Potential fix for #2: Treating unknown data as THUMB code, not bytes.

Author: aarant

README.md

@@ -19,6 +19,7 @@ Luvdis is a tool for disassembling GBA ROMs. Features include:
    - [From Releases](#from-releases)
    - [From latest source](#from-latest-source)
 - [Usage](#usage)
+  - [FAQ](#faq)
   - [Options](#options)
   - [ROM detection](#rom-detection)
 
@@ -73,6 +74,17 @@ To disassemble only part of a ROM, say, up to the start of read-only data, provi
 $ luvdis rom.gba --start 0x0800024C --stop 0x0x81b32b4 -o rom.s
 ```
 
+### FAQ
+
+#### How can I get rid of large blocks of raw bytes in the disassembly?
+
+By default, Luvdis treats areas of a ROM that it can't determine are executable as byte data. You can change this behavior
+with the `default_mode` option:
+
+```sh
+$ luvdis rom.gba --default_mode THUMB -o rom.s
+```
+
 ### Options
 
 ```
@@ -105,6 +117,9 @@ Options:
   --min-length INTEGER RANGE  Minimum valid instruction length required in
                               order to 'guess' a function. Must be at least 1,
                               defaults to 3.
+  --default-mode [THUMB|BYTE|WORD]
+                              Default disassembly mode when the nature of
+                              an address is unknown. Defaults to 'BYTE'.
   --help                      Show this message and exit.
 ```
 

luvdis/__init__.py

@@ -2,6 +2,6 @@
 
 Copyright (C) 2020 A. Antonitis. Licensed under the MIT license.
 """
-__version__ = '0.6.1'
+__version__ = '0.7.0'
 __doc__ = __doc__.replace('__version__', __version__)
 url = __url__ = 'https://github.com/arantonitis/luvdis'

luvdis/__main__.py

@@ -7,7 +7,7 @@
 from luvdis.config import read_config
 from luvdis.common import eprint, set_debug
 from luvdis.rom import ROM
-from luvdis.analyze import State, BASE_ADDRESS, END_ADDRESS
+from luvdis.analyze import State, BASE_ADDRESS, END_ADDRESS, THUMB, BYTE, WORD
 
 
 class AddressInt(click.ParamType):
@@ -31,6 +31,7 @@ def convert(self, value, param, ctx):
 
 
 ADDRESS_INT = AddressInt()
+MODE_MAP = {'byte': BYTE, 'thumb': THUMB, 'word': WORD}
 
 
 @click.group(cls=DefaultGroup, default='disasm', default_if_no_args=True)
@@ -47,7 +48,7 @@ def main():
                    'output path.')
 @click.option('-c', '--config', type=click.Path(exists=True, dir_okay=False, readable=True),
               help='Function configuration file.')
-@click.option('-co', '--config-out', 'config_out', type=click.Path(writable=True, dir_okay=False),
+@click.option('-co', '--config-out', type=click.Path(writable=True, dir_okay=False),
               help="Output configuration. If any functions are 'guessed' by Luvdis, they will appear here.")
 @click.option('-D', '--debug', is_flag=True, help='Turn on/off debugging behavior.')
 @click.option('--start', type=ADDRESS_INT, default=BASE_ADDRESS,
@@ -58,15 +59,18 @@ def main():
               help="Assembler macro file to '.include' in disassembly. If not specified, default macros are embedded.")
 @click.option('--guess/--no-guess', default=True,
               help='Turn on/off function guessing & discovery. Default is to perform guessing.')
-@click.option('--min-calls', 'min_calls', type=click.IntRange(1), default=2,
+@click.option('--min-calls', type=click.IntRange(1), default=2,
               help="Minimum number of calls to a function required in order to 'guess' it. Must be at least 1, "
                    "defaults to 2.")
-@click.option('--min-length', 'min_length', type=click.IntRange(1), default=3,
+@click.option('--min-length', type=click.IntRange(1), default=3,
               help="Minimum valid instruction length required in order to 'guess' a function. Must be at least 1, "
                    "defaults to 3.")
-def disasm(rom, output, config, config_out, debug, start, stop, macros, guess, min_calls, min_length, **kwargs):
+@click.option('--default-mode', type=click.Choice(('THUMB', 'BYTE', 'WORD'), case_sensitive=False), default='BYTE',
+              help="Default disassembly mode when the nature of an address is unknown. Defaults to 'BYTE'.")
+def disasm(rom, output, config, config_out, debug, start, stop, macros, guess, min_calls, min_length, default_mode,
+           **kw):
     """ Analyze and disassemble a GBA ROM. """
-    for k, v in kwargs.items():
+    for k, v in kw.items():
         print(k, v)
     set_debug(debug)
     functions = read_config(config) if config else None
@@ -76,7 +80,7 @@ def disasm(rom, output, config, config_out, debug, start, stop, macros, guess, m
     if output in (None, '-'):
         output = None
         eprint(f'No output file specified. Printing to stdout.')
-    state.dump(rom, output, config_out)
+    state.dump(rom, output, config_out, default_mode=MODE_MAP[default_mode.lower()])
 
 
 @main.command(name='info')

luvdis/analyze.py

@@ -382,7 +382,7 @@ def __init__(self, functions=None, min_calls=2, min_length=3, start=BASE_ADDRESS
                 else:
                     name = value
                 self.unexpanded[addr] = name
-        self.functions = {}  # addr -> (name, end)
+        self.functions = {}  # addr -> (name, end_address)
         self.not_funcs = set()
         self.min_calls, self.min_length, self.start, self.stop = min_calls, min_length, start, stop
         self.macros = macros
@@ -468,7 +468,7 @@ def analyze_func(self, rom, addr, state=None):
                         exit_behaved = False
                         end = rom.size | 0x08000000 if ins is None else ins.address
                         break
-                    elif ins.id == Opcode.ldr:  # Mark target as WORD
+                    elif ins.id == Opcode.ldr:  # Mark load target as WORD
                         end = addr = ins.address+2
                         target = ins.target
                         if target < self.stop:  # TODO: Is this necessary?
@@ -559,25 +559,27 @@ def label_for(self, addr):
                 return name
         return f'_{addr:08X}'
 
-    def dump(self, rom, path=None, config_output=None):
-        if config_output:  # Optionally, write updated function list
+    def dump(self, rom, path=None, config_output=None, default_mode=BYTE):
+        if config_output:  # Optionally write updated function list
             addr_map = {addr: (name, self.module_addrs.get(addr, None)) for addr, (name, _) in self.functions.items()}
             write_config(addr_map, config_output)
         # Setup initial module & file
         folder, module = os.path.split(path) if path else (None, None)
-        if DEBUG and path:  # Output function range info if debugging
+        if DEBUG and path:
             import pickle
+            # Output function range info if debugging
             with open(os.path.join(folder, 'funcs.pickle'), 'wb') as f:
                 pickle.dump(self.debug_ranges, f)
+            # Also output a linker script
             fl = open('luvdis.ld', 'w')
         f = None if path else sys.stdout
         # Setup start and end addresses
         addr = self.start
-        if type(self.stop) is float:  # End is the final address in the ROM
+        if type(self.stop) is float:  # End at the final address in the ROM
             end = rom.size | BASE_ADDRESS
         else:
             end = min(rom.size, self.stop & 0xffffff) | BASE_ADDRESS
-        if addr not in self.module_addrs and module:  # Set module of initial address to the path output
+        if addr not in self.module_addrs and module:  # Mark the very first address as belonging to the initial module
             self.module_addrs[addr] = module
         mode, flags, bytecount = BYTE, 0, 0
         # Initialize progress bar & messages
@@ -594,8 +596,8 @@ def warn(*args):
             old_mode = mode
 
             # Switch output modes
-            if addr_flags == 0 and flags != 0:  # Switch to byte mode when address flags are zero
-                mode = BYTE
+            if addr_flags == 0 and flags != 0:  # Switch to default mode when address flags are zero
+                mode = default_mode  # By default, BYTE mode
             elif addr_flags & FLAG_EXEC and not (flags & FLAG_EXEC):  # Output code
                 mode = THUMB
             elif addr_flags & FLAG_WORD and not (flags & FLAG_WORD) and not (addr_flags & FLAG_EXEC):  # Output words
@@ -638,7 +640,7 @@ def warn(*args):
             # Switch module output
             if f is not sys.stdout and addr in self.module_addrs:  # Address has module info
                 new_module = self.module_addrs[addr]
-                if new_module != module or f is None:  # New/first module seen
+                if new_module != module or f is None:  # Entering new/first module
                     module = new_module
                     path = os.path.join(folder, module)
                     eprint(f"{addr:08X}: module '{path}'")
@@ -651,12 +653,12 @@ def warn(*args):
                     f = open(path, 'w', buffering=1)
                     f.write(ASM_PRELUDE)
                     f.write(f'.include "{self.macros}"\n' if self.macros else MACROS)
-                    bytecount = 0  # Reset byte bytecount
-                    if DEBUG:  # Output link script if debugging
+                    bytecount = 0  # Reset bytecount
+                    if DEBUG:  # Output linker script if debugging
                         fl.write(f'{path[:-2]}.o(.text);\n')
 
             # Emit code or data
-            if mode == THUMB:
+            if mode == THUMB:  # THUMB code
                 offset = ins.size
                 if ins.id == Opcode.bl or ins.id in BRANCHES:
                     target = ins.target
@@ -672,7 +674,7 @@ def warn(*args):
                             emit = f'.2byte 0x{i:04X} @ {ins.mnemonic} _{target:08X}'
                 elif ins.id == Opcode.bx:
                     value = rom.read(addr, 2)
-                    # Assembler cannot emit bx with nonzero rd, see THUMB.5 TODO: Should these be illegal?
+                    # Assembler will not emit bx with nonzero rd, see THUMB.5 TODO: Should these be treated as illegal?
                     emit = f'.inst 0x{value:04X}' if value & 3 != 0 else str(ins)
                 elif ins.id == Opcode.ldr and isinstance(ins, Thumb6):  # Convert PC-relative loads into labels
                     target = ins.target
@@ -700,10 +702,7 @@ def warn(*args):
                     value = self.label_for(value-1)
                 else:
                     value = f'0x{value:08X}'
-                if label:
-                    emit = f'{label} .4byte {value}'
-                else:
-                    emit = f'\t.4byte {value}'
+                emit = f'{label} .4byte {value}' if label else f'\t.4byte {value}'
                 if DEBUG:
                     comment += f' @ {addr_flags}'
                 f.write(f'{emit}{comment}\n')
@@ -729,7 +728,7 @@ def warn(*args):
             flags = addr_flags
             addr += offset
             bar.update(offset)
-        # Close current module
+        # Done with output; close file handles and cleanup
         if f is not sys.stdout and f:
             if bytecount:
                 f.write('\n')