Re: [DynInst_API:] Cross-Platform Binary Analysis

[Bill Williams <bill@xxxxxxxxxxx>]

On 09/30/2013 07:55 AM, Stefan Stattelmann wrote:
> Hi Bill,
>
> compiling with $PLATFORM set to ppc32_linux did the trick.
>
> Another issue I ran into was that bare-metal PowerPC code does not seem
> to work. As far as I could reconstruct the problem, loading the ELF data
> section produces an error somewhere in Object-elf.C.
> No problem with PowerPC Linux binaries though.
Glad to hear it. What are you defining as bare-metal PowerPC code? File 
full of assembled instructions, no headers etc?
I've puttered around in the past with a mmapped CodeSource 
implementation for ParseAPI that would bypass symtab/symlite and just 
treat a file/chunk of memory as a code source; it's pretty trivial but 
I've never polished it up for public use. If you're looking for 
bare-metal analysis, that's the route I'd take...I can try to dig up 
some of my old code if you need pointers/a skeleton.
> Thanks for your help.
>
> Stefan
>
>
>
> From: Bill Williams <bill@xxxxxxxxxxx>
> To: dyninst-api@xxxxxxxxxxx,
> Date: 24.09.2013 17:28
> Subject: Re: [DynInst_API:] Cross-Platform Binary Analysis
> Sent by: "Dyninst-api" <dyninst-api-bounces@xxxxxxxxxxx>
> ------------------------------------------------------------------------
>
>
>
> On 09/24/2013 07:34 AM, Stefan Stattelmann wrote:
>  > Hi,
>  >
>  > I would like to use Dyninst for cross-platform binary analysis, e.g.,
>  > analyzing PowerPC code on a x86 machine. Is this currently supported and
>  > if so, how? From my experiments so far it seems like binaries for a
>  > different CPU architectures are not recognized automatically and thus
>  > cannot be decoded.
>  >
>  > Thanks and best regards,
>  > Stefan
>  >
> Stefan--
>
> The support for this is limited, but if you're looking for binary
> analysis and not instrumentation, it shouldn't be too hard to get it
> working. I believe, as a first approximation, that if you build the
> analysis components (symtab, instructionapi, parseapi, dataflowapi) with
> your $PLATFORM set to your target architecture, that should give you
> components that can analyze cross-architecture binaries (but not native
> ones). (Note: it is likely that there will be some kinks to work out
> there; if/when you run into trouble, please let us know.)
>
> Improving this support has long been on our TODO list; the right thing
> is to get rid of any lingering per-architecture conditional compilation
> and properly abstract the platform-specific bits into classes (with
> machinery to auto-detect, as you note). This could happen for 9.0 but
> almost certainly not for 8.2.
>
>  >
>  > _______________________________________________
>  > Dyninst-api mailing list
>  > Dyninst-api@xxxxxxxxxxx
>  > https://lists.cs.wisc.edu/mailman/listinfo/dyninst-api
>  >
>
>
> --
> --bw
>
> Bill Williams
> Paradyn Project
> bill@xxxxxxxxxxx
> _______________________________________________
> Dyninst-api mailing list
> Dyninst-api@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/dyninst-api
>
>
>
> _______________________________________________
> Dyninst-api mailing list
> Dyninst-api@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/dyninst-api

--
--bw

Bill Williams
Paradyn Project
bill@xxxxxxxxxxx