Merge branch 'code-pe-export-are-functions' into 'main'

PE exports that correspond to code should be function entries

See merge request rewriting/ddisasm!1171

CHANGELOG.md

@@ -5,6 +5,7 @@
 * Better data access computation for MIPS binaries.
 * Detect incremental linking regions in PE binaries.
 * Create elfStackSize and elfStackExec auxdata from ELF PT_GNU_STACK segments.
+* In PE binaries, every exported code symbol is considered a function entry.
 * Requires gtirb >=1.12.1, gtirb-pprinter >=2.0.0
 
 # 1.7.0

examples/asm_examples/ex_dll_export_thunk/Makefile.windows

@@ -0,0 +1,5 @@
+all:
+	ml64.exe /c ex.asm
+	link.exe ex.obj /DLL /SUBSYSTEM:WINDOWS
+clean:
+	del /F /Q *.dll *.obj

examples/asm_examples/ex_dll_export_thunk/ex.asm

@@ -0,0 +1,33 @@
+INCLUDELIB libcmt
+
+EXTERN __ImageBase:BYTE
+
+EXTRN puts:PROC
+
+.CODE
+
+
+print_ok1 PROC EXPORT
+                JMP print_ok
+print_ok1 ENDP
+
+print_ok2 PROC EXPORT
+                JMP print_ok
+print_ok2 ENDP
+
+print_ok3 PROC EXPORT
+                JMP print_ok
+print_ok3 ENDP
+
+
+print_ok:
+                LEA   RCX, OFFSET $OK
+                CALL  puts
+
+exit:
+                RET 0
+.DATA
+$OK DB	"ok", 00H
+
+
+END

src/datalog/basic_function_inference.dl

@@ -52,6 +52,7 @@ function_entry_initial(EA):-
         function_symbol(EA,_);
         start_function(EA);
         loaded_section(EA,_,_);
+        pe_export_entry(EA,_,_);
         // TODO: Fold pc_load_call into generic arch component.
         direct_call(EA_caller,EA), code(EA_caller), !pc_load_call(EA_caller,EA)
     ).

tests/function_inference_test.py

@@ -1,11 +1,14 @@
 import platform
 import unittest
-from disassemble_reassemble_check import compile, disassemble, cd
+from disassemble_reassemble_check import compile, disassemble, cd, make
 import gtirb
+import os
+import subprocess
 
 from pathlib import Path
 
 ex_dir = Path("./examples/")
+ex_asm_dir = ex_dir / "asm_examples"
 
 
 class TestFunctionInference(unittest.TestCase):
@@ -386,5 +389,30 @@ def test_functions_virtualDispatch_clang(self):
         )
 
 
+class PEFunctionInferenceTests(unittest.TestCase):
+    @unittest.skipUnless(
+        platform.system() == "Windows"
+        and os.environ.get("VSCMD_ARG_TGT_ARCH") == "x64",
+        "This test is Windows (x64) only.",
+    )
+    def test_code_exports_are_functions(self):
+        """
+        Test that any code that is exported is
+        considered as a function entry.
+        """
+        with cd(ex_asm_dir / "ex_dll_export_thunk"):
+            subprocess.run(make("all"), stdout=subprocess.DEVNULL)
+            self.assertTrue(disassemble("ex.dll", format="--ir")[0])
+            ir = gtirb.IR.load_protobuf("ex.dll.gtirb")
+            module = ir.modules[0]
+            functionNames = {
+                sym.name
+                for sym in module.aux_data["functionNames"].data.values()
+            }
+            self.assertIn("print_ok1", functionNames)
+            self.assertIn("print_ok2", functionNames)
+            self.assertIn("print_ok3", functionNames)
+
+
 if __name__ == "__main__":
     unittest.main()