[PATCH] Ensure proper alignment when reading variadic arguments for x86-32 and x86-64

Mon Aug 18 16:59:23 PDT 2014

Hi Duncan,
Sorry for the long delay. I am not calling ensureMaxAlignment here, because
the semantics of ensureMaxAlignment are too strong for this situation.
Empirically, invoking ensureMaxAlignment seems to ensure %rsp is always
aligned to particular byte width. The tests test/CodeGen/X86/red-zone.ll
and test/CodeGen/X86/red-zone.ll expose the difference in behavior. When
ensureMaxAlignment is called unconditionally, the "subq    $4, %rsp -
addq    $4, %rsp" pairing becomes "pushq   %rax - popq    %rax."

The goal of the patch is to implement section 3.2.3 of the AMD64 ABI
correctly. The controlling sentence is, "The size of each argument gets
rounded up to eightbytes. Therefore the stack will always be eightbyte
aligned." The equivalent sentence in the i386 ABI page 37 says, "At all
times, the stack pointer should point to a word-aligned area." For both
architectures, the stack pointer is not being rounded up to the nearest
eightbyte or word between the last normal argument and the first variadic
argument.

I have attached an updated patch adding the AlignStack helper function as
per your suggestion and a simplified test case based on FileCheck. Thanks
for your help and suggestions.

Tom

On Mon, Aug 4, 2014 at 7:44 PM, Duncan P. N. Exon Smith <
dexonsmith at apple.com> wrote:

> Usually just replying weekly with the text "ping" and the
> reattached patch is your best bet.
>
> I had a look, and the patch seems reasonable.  However, you
> should use `FileCheck` [1] instead of `grep` and `count`.
>
> [1]: http://llvm.org/docs/CommandGuide/FileCheck.html
>
> Also, I find the name `AllocateStack()` confusing in this
> context.  I think it would be better to extract out a helper
> called `AlignStack()` that your new code and `AllocateStack()`
> both call.
>
> Question: why aren't you calling `ensureMaxAlignment()` here?
>
> > On 2014-Aug-04, at 14:10, Thomas Jablin <tjablin at gmail.com> wrote:
> >
> > Is there anything I can do to accelerate the review of this patch? I
> originally submitted it to llvm-commits on June 24.
> > Tom
> >
> >
> > On Mon, Jul 28, 2014 at 8:34 AM, Thomas Jablin <tjablin at gmail.com>
> wrote:
> > Adding Elena since she most recently touched this function.
> >
> >
> > On Mon, Jul 28, 2014 at 8:15 AM, Thomas Jablin <tjablin at gmail.com>
> wrote:
> > This patch still requires a reviewer.
> >
> >
> > On Mon, Jul 14, 2014 at 4:07 PM, Thomas Jablin <tjablin at gmail.com>
> wrote:
> > I'm still waiting for a review. This is a patch for PR14792.
> > Tom
> >
> >
> > On Mon, Jun 30, 2014 at 12:34 PM, Thomas Jablin <tjablin at gmail.com>
> wrote:
> > I'm still waiting for a review. Maybe Nadav since he owns the x86
> backend?
> >
> >
> > On Tue, Jun 24, 2014 at 12:36 PM, Thomas Jablin <tjablin at gmail.com>
> wrote:
> > Hi All,
> > Please find enclosed an updated and simplified patch along with a
> additional test for properly aligned variadic arguments.
> > Tom
> >
> > On Fri, Jul 19, 2013 at 8:36 PM, Eli Friedman <eli.friedman at gmail.com>
> wrote:
> > On Thu, Jul 18, 2013 at 10:31 PM, Yun-Wei Lee <lee2041412 at gmail.com>
> wrote:
> > > Thank for your suggestion.
> > >
> > > I have tried to use GetAlignedArgumentStackSize.
> > > It is so strange that this function may output uncorrectly.
> > > For example, the StackSize will be 40 when handling the third function
> call
> > > in main.
> > > If I modified the this function, it will fail one test.
> > > I wonder this may be another bug in llvm......
> > > And that's why I didn't use GetAlignedArgumentStackSize previously.
> > >
> > > Thank you!
> >
> > I'm sorry, I'm not that familiar with the code.  A brief look tells me
> > that it's likely none of the patches are quite right;
> > http://llvm.org/bugs/attachment.cgi?id=10897 looks closer, but still
> > not quite right.
> >
> > -Eli
> >
> > >
> > > On Thu, Jul 18, 2013 at 7:56 PM, Eli Friedman <eli.friedman at gmail.com>
> > > wrote:
> > >>
> > >> On Thu, Jul 18, 2013 at 8:36 AM, Yun-Wei Lee <lee2041412 at gmail.com>
> wrote:
> > >> > http://llvm.org/bugs/show_bug.cgi?id=14792
> > >> >
> > >> > Problem:
> > >> >   In the i386 ABI Page 3-10, it said that the stack is aligned.
> However,
> > >> > the
> > >> > two example code show that does not handle the alignment correctly
> when
> > >> > using variadic function. For example, if the size of the first
> argument
> > >> > is
> > >> > 17, the overflow_arg_area in va_list will be set to "address of
> first
> > >> > argument + 16" instead of "address of first argument + 24" after
> calling
> > >> > va_start.
> > >> >   In addition, #6636 showed the same problem because in AMD64,
> arguments
> > >> > is
> > >> > passed by register at first, then pass by memory when run out of
> > >> > register
> > >> > (AMD64 ABI 3.5.7 rule 10).
> > >> >
> > >> > Why this problem happened?
> > >> >   When calling va_start to set va_list, overflow_arg_area is not set
> > >> > correctly. To set the overflow_arg_area correctly, we need to get
> the
> > >> > FrameIndex correctly. Now, here comes the problem, llvm doesn't
> handle
> > >> > it
> > >> > correctly. It accounts for StackSize to compute the FrameIndex, and
> if
> > >> > the
> > >> > StackSize is not aligned, it will compute the wrong FrameIndex. As a
> > >> > result
> > >> > overflow_arg_area will not be set correctly.
> > >> >
> > >> > My Solution:
> > >> > 1. Record the Align if it is located in Memory.
> > >> > 2. If it is variadic function and needs to set FrameIndex, adjust
> the
> > >> > stacksize.
> > >>
> > >> Please read http://llvm.org/docs/DeveloperPolicy.html .  In
> > >> particular, patches should be sent to llvm-commits, and patches should
> > >> generally include a regression test.
> > >>
> > >> In terms of the code, you might want to consider using
> > >> llvm::RoundUpToAlignment.
> > >>
> > >> -Eli
> > >
> > >
> > _______________________________________________
> > LLVM Developers mailing list
> > LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
> >
> >
> >
> >
> >
> >
> >
> > _______________________________________________
> > llvm-commits mailing list
> > llvm-commits at cs.uiuc.edu
> > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140818/7999b0d0/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: variadicAlignment.patch
Type: text/x-patch
Size: 2875 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140818/7999b0d0/attachment.bin>