[LLVMdev] New ARM disassembler - unpredictable instructions

[James Molloy]

Hi Owen,

 

I've got a load of examples for this, but I'll pick one.

 

echo '0xB0 0x00 0x00 0x00' |/work/llvm-cmake2/build/bin/llvm-mc -arch arm
--disassemble -

        strheq  r0, [r0], -r0

 

This instruction encoding is marked UNPREDICTABLE. The question is: does a
disassembly of it have value? One use case put to me was in a debugger,
where one needs to know if an instruction was ARM-mode or Thumb-mode so it
can correctly guess the mode of the previous instruction in the stream and
disassemble that correctly.

 

So I think that "return false" on all UNPREDICTABLEs is losing functionality
that could potentially be useful. Naturally you don't want to do the current
behaviour and have UNPREDICTABLEs erroneously decoded as valid instructions
(as it looks above, until you try to assemble it at which point an error is
raised).

 

Our internal disassembly mechanism prints the following:

 

        strheq  r0, [r0], -r0 ; ?

 

With the trailing "; ?" meaning that the instruction is unpredictable. It
also gives a hint that the disassembly cannot be correctly assembled -
there might be a (correct) assumption otherwise that assembly and
disassembly are the exact opposites of each other, and anything that is
successfully disassembled must be able to be assembled again (indeed, this
can be an extremely useful test for a disassembler/assembler combo such as
MC).

 

The question is whether or not a "soft fail" such as the above has value - I
think it does. What do you think?

 

Cheers,

 

James 

 

From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On
Behalf Of Owen Anderson
Sent: 11 August 2011 18:06
To: James Molloy
Cc: llvmdev at cs.uiuc.edu
Subject: Re: [LLVMdev] New ARM disassembler - unpredictable instructions

 

 

On Aug 11, 2011, at 1:32 AM, James Molloy wrote:





Hi,

 

I'm looking at the new ARM disassembler, and performing testing to see if
there are any obvious bugs I can fix. Overall it looks really nice - I've
only found one obvious bug that I'm packaging up a patch for now.

 

Glad you're liking it!  I haven't really done comprehensive testing yet-it
passes all the decoding tests that the previous disassembler passed, but I'm
sure there are many cases that it does not yet get right.  Patches welcome.





Currently it does not provide information as to whether a particular
bitpattern was UNPREDICTABLE or not. For example, an invalid bitpattern may
return correct disassembly for an alternative, predictable instruction
meaning that the sequence X -> disassemble -> assemble -> Y could end up
with X being different to Y.

 

Can you give an example of this?

 

--Owen
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20110812/41cbbe47/attachment.html>