llvm dev - Jan 2015 - [LLVMdev] DataFlowSanitizer using wrong memory layout

Hi all,

Any one tried using DataFlowSanitizer on Linux x86_64? 

I tried on:
 3.13.0-44-generic #73~precise1-Ubuntu SMP Wed Dec 17 00:39:15 UTC 2014 x86_64
x86_64 x86_64 GNU/Linux

It assumes wrong memory layout and remaps application code segment as shadow
memory, thus causing SIGSEV (Segmentation fault).


Is this know? and fix under way?

-Thanks,
Aravind
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150115/0ca8b6b0/attachment.html>

On Thu, Jan 15, 2015 at 10:58:42AM -0800, machiry aravind
wrote:
> Hi all,
> 
> Any one tried using DataFlowSanitizer on Linux x86_64? 
DFSan is developed and tested regularly on Linux x86_64.
> I tried on:
>  3.13.0-44-generic #73~precise1-Ubuntu SMP Wed Dec 17 00:39:15 UTC 2014
x86_64 x86_64 x86_64 GNU/Linux
> 
> It assumes wrong memory layout and remaps application code segment as
shadow memory, thus causing SIGSEV (Segmentation fault).
> 
> 
> Is this know? and fix under way?
I've found this to occur if the application is not built as a
position-independent executable, but we've explicitly taught the Clang
driver to produce position-independent executables if DFSan is enabled,
so it isn't clear why this might be happening to you.

How exactly are you building your application? Does a simple C hello world
program work? Do the tests (make check-dfsan) pass?

Thanks,
-- 
Peter

ah!

I has other passes to run and missed enabling PIC.

Thanks, this works now.

-Aravind
> Date: Thu, 15 Jan 2015 11:34:33 -0800
> From: peter at pcc.me.uk
> To: machiry_msidc at hotmail.com
> CC: llvmdev at cs.uiuc.edu
> Subject: Re: [LLVMdev] DataFlowSanitizer using wrong memory layout
> 
> On Thu, Jan 15, 2015 at 10:58:42AM -0800, machiry aravind wrote:
> > Hi all,
> > 
> > Any one tried using DataFlowSanitizer on Linux x86_64? 
> 
> DFSan is developed and tested regularly on Linux x86_64.
> 
> > I tried on:
> >  3.13.0-44-generic #73~precise1-Ubuntu SMP Wed Dec 17 00:39:15 UTC
2014 x86_64 x86_64 x86_64 GNU/Linux
> > 
> > It assumes wrong memory layout and remaps application code segment as
shadow memory, thus causing SIGSEV (Segmentation fault).
> > 
> > 
> > Is this know? and fix under way?
> 
> I've found this to occur if the application is not built as a
> position-independent executable, but we've explicitly taught the Clang
> driver to produce position-independent executables if DFSan is enabled,
> so it isn't clear why this might be happening to you.
> 
> How exactly are you building your application? Does a simple C hello world
> program work? Do the tests (make check-dfsan) pass?
> 
> Thanks,
> -- 
> Peter
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150115/2e8e8b19/attachment.html>