[PATCH] Don't force the use of a base pointer with MS inline asm

[Reid Kleckner]

Jakob, can you take a look at this?

Unfortunately, the fatal error I added actually comes up far more often
than I expected because LLVM will try to align stack frames for code like:

void foo() {
  struct { int x; int y; } p;  // LLVM wants this to be 8 byte aligned. Why?
  __asm { ... adjust esp and use p }
}

I'd prefer it if LLVM failed instead of miscompiling this code, though.

I wasn't able to do a complete fix to pick the base register dynamically
because I can't figure out how to pull the constraints off inline asm in MI.

I would also reject gcc inline asm that clobbers esp and in the presence of
stack realignment if I could see the constraints.


On Tue, Aug 13, 2013 at 6:09 PM, Reid Kleckner <rnk at google.com> wrote:

>     - Add a fatal error.
>
> http://llvm-reviews.chandlerc.com/D1317
>
> CHANGE SINCE LAST DIFF
>   http://llvm-reviews.chandlerc.com/D1317?vs=3269&id=3460#toc
>
> Files:
>   lib/Target/X86/X86RegisterInfo.cpp
>   test/CodeGen/X86/inline-asm-stack-realign.ll
>   test/CodeGen/X86/ms-inline-asm.ll
>
> Index: lib/Target/X86/X86RegisterInfo.cpp
> ===================================================================
> --- lib/Target/X86/X86RegisterInfo.cpp
> +++ lib/Target/X86/X86RegisterInfo.cpp
> @@ -340,6 +340,11 @@
>          "Stack realignment in presence of dynamic allocas is not
> supported with"
>          "this calling convention.");
>
> +    if (MF.hasMSInlineAsm())
> +      report_fatal_error(
> +        "Stack realignment in the presence of stack-adjusting inline
> assembly "
> +        "is not supported");
> +
>      for (MCSubRegIterator I(getBaseRegister(), this,
> /*IncludeSelf=*/true);
>           I.isValid(); ++I)
>        Reserved.set(*I);
> @@ -396,18 +401,16 @@
>     if (!EnableBasePointer)
>       return false;
>
> -   // When we need stack realignment and there are dynamic allocas, we
> can't
> -   // reference off of the stack pointer, so we reserve a base pointer.
> -   //
> -   // This is also true if the function contain MS-style inline assembly.
>  We
> -   // do this because if any stack changes occur in the inline assembly,
> e.g.,
> -   // "pusha", then any C local variable or C argument references in the
> -   // inline assembly will be wrong because the SP is not properly
> tracked.
> -   if ((needsStackRealignment(MF) && MFI->hasVarSizedObjects()) ||
> -       MF.hasMSInlineAsm())
> -     return true;
> -
> -   return false;
> +   // When we need stack realignment, we can't address the stack from the
> frame
> +   // pointer.  When we have dynamic allocas or MS inline asm, we can't
> address
> +   // variables from the stack pointer.  MS inline asm can reference
> locals
> +   // while also adjusting the stack pointer.  When we can't use both the
> SP and
> +   // the FP, we need a separate base pointer register.
> +   // FIXME: gcc inline asm can also adjust the stack pointer if it lists
> esp as
> +   // a clobber, but LLVM does not support that.
> +   bool CantUseFP = needsStackRealignment(MF);
> +   bool CantUseSP = MFI->hasVarSizedObjects() || MF.hasMSInlineAsm();
> +   return CantUseFP && CantUseSP;
>  }
>
>  bool X86RegisterInfo::canRealignStack(const MachineFunction &MF) const {
> Index: test/CodeGen/X86/inline-asm-stack-realign.ll
> ===================================================================
> --- /dev/null
> +++ test/CodeGen/X86/inline-asm-stack-realign.ll
> @@ -0,0 +1,16 @@
> +; RUN: not llc -march x86 %s 2>&1 | FileCheck %s
> +
> +; Realigning the stack and referring to local stack vars through memory in
> +; inline asm with this many clobbers is impossible.
> +; FIXME: LLVM could do better here by looking at the list of clobbers.
> +
> +; CHECK: Stack realignment in presence of MS inline asm is not supported
> +
> +define i32 @foo() {
> +entry:
> +  %r = alloca i32, align 16
> +  store i32 -1, i32* %r, align 16
> +  call void asm sideeffect inteldialect "pusha\0A\09xor eax, eax\0A\09xor
> ebx, ebx\0A\09xor ecx, ecx\0A\09xor edx, edx\0A\09xor esi, esi\0A\09xor
> edi, edi\0A\09mov dword ptr $0, esi\0A\09popa",
> "=*m,~{eax},~{ebx},~{ecx},~{edi},~{edx},~{esi},~{dirflag},~{fpsr},~{flags}"(i32*
> %r) #1
> +  %0 = load i32* %r, align 16
> +  ret i32 %0
> +}
> Index: test/CodeGen/X86/ms-inline-asm.ll
> ===================================================================
> --- test/CodeGen/X86/ms-inline-asm.ll
> +++ test/CodeGen/X86/ms-inline-asm.ll
> @@ -103,8 +103,8 @@
>  ; CHECK: {{## InlineAsm End|#NO_APP}}
>  ; CHECK: {{## InlineAsm Start|#APP}}
>  ; CHECK: .intel_syntax
> -; CHECK: mov dword ptr [esi], edi
> +; CHECK: mov dword ptr [ebp - 8], edi
>  ; CHECK: .att_syntax
>  ; CHECK: {{## InlineAsm End|#NO_APP}}
> -; CHECK: movl (%esi), %eax
> +; CHECK: movl -8(%ebp), %eax
>  }
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130816/e7d12c09/attachment.html>