macaw/base
Tristan Ravitch 6a4f406c68 Revisit handling of tail calls
It turns out that we have to be more conservative with tail call identification,
as incorrectly identifying a block as the target of a tail call (instead of a
branch) can cause other branch classifiers to fail if that block is the target
of another jump.

Ultimately, we will need to give up some tail call recognition (since they are
in general indistinguishable from jumps), and instead only identify known call
targets as tail call candidates.

With additional global analysis we could do better.

Fixes #294
2022-06-27 15:02:43 -07:00
..
src/Data/Macaw Revisit handling of tail calls 2022-06-27 15:02:43 -07:00
ChangeLog.md Adapt to dynSymEntry being definition-aware in elf-edit 2022-04-19 12:47:24 -04:00
LICENSE Update license information. 2017-09-27 15:59:06 -07:00
macaw-base.cabal Drop support for GHC 8.6 2022-01-10 16:40:23 -05:00
README.rst Haddock and README fixes. 2019-01-08 16:38:38 -08:00

The macaw library implements architecture-independent binary code
discovery.  Support for specific architectures is provided by
implementing the semantics of that architecture.  The library is
written in terms of an abstract interface to memory, for which an ELF
backend is provided (via the elf-edit_ library).  The basic code
discovery is based on a variant of Value Set Analysis (VSA).

The most important user-facing abstractions are:

* The ``Memory`` type, defined in ``Data.Macaw.Memory``, which provides an abstract interface to an address space containing both code and data.
* The ``memoryForElfSegments`` function is a useful helper to produce a ``Memory`` from an ELF file.
* The ``cfgFromAddrs`` function, defined in ``Data.Macaw.Discovery``, which performs code discovery on a ``Memory`` given some initial parameters (semantics to use via ``ArchitectureInfo`` and some entry points).
* The ``DiscoveryInfo`` type, which is the result of ``cfgFromAddrs``; it contains a collection of ``DiscoveryFunInfo`` records, each of which represents a discovered function.  Every basic block is assigned to at least one function.

Architecture-specific code goes into separate libraries.  X86-specific code is in the macaw-x86 repo.

An abbreviated example of using macaw on an X86_64 ELF file looks like::

  import qualified Data.Map as M
  import qualified Data.ElfEdit as E
  import qualified Data.Parameterized.Some as PU
  import qualified Data.Macaw.X86 as MX86
  import qualified Data.Macaw.Memory.ElfLoader as ML
  import qualified Data.Macaw.Discovery as MD

  discoverCode :: E.Elf Word64 -> (forall ids . MD.DiscoveryInfo X86_64 ids -> a) -> a
  discoverCode elf k =
    case ML.resolveElfContents ML.defaultLoadOptions elf of
      Left e -> error (show e)
      Right (_, _, Nothing, _) -> error "Unable to determine entry point"
      Right (warn, mem, Just entryPoint, _) -> do
        mapM_ print warn
        case MD.cfgFromAddrs MX86.x86_64_linux_info mem M.empty [entryPoint] [] of
        PU.Some di -> k di


In the callback ``k``, the ``DiscoveryInfo`` can be analyzed as desired.

Implementing support for an architecture is more involved and requires implementing an ``ArchitectureInfo``, which is defined in ``Data.Macaw.Architecture.Info``.  This structure contains architecture-specific information like:

* The pointer width
* A disassembler from bytes to abstract instructions
* ABI information regarding registers and calling conventions
* A transfer function for architecture-specific features not represented in the common IR

.. _elf-edit: https://github.com/GaloisInc/elf-edit
.. _flexdis86: https://github.com/GaloisInc/flexdis86