adds an option to directly use ogre files as a loader (BinaryAnalysis…

…Platform#1392) The feature is implemented directly in the `Image.create` function so that it can be used programmatically. The `backend` parameter can now accept an explicit file path instead of the loader name. The documentation of the `loader` command-line parameter is accordingly updated. To prevent clashes between existing loaders and filenames, only existing filenames with explicit paths are treated as ogre specifications, i.e., the path must start with `./`, `../`, or `/` in Unix. Example of usage, ``` $ bap testsuite/bin/arm-linux-gnueabi-echo -dogre:echo.ogre -dbir | grep print_endline 000007da: sub print_endline(print_endline_result) 00000804: print_endline_result :: out u32 = R0 0000031e: call @print_endline with return %00000320 $ sed 's/print_endline/printl/g' echo.ogre > echo.mangled.ogre $ bap testsuite/bin/arm-linux-gnueabi-echo --loader=./echo.mangled.ogre -dbir | grep printl 00000339: call @printl with return %0000033b 000007df: sub printl(printl_result) 00000807: printl_result :: out u32 = R0 ``` You can also use `bap specification ./exe` to obtain the initial OGRE specification for a file (without disassembling it) and the edit it to your taste. And, of course, you can write a specification from scratch.
thestr4ng3r · Jan 6, 2022 · 6af0e78 · 6af0e78
1 parent 5be8fef
commit 6af0e78
Show file tree

Hide file tree

Showing 3 changed files with 43 additions and 12 deletions.
diff --git a/lib/bap/bap.mli b/lib/bap/bap.mli
@@ -5506,10 +5506,20 @@ module Std : sig
     type result = (t * Error.t list) Or_error.t
 
     (** [create ?backend filename] creates an image of the file specified
-        specified by the [filename]. If [backend] is equal to "auto", then
-        all backends are tried in order. If only one backend can read this
-        file (i.e., there is no ambiguity), then image is returned. If
-        [backend] is not specified, then the LLVM backend is used. *)
+        by the [filename]. If [backend] is not specified, then
+        all availabe backends are used and their information is
+        merged. If the information provided by all backends agree (i.e.,
+        there's no conflicting information), then image is returned.
+        If [backend] is an explicit file path, then it is read as an
+        OGRE file and used for loading. Otherwise, [backend] should be
+        a name of one of the backends registered either with
+        [register_backend] or [register_loader]. See
+        [available_backends] for the list of available backends.
+
+        @since 2.5.0 accepts backend accepts an explicit file path,
+        note a file path is explicit if it exists and
+        [Fn.non Filename.is_implicit].
+    *)
     val create : ?backend:string -> path -> result
 
     (** [of_string ?backend ~data] creates an image from the specified
@@ -9084,7 +9094,7 @@ module Std : sig
         is translated to,
         {v
            #11 := #9 + 13
-           #12 := 11 * #11 
+           #12 := 11 * #11
            #10 := #12 - 17
         v}
         @since 2.5.0 *)

diff --git a/lib/bap_image/bap_image.ml b/lib/bap_image/bap_image.ml
@@ -664,15 +664,24 @@ let merge_loaders () : loader =
 
 let get_loader = function
   | None -> merge_loaders ()
+  | Some name when Sys.file_exists name &&
+                   Fn.non Filename.is_implicit name ->
+    let local _ = match Ogre.Doc.from_file name with
+      | Error err -> Error err
+      | Ok doc -> Ok (Some doc) in
+    (module struct
+      let from_file = local
+      let from_data = local
+    end)
   | Some name -> match Hashtbl.find backends name with
     | Some loader -> loader
     | None ->
-      let module Failure = struct
-        let fail _ = Result.failf "Unknown image loader %s" name ()
+      let fail _ = Result.failf "Unknown image loader %s" name () in
+      (module struct
         let from_file = fail
         let from_data = fail
-      end in
-      (module Failure)
+      end)
+
 
 let invoke load data arg = match load arg with
   | Ok (Some doc) -> from_spec (data arg) doc

diff --git a/plugins/disassemble/disassemble_main.ml b/plugins/disassemble/disassemble_main.ml
@@ -243,7 +243,18 @@ let input = Extension.Command.argument
 let loader =
   Extension.Command.parameter
     ~doc:"Use the specified loader.
-          Use the loader `raw' to load unstructured files"
+          The loader could be either an identifier or a filename. \
+          The filename has to be explicit, i.e., to start with an \
+          explicit reference to the root directory or to the current \
+          directory (e.g., $(b,./), or $(b,../), or $(b,/) in Unix). \
+          The contents of the file should be a well-formed OGRE \
+          document that contains the necessary meta-information \
+          about the binary. \
+          The default loader is named $(b,llvm) and uses LLVM loaders \
+          to parse the input binary and supports ELF, MachO, and COFF \
+          (including Windows PE), formats. \
+          To load unstructured files use the $(b,raw) loader and \
+          specify the loader parameters via the $(b,raw) plugin."
     Extension.Type.(string =? "llvm")
     "loader"
 
@@ -362,13 +373,14 @@ let save_knowledge ~had_knowledge ~update digest = function
   | Some _ -> ()
 
 
-
 let create_and_process input outputs passes loader target update kb ctxt =
+  let uses_file_loader = Sys.file_exists loader &&
+                         Fn.non Filename.is_implicit loader in
   let package = input in
   let digest = make_digest [
       Extension.Configuration.digest ctxt;
       Caml.Digest.file input;
-      loader;
+      if uses_file_loader then Caml.Digest.file loader else loader;
     ] in
   let had_knowledge = load_knowledge digest kb in
   let input = Project.Input.load ~target ~loader input in