[PATCH] CodeGen: Resize LifetimeExtendedCleanupHeader to avoid alignment issues

[Richard Smith]

LGTM

On Tue, Jun 30, 2015 at 4:58 PM, Justin Bogner <mail at justinbogner.com>
wrote:

> ping
>
> Justin Bogner <mail at justinbogner.com> writes:
> > The LifetimeExtendedCleanupHeader is carefully fit into 32 bytes,
> > meaning that cleanups on the LifetimeExtendedCleanupStack are *always*
> > allocated at a misaligned address and cause undefined behaviour.
> >
> > There are two ways to solve this - add padding after the header when
> > we allocated our cleanups, or just simplify the header and let it use
> > 64 bits in the first place. I've opted for the latter, and added a
> > static assert to avoid the issue in the future.
> >
> > This does waste some space in the 32-bit case. It wouldn't be too hard
> > to make the size of this struct dependent on the host machine if
> > that's an issue. WDYT?
> >
> > commit 1f1d41a60f5106e9e89fbc917bfc0826b5caad02
> > Author: Justin Bogner <mail at justinbogner.com>
> > Date:   Mon Jun 22 16:02:17 2015 -0700
> >
> >     CodeGen: Resize LifetimeExtendedCleanupHeader to avoid alignment
> issues
> >
> >     The LifetimeExtendedCleanupHeader is carefully fit into 32 bytes,
> >     meaning that cleanups on the LifetimeExtendedCleanupStack are
> *always*
> >     allocated at a misaligned address and cause undefined behaviour.
> >
> >     There are two ways to solve this - add padding after the header when
> >     we allocated our cleanups, or just simplify the header and let it use
> >     64 bits in the first place. I've opted for the latter, and added a
> >     static assert to avoid the issue in the future.
> >
> > diff --git a/lib/CodeGen/CodeGenFunction.h
> b/lib/CodeGen/CodeGenFunction.h
> > index 45475d1..0891b5d 100644
> > --- a/lib/CodeGen/CodeGenFunction.h
> > +++ b/lib/CodeGen/CodeGenFunction.h
> > @@ -283,12 +283,12 @@ public:
> >    /// Header for data within LifetimeExtendedCleanupStack.
> >    struct LifetimeExtendedCleanupHeader {
> >      /// The size of the following cleanup object.
> > -    unsigned Size : 29;
> > +    unsigned Size;
> >      /// The kind of cleanup to push: a value from the CleanupKind
> enumeration.
> > -    unsigned Kind : 3;
> > +    CleanupKind Kind;
> >
> > -    size_t getSize() const { return size_t(Size); }
> > -    CleanupKind getKind() const { return
> static_cast<CleanupKind>(Kind); }
> > +    size_t getSize() const { return Size; }
> > +    CleanupKind getKind() const { return Kind; }
> >    };
> >
> >    /// i32s containing the indexes of the cleanup destinations.
> > @@ -388,6 +388,8 @@ public:
> >      LifetimeExtendedCleanupStack.resize(
> >          LifetimeExtendedCleanupStack.size() + sizeof(Header) +
> Header.Size);
> >
> > +    static_assert(sizeof(Header) % llvm::AlignOf<T>::Alignment == 0,
> > +                  "Cleanup will be allocated on misaligned address");
> >      char *Buffer = &LifetimeExtendedCleanupStack[OldSize];
> >      new (Buffer) LifetimeExtendedCleanupHeader(Header);
> >      new (Buffer + sizeof(Header)) T(A...);
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20150630/4e239421/attachment.html>