[llvm] r207615 - ARM: support stack probe emission for Windows on ARM

Kostya Serebryany kcc at google.com
Wed Apr 30 23:18:29 PDT 2014


Indeed, thanks!
check-llvm is now lsan-clean


On Thu, May 1, 2014 at 8:28 AM, Saleem Abdulrasool <compnerd at compnerd.org>wrote:

> On Wed, Apr 30, 2014 at 1:42 AM, Kostya Serebryany <kcc at google.com> wrote:
>
>> I observe a leak report when running this test under lsan or valgrind (
>> http://llvm.org/bugs/show_bug.cgi?id=19521).
>> Please check.
>>
>
> Thank you for bringing this to my attention.  I believe that the leak
> should be addressed by SVN  r207737.  At the very least, valgrind seems to
> be happy with it now.
>
> Indirect leak of 136 byte(s) in 1 object(s) allocated from:
>>
>>     #0 0x65f540 in operator new(unsigned long)
>> projects/compiler-rt/lib/asan/asan_new_delete.cc:62
>>     #1 0x20909d3 in llvm::User::operator new(unsigned long, unsigned int)
>> lib/IR/User.cpp:60
>>     #2 0xa8bb12 in Twine include/llvm/IR/Function.h:131
>>
>>     #3 0xa8bb12 in
>> llvm::ARMFrameLowering::emitPrologue(llvm::MachineFunction&) const
>> lib/Target/ARM/ARMFrameLowering.cpp:322
>>     #4 0x1c49990 in
>> llvm::PEI::insertPrologEpilogCode(llvm::MachineFunction&)
>> lib/CodeGen/PrologEpilogInserter.cpp:671
>>     #5 0x1c43127 in
>> llvm::PEI::runOnMachineFunction(llvm::MachineFunction&)
>> lib/CodeGen/PrologEpilogInserter.cpp:155
>>     #6 0x1b2cbfb in
>> llvm::MachineFunctionPass::runOnFunction(llvm::Function&)
>> lib/CodeGen/MachineFunctionPass.cpp:33
>>     #7 0x204ee04 in llvm::FPPassManager::runOnFunction(llvm::Function&)
>> lib/IR/LegacyPassManager.cpp:1545
>>     #8 0x204f315 in llvm::FPPassManager::runOnModule(llvm::Module&)
>> lib/IR/LegacyPassManager.cpp:1567
>>     #9 0x2050014 in runOnModule lib/IR/LegacyPassManager.cpp:1625
>>
>>     #10 0x2050014 in llvm::legacy::PassManagerImpl::run(llvm::Module&)
>> lib/IR/LegacyPassManager.cpp:1734
>>     #11 0x685d57 in compileModule tools/llc/llc.cpp:356
>>
>>     #12 0x685d57 in main tools/llc/llc.cpp:202
>>
>>     #13 0x7fd3756f276c in __libc_start_main
>> /build/buildd/eglibc-2.15/csu/libc-start.c:226
>>
>>
>> On Wed, Apr 30, 2014 at 11:05 AM, Saleem Abdulrasool <
>> compnerd at compnerd.org> wrote:
>>
>>> Author: compnerd
>>> Date: Wed Apr 30 02:05:07 2014
>>> New Revision: 207615
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=207615&view=rev
>>> Log:
>>> ARM: support stack probe emission for Windows on ARM
>>>
>>> This introduces the stack lowering emission of the stack probe function
>>> for
>>> Windows on ARM. The stack on Windows on ARM is a dynamically paged stack
>>> where
>>> any page allocation which crosses a page boundary of the following guard
>>> page
>>> will cause a page fault. This page fault must be handled by the kernel to
>>> ensure that the page is faulted in. If this does not occur and a write
>>> access
>>> any memory beyond that, the page fault will go unserviced, resulting in
>>> an
>>> abnormal program termination.
>>>
>>> The watermark for the stack probe appears to be at 4080 bytes (for
>>> accommodating the stack guard canaries and stack alignment) when SSP is
>>> enabled.  Otherwise, the stack probe is emitted on the page size
>>> boundary of
>>> 4096 bytes.
>>>
>>> Added:
>>>     llvm/trunk/test/CodeGen/ARM/Windows/chkstk.ll
>>> Modified:
>>>     llvm/trunk/docs/Extensions.rst
>>>     llvm/trunk/lib/Target/ARM/ARMFrameLowering.cpp
>>>
>>> Modified: llvm/trunk/docs/Extensions.rst
>>> URL:
>>> http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/Extensions.rst?rev=207615&r1=207614&r2=207615&view=diff
>>>
>>> ==============================================================================
>>> --- llvm/trunk/docs/Extensions.rst (original)
>>> +++ llvm/trunk/docs/Extensions.rst Wed Apr 30 02:05:07 2014
>>> @@ -159,3 +159,34 @@ different COMDATs:
>>>    .globl Symbol2
>>>    Symbol2:
>>>    .long 1
>>> +
>>> +Target Specific Behaviour
>>> +=========================
>>> +
>>> +Windows on ARM
>>> +--------------
>>> +
>>> +Stack Probe Emission
>>> +^^^^^^^^^^^^^^^^^^^^
>>> +
>>> +The reference implementation (Microsoft Visual Studio 2012) emits stack
>>> probes
>>> +in the following fashion:
>>> +
>>> +.. code-block:: gas
>>> +
>>> +  movw r4, #constant
>>> +  bl __chkstk
>>> +  sub.w sp, sp, r4
>>> +
>>> +However, this has the limitation of 32 MiB (±16MiB).  In order to
>>> accomodate
>>> +larger binaries, LLVM supports the use of ``-mcode-model=large`` to
>>> allow a 4GiB
>>> +range via a slight deviation.  It will generate an indirect jump as
>>> follows:
>>> +
>>> +.. code-block:: gas
>>> +
>>> +  movw r4, #constant
>>> +  movw r12, :lower16:__chkstk
>>> +  movt r12, :upper16:__chkstk
>>> +  blx r12
>>> +  sub.w sp, sp, r4
>>> +
>>>
>>> Modified: llvm/trunk/lib/Target/ARM/ARMFrameLowering.cpp
>>> URL:
>>> http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMFrameLowering.cpp?rev=207615&r1=207614&r2=207615&view=diff
>>>
>>> ==============================================================================
>>> --- llvm/trunk/lib/Target/ARM/ARMFrameLowering.cpp (original)
>>> +++ llvm/trunk/lib/Target/ARM/ARMFrameLowering.cpp Wed Apr 30 02:05:07
>>> 2014
>>> @@ -25,6 +25,7 @@
>>>  #include "llvm/CodeGen/RegisterScavenging.h"
>>>  #include "llvm/IR/CallingConv.h"
>>>  #include "llvm/IR/Function.h"
>>> +#include "llvm/IR/Module.h"
>>>  #include "llvm/MC/MCContext.h"
>>>  #include "llvm/Support/CommandLine.h"
>>>  #include "llvm/Target/TargetOptions.h"
>>> @@ -142,6 +143,14 @@ static int sizeOfSPAdjustment(const Mach
>>>    return count;
>>>  }
>>>
>>> +static bool WindowsRequiresStackProbe(const MachineFunction &MF,
>>> +                                      size_t StackSizeInBytes) {
>>> +  const MachineFrameInfo *MFI = MF.getFrameInfo();
>>> +  if (MFI->getStackProtectorIndex() > 0)
>>> +    return StackSizeInBytes >= 4080;
>>> +  return StackSizeInBytes >= 4096;
>>> +}
>>> +
>>>  void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
>>>    MachineBasicBlock &MBB = MF.front();
>>>    MachineBasicBlock::iterator MBBI = MBB.begin();
>>> @@ -149,15 +158,16 @@ void ARMFrameLowering::emitPrologue(Mach
>>>    ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
>>>    MachineModuleInfo &MMI = MF.getMMI();
>>>    MCContext &Context = MMI.getContext();
>>> +  const TargetMachine &TM = MF.getTarget();
>>>    const MCRegisterInfo *MRI = Context.getRegisterInfo();
>>>    const ARMBaseRegisterInfo *RegInfo =
>>> -    static_cast<const
>>> ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
>>> +    static_cast<const ARMBaseRegisterInfo*>(TM.getRegisterInfo());
>>>    const ARMBaseInstrInfo &TII =
>>> -    *static_cast<const
>>> ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
>>> +    *static_cast<const ARMBaseInstrInfo*>(TM.getInstrInfo());
>>>    assert(!AFI->isThumb1OnlyFunction() &&
>>>           "This emitPrologue does not support Thumb1!");
>>>    bool isARM = !AFI->isThumbFunction();
>>> -  unsigned Align =
>>> MF.getTarget().getFrameLowering()->getStackAlignment();
>>> +  unsigned Align = TM.getFrameLowering()->getStackAlignment();
>>>    unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize(Align);
>>>    unsigned NumBytes = MFI->getStackSize();
>>>    const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
>>> @@ -187,7 +197,8 @@ void ARMFrameLowering::emitPrologue(Mach
>>>          .addCFIIndex(CFIIndex);
>>>    }
>>>
>>> -  if (!AFI->hasStackFrame()) {
>>> +  if (!AFI->hasStackFrame() &&
>>> +      (!STI.isTargetWindows() || !WindowsRequiresStackProbe(MF,
>>> NumBytes))) {
>>>      if (NumBytes - ArgRegsSaveSize != 0) {
>>>        emitSPUpdate(isARM, MBB, MBBI, dl, TII, -(NumBytes -
>>> ArgRegsSaveSize),
>>>                     MachineInstr::FrameSetup);
>>> @@ -284,6 +295,50 @@ void ARMFrameLowering::emitPrologue(Mach
>>>    } else
>>>      NumBytes = DPRCSOffset;
>>>
>>> +  if (STI.isTargetWindows() && WindowsRequiresStackProbe(MF, NumBytes))
>>> {
>>> +    uint32_t NumWords = NumBytes >> 2;
>>> +
>>> +    if (NumWords < 65536)
>>> +      AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi16),
>>> ARM::R4)
>>> +                     .addImm(NumWords));
>>> +    else
>>> +      BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi32imm), ARM::R4)
>>> +        .addImm(NumWords);
>>> +
>>> +    switch (TM.getCodeModel()) {
>>> +    case CodeModel::Small:
>>> +    case CodeModel::Medium:
>>> +    case CodeModel::Default:
>>> +    case CodeModel::Kernel:
>>> +      BuildMI(MBB, MBBI, dl, TII.get(ARM::tBL))
>>> +        .addImm((unsigned)ARMCC::AL).addReg(0)
>>> +        .addExternalSymbol("__chkstk")
>>> +        .addReg(ARM::R4, RegState::Implicit);
>>> +      break;
>>> +    case CodeModel::Large:
>>> +    case CodeModel::JITDefault: {
>>> +      LLVMContext &Ctx = MF.getMMI().getModule()->getContext();
>>> +      const GlobalValue *F =
>>> +        Function::Create(FunctionType::get(Type::getVoidTy(Ctx), false),
>>> +                         GlobalValue::AvailableExternallyLinkage,
>>> "__chkstk");
>>> +
>>> +      BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi32imm), ARM::R12)
>>> +        .addGlobalAddress(F);
>>> +      BuildMI(MBB, MBBI, dl, TII.get(ARM::BLX))
>>> +        .addReg(ARM::R12, RegState::Kill)
>>> +        .addReg(ARM::R4, RegState::Implicit);
>>> +      break;
>>> +    }
>>> +    }
>>> +
>>> +    AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl,
>>> TII.get(ARM::t2SUBrr),
>>> +                                        ARM::SP)
>>> +                                .addReg(ARM::SP, RegState::Define)
>>> +                                .addReg(ARM::R4, RegState::Kill)
>>> +                                .setMIFlags(MachineInstr::FrameSetup)));
>>> +    NumBytes = 0;
>>> +  }
>>> +
>>>    unsigned adjustedGPRCS1Size = GPRCS1Size;
>>>    if (NumBytes) {
>>>      // Adjust SP after all the callee-save spills.
>>>
>>> Added: llvm/trunk/test/CodeGen/ARM/Windows/chkstk.ll
>>> URL:
>>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/Windows/chkstk.ll?rev=207615&view=auto
>>>
>>> ==============================================================================
>>> --- llvm/trunk/test/CodeGen/ARM/Windows/chkstk.ll (added)
>>> +++ llvm/trunk/test/CodeGen/ARM/Windows/chkstk.ll Wed Apr 30 02:05:07
>>> 2014
>>> @@ -0,0 +1,24 @@
>>> +; RUN: llc -mtriple=thumbv7-windows -mcpu=cortex-a9 %s -o - \
>>> +; RUN:  | FileCheck -check-prefix CHECK-DEFAULT-CODE-MODEL %s
>>> +
>>> +; RUN: llc -mtriple=thumbv7-windows -mcpu=cortex-a9 -code-model=large
>>> %s -o - \
>>> +; RUN:  | FileCheck -check-prefix CHECK-LARGE-CODE-MODEL %s
>>> +
>>> +define arm_aapcs_vfpcc void @check_watermark() {
>>> +entry:
>>> +  %buffer = alloca [4096 x i8], align 1
>>> +  ret void
>>> +}
>>> +
>>> +; CHECK-DEFAULT-CODE-MODEL: check_watermark:
>>> +; CHECK-DEFAULT-CODE-MODEL:    movw r4, #1024
>>> +; CHECK-DEFAULT-CODE-MODEL:    bl __chkstk
>>> +; CHECK-DEFAULT-CODE-MODEL:    sub.w sp, sp, r4
>>> +
>>> +; CHECK-LARGE-CODE-MODEL: check_watermark:
>>> +; CHECK-LARGE-CODE-MODEL:      movw r4, #1024
>>> +; CHECK-LARGE-CODE-MODEL:      movw r12, :lower16:__chkstk
>>> +; CHECK-LARGE-CODE-MODEL:      movt r12, :upper16:__chkstk
>>> +; CHECK-LARGE-CODE-MODEL:      blx r12
>>> +; CHECK-LARGE-CODE-MODEL:      sub.w sp, sp, r4
>>> +
>>>
>>>
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>>
>>
>>
>
>
> --
> Saleem Abdulrasool
> compnerd (at) compnerd (dot) org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140501/a11c881e/attachment.html>


More information about the llvm-commits mailing list