
+  // LoopExit:
+  //   MOV SP, TargetReg
+  BuildMI(*ExitMBB, ExitMBB->end(), DL, TII->get(AArch64::ADDXri), AArch64::SP)
+      .addReg(TargetReg)
+      .addImm(0)
+      .addImm(AArch64_AM::getShifterImm(AArch64_AM::LSL, 0))
+      .setMIFlags(Flags);
+
+  //   STR XZR, [SP]
+  BuildMI(*ExitMBB, ExitMBB->end(), DL, TII->get(AArch64::STRXui))
+      .addReg(AArch64::XZR)
+      .addReg(AArch64::SP)
+      .addImm(0)
+      .setMIFlags(Flags);
+
+  ExitMBB->splice(ExitMBB->end(), &MBB, std::next(MBBI), MBB.end());
+  ExitMBB->transferSuccessorsAndUpdatePHIs(&MBB);
+
+  LoopTestMBB->addSuccessor(ExitMBB);
+  LoopTestMBB->addSuccessor(LoopBodyMBB);
+  LoopBodyMBB->addSuccessor(LoopTestMBB);
+  MBB.addSuccessor(LoopTestMBB);
+
+  // Update liveins.
+  if (MF.getRegInfo().reservedRegsFrozen()) {
+    recomputeLiveIns(*LoopTestMBB);
+    recomputeLiveIns(*LoopBodyMBB);
+    recomputeLiveIns(*ExitMBB);
+  }
+
+  return ExitMBB->begin();
+}
+
 #define GET_INSTRINFO_HELPERS
 #define GET_INSTRMAP_INFO
 #include "AArch64GenInstrInfo.inc"

llvm/lib/Target/AArch64/AArch64InstrInfo.cpp

@@ -383,6 +383,13 @@ class AArch64InstrInfo final : public AArch64GenInstrInfo {
   bool isLegalAddressingMode(unsigned NumBytes, int64_t Offset,
                              unsigned Scale) const;
 
+  // Decrement the SP, issuing probes along the way. `TargetReg` is the new top
+  // of the stack. `FrameSetup` is passed as true, if the allocation is a part
+  // of constructing the activation frame of a function.
+  MachineBasicBlock::iterator probedStackAlloc(MachineBasicBlock::iterator MBBI,
+                                               Register TargetReg,
+                                               bool FrameSetup) const;
+
 #define GET_INSTRINFO_HELPER_DECLS
 #include "AArch64GenInstrInfo.inc"
 

llvm/lib/Target/AArch64/AArch64InstrInfo.h

@@ -936,7 +936,8 @@ include "SMEInstrFormats.td"
 // Miscellaneous instructions.
 //===----------------------------------------------------------------------===//
 
-let Defs = [SP], Uses = [SP], hasSideEffects = 1, isCodeGenOnly = 1 in {
+let hasSideEffects = 1, isCodeGenOnly = 1 in {
+let Defs = [SP], Uses = [SP] in {
 // We set Sched to empty list because we expect these instructions to simply get
 // removed in most cases.
 def ADJCALLSTACKDOWN : Pseudo<(outs), (ins i32imm:$amt1, i32imm:$amt2),
@@ -945,7 +946,27 @@ def ADJCALLSTACKDOWN : Pseudo<(outs), (ins i32imm:$amt1, i32imm:$amt2),
 def ADJCALLSTACKUP : Pseudo<(outs), (ins i32imm:$amt1, i32imm:$amt2),
                             [(AArch64callseq_end timm:$amt1, timm:$amt2)]>,
                             Sched<[]>;
-} // Defs = [SP], Uses = [SP], hasSideEffects = 1, isCodeGenOnly = 1
+
+}
+
+let Defs = [SP, NZCV], Uses = [SP] in {
+// Probed stack allocation of a constant size, used in function prologues when
+// stack-clash protection is enabled.
+def PROBED_STACKALLOC : Pseudo<(outs GPR64:$scratch),
+                               (ins i64imm:$stacksize, i64imm:$fixed_offset,
+                                i64imm:$scalable_offset),
+                               []>,
+                               Sched<[]>;
+
+// Probed stack allocation of a variable size, used in function prologues when
+// stack-clash protection is enabled.
+def PROBED_STACKALLOC_VAR : Pseudo<(outs),
+                                   (ins GPR64sp:$target),
+                                   []>,
+                                   Sched<[]>;
+
+} // Defs = [SP, NZCV], Uses = [SP] in 
+} // hasSideEffects = 1, isCodeGenOnly = 1
 
 let isReMaterializable = 1, isCodeGenOnly = 1 in {
 // FIXME: The following pseudo instructions are only needed because remat

llvm/lib/Target/AArch64/AArch64InstrInfo.td

@@ -97,14 +97,45 @@ AArch64FunctionInfo::AArch64FunctionInfo(const Function &F,
     if (const auto *BTE = mdconst::extract_or_null<ConstantInt>(
             F.getParent()->getModuleFlag("branch-target-enforcement")))
       BranchTargetEnforcement = BTE->getZExtValue();
-    return;
+  } else {
+    const StringRef BTIEnable =
+        F.getFnAttribute("branch-target-enforcement").getValueAsString();
+    assert(BTIEnable.equals_insensitive("true") ||
+           BTIEnable.equals_insensitive("false"));
+    BranchTargetEnforcement = BTIEnable.equals_insensitive("true");
   }
 
-  const StringRef BTIEnable =
-      F.getFnAttribute("branch-target-enforcement").getValueAsString();
-  assert(BTIEnable.equals_insensitive("true") ||
-         BTIEnable.equals_insensitive("false"));
-  BranchTargetEnforcement = BTIEnable.equals_insensitive("true");
+  // The default stack probe size is 4096 if the function has no
+  // stack-probe-size attribute. This is a safe default because it is the
+  // smallest possible guard page size.
+  uint64_t ProbeSize = 4096;
+  if (F.hasFnAttribute("stack-probe-size"))
+    ProbeSize = F.getFnAttributeAsParsedInteger("stack-probe-size");
+  else if (const auto *PS = mdconst::extract_or_null<ConstantInt>(
+               F.getParent()->getModuleFlag("stack-probe-size")))
+    ProbeSize = PS->getZExtValue();
+  assert(int64_t(ProbeSize) > 0 && "Invalid stack probe size");
+
+  if (STI->isTargetWindows()) {
+    if (!F.hasFnAttribute("no-stack-arg-probe"))
+      StackProbeSize = ProbeSize;
+  } else {
+    // Round down to the stack alignment.
+    uint64_t StackAlign =
+        STI->getFrameLowering()->getTransientStackAlign().value();
+    ProbeSize = std::max(StackAlign, ProbeSize & ~(StackAlign - 1U));
+    StringRef ProbeKind;
+    if (F.hasFnAttribute("probe-stack"))
+      ProbeKind = F.getFnAttribute("probe-stack").getValueAsString();
+    else if (const auto *PS = dyn_cast_or_null<MDString>(
+                 F.getParent()->getModuleFlag("probe-stack")))
+      ProbeKind = PS->getString();
+    if (ProbeKind.size()) {
+      if (ProbeKind != "inline-asm")
+        report_fatal_error("Unsupported stack probing method");
+      StackProbeSize = ProbeSize;
+    }
+  }
 }
 
 MachineFunctionInfo *AArch64FunctionInfo::clone(

llvm/lib/Target/AArch64/AArch64MachineFunctionInfo.cpp

@@ -194,6 +194,8 @@ class AArch64FunctionInfo final : public MachineFunctionInfo {
   /// True if the function need asynchronous unwind information.
   mutable std::optional<bool> NeedsAsyncDwarfUnwindInfo;
 
+  int64_t StackProbeSize = 0;
+
 public:
   AArch64FunctionInfo(const Function &F, const AArch64Subtarget *STI);
 
@@ -456,6 +458,10 @@ class AArch64FunctionInfo final : public MachineFunctionInfo {
     HasStreamingModeChanges = HasChanges;
   }
 
+  bool hasStackProbing() const { return StackProbeSize != 0; }
+
+  int64_t getStackProbeSize() const { return StackProbeSize; }
+
 private:
   // Hold the lists of LOHs.
   MILOHContainer LOHContainerSet;