elf_mem_map 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414
  1. #!/usr/bin/env python3
  2. import argparse
  3. import elftools.elf.elffile
  4. import elftools.dwarf.descriptions
  5. from collections import namedtuple
  6. from struct import unpack
  7. import re
  8. SRAM_OFFSET = 0x800000
  9. EEPROM_OFFSET = 0x810000
  10. FILL_BYTE = b'\0'
  11. Entry = namedtuple('Entry', ['name', 'loc', 'size', 'declpos'])
  12. Member = namedtuple('Member', ['name', 'off', 'size'])
  13. def array_inc(loc, dim, idx=0):
  14. if idx == len(dim):
  15. return True
  16. loc[idx] += 1
  17. if loc[idx] == dim[idx]:
  18. loc[idx] = 0
  19. return array_inc(loc, dim, idx+1)
  20. return False
  21. def get_type_size(type_DIE):
  22. while True:
  23. if 'DW_AT_byte_size' in type_DIE.attributes:
  24. return type_DIE, type_DIE.attributes.get('DW_AT_byte_size').value
  25. if 'DW_AT_type' not in type_DIE.attributes:
  26. return None
  27. type_DIE = type_DIE.get_DIE_from_attribute('DW_AT_type')
  28. def get_type_arrsize(type_DIE):
  29. size = get_type_size(type_DIE)
  30. if size is None:
  31. return None
  32. byte_size = size[1]
  33. if size[0].tag != 'DW_TAG_pointer_type':
  34. array_DIE = get_type_def(type_DIE, 'DW_TAG_array_type')
  35. if array_DIE is not None:
  36. for range_DIE in array_DIE.iter_children():
  37. if range_DIE.tag == 'DW_TAG_subrange_type' and \
  38. 'DW_AT_upper_bound' in range_DIE.attributes:
  39. dim = range_DIE.attributes['DW_AT_upper_bound'].value + 1
  40. byte_size *= dim
  41. return byte_size
  42. def get_type_def(type_DIE, type_tag):
  43. while True:
  44. if type_DIE.tag == type_tag:
  45. return type_DIE
  46. if 'DW_AT_type' not in type_DIE.attributes:
  47. return None
  48. type_DIE = type_DIE.get_DIE_from_attribute('DW_AT_type')
  49. def get_FORM_block1(attr):
  50. if attr.form != 'DW_FORM_block1':
  51. return None
  52. if attr.value[0] == 3: # OP_addr
  53. return int.from_bytes(attr.value[1:], 'little')
  54. if attr.value[0] == 35: # OP_plus_uconst (ULEB128)
  55. v = 0
  56. s = 0
  57. for b in attr.value[1:]:
  58. v |= (b & 0x7f) << s
  59. if b & 0x80 == 0:
  60. break
  61. s += 7
  62. return v
  63. return None
  64. def get_array_dims(DIE):
  65. array_DIE = get_type_def(DIE, 'DW_TAG_array_type')
  66. if array_DIE is None:
  67. return []
  68. array_dim = []
  69. for range_DIE in array_DIE.iter_children():
  70. if range_DIE.tag == 'DW_TAG_subrange_type' and \
  71. 'DW_AT_upper_bound' in range_DIE.attributes:
  72. array_dim.append(range_DIE.attributes['DW_AT_upper_bound'].value + 1)
  73. return array_dim
  74. def get_struct_members(DIE, entry, expand_structs, struct_gaps):
  75. struct_DIE = get_type_def(DIE, 'DW_TAG_structure_type')
  76. if struct_DIE is None:
  77. return []
  78. members = []
  79. for member_DIE in struct_DIE.iter_children():
  80. if member_DIE.tag == 'DW_TAG_member' and 'DW_AT_name' in member_DIE.attributes:
  81. m_name = member_DIE.attributes['DW_AT_name'].value.decode('ascii')
  82. m_off = get_FORM_block1(member_DIE.attributes['DW_AT_data_member_location'])
  83. m_byte_size = get_type_size(member_DIE)[1]
  84. # still expand member arrays
  85. m_array_dim = get_array_dims(member_DIE)
  86. if m_byte_size == 1 and len(m_array_dim) > 1:
  87. # likely string, remove one dimension
  88. m_byte_size *= m_array_dim.pop()
  89. if len(m_array_dim) == 0 or (len(m_array_dim) == 1 and m_array_dim[0] == 1):
  90. # plain entry
  91. members.append(Member(m_name, m_off, m_byte_size))
  92. elif len(m_array_dim) == 1 and m_byte_size == 1:
  93. # likely string, avoid expansion
  94. members.append(Member(m_name + '[]', m_off, m_array_dim[0]))
  95. else:
  96. # expand array entries
  97. m_array_pos = m_off
  98. m_array_loc = [0] * len(m_array_dim)
  99. while True:
  100. # location index
  101. sfx = ''
  102. for d in range(len(m_array_dim)):
  103. sfx += '[{}]'.format(str(m_array_loc[d]).rjust(len(str(m_array_dim[d]-1)), '0'))
  104. members.append(Member(m_name + sfx, m_array_pos, m_byte_size))
  105. # advance
  106. if array_inc(m_array_loc, m_array_dim):
  107. break
  108. m_array_pos += m_byte_size
  109. if struct_gaps and len(members):
  110. # fill gaps in the middle
  111. members = list(sorted(members, key=lambda x: x.off))
  112. last_end = 0
  113. for n in range(len(members)):
  114. member = members[n]
  115. if member.off > last_end:
  116. members.append(Member('*UNKNOWN*', last_end, member.off - last_end))
  117. last_end = member.off + member.size
  118. if struct_gaps and len(members):
  119. # fill gap at the end
  120. members = list(sorted(members, key=lambda x: x.off))
  121. last = members[-1]
  122. last_end = last.off + last.size
  123. if entry.size > last_end:
  124. members.append(Member('*UNKNOWN*', last_end, entry.size - last_end))
  125. return members
  126. def get_elf_globals(path, expand_structs, struct_gaps=True):
  127. fd = open(path, "rb")
  128. if fd is None:
  129. return
  130. elffile = elftools.elf.elffile.ELFFile(fd)
  131. if elffile is None or not elffile.has_dwarf_info():
  132. return
  133. # probably not needed, since we're decoding expressions manually
  134. elftools.dwarf.descriptions.set_global_machine_arch(elffile.get_machine_arch())
  135. dwarfinfo = elffile.get_dwarf_info()
  136. grefs = []
  137. for CU in dwarfinfo.iter_CUs():
  138. file_entries = dwarfinfo.line_program_for_CU(CU).header["file_entry"]
  139. for DIE in CU.iter_DIEs():
  140. # handle only variable types
  141. if DIE.tag != 'DW_TAG_variable':
  142. continue
  143. if 'DW_AT_location' not in DIE.attributes:
  144. continue
  145. if 'DW_AT_name' not in DIE.attributes and \
  146. 'DW_AT_abstract_origin' not in DIE.attributes:
  147. continue
  148. # handle locations encoded directly as DW_OP_addr (leaf globals)
  149. loc = get_FORM_block1(DIE.attributes['DW_AT_location'])
  150. if loc is None or loc < SRAM_OFFSET or loc >= EEPROM_OFFSET:
  151. continue
  152. loc -= SRAM_OFFSET
  153. # variable name/type
  154. if 'DW_AT_name' not in DIE.attributes and \
  155. 'DW_AT_abstract_origin' in DIE.attributes:
  156. DIE = DIE.get_DIE_from_attribute('DW_AT_abstract_origin')
  157. if 'DW_AT_location' in DIE.attributes:
  158. # duplicate reference (handled directly), skip
  159. continue
  160. if 'DW_AT_name' not in DIE.attributes:
  161. continue
  162. if 'DW_AT_type' not in DIE.attributes:
  163. continue
  164. name = DIE.attributes['DW_AT_name'].value.decode('ascii')
  165. # get final storage size
  166. size = get_type_size(DIE)
  167. if size is None:
  168. continue
  169. byte_size = size[1]
  170. # location of main definition
  171. declpos = ''
  172. if 'DW_AT_decl_file' in DIE.attributes and \
  173. 'DW_AT_decl_line' in DIE.attributes:
  174. line = DIE.attributes['DW_AT_decl_line'].value
  175. fname = DIE.attributes['DW_AT_decl_file'].value
  176. if fname and fname - 1 < len(file_entries):
  177. fname = file_entries[fname-1].name.decode('ascii')
  178. declpos = '{}:{}'.format(fname, line)
  179. # fetch array dimensions (if known)
  180. array_dim = get_array_dims(DIE)
  181. # fetch structure members (one level only)
  182. entry = Entry(name, loc, byte_size, declpos)
  183. if not expand_structs or size[0].tag == 'DW_TAG_pointer_type':
  184. members = []
  185. else:
  186. members = get_struct_members(DIE, entry, expand_structs, struct_gaps)
  187. def expand_members(entry, members):
  188. if len(members) == 0:
  189. grefs.append(entry)
  190. else:
  191. for member in members:
  192. grefs.append(Entry(entry.name + '.' + member.name,
  193. entry.loc + member.off, member.size,
  194. entry.declpos))
  195. if byte_size == 1 and len(array_dim) > 1:
  196. # likely string, remove one dimension
  197. byte_size *= array_dim.pop()
  198. if len(array_dim) == 0 or (len(array_dim) == 1 and array_dim[0] == 1):
  199. # plain entry
  200. expand_members(entry, members)
  201. elif len(array_dim) == 1 and byte_size == 1:
  202. # likely string, avoid expansion
  203. grefs.append(Entry(entry.name + '[]', entry.loc,
  204. array_dim[0], entry.declpos))
  205. else:
  206. # expand array entries
  207. array_pos = loc
  208. array_loc = [0] * len(array_dim)
  209. while True:
  210. # location index
  211. sfx = ''
  212. for d in range(len(array_dim)):
  213. sfx += '[{}]'.format(str(array_loc[d]).rjust(len(str(array_dim[d]-1)), '0'))
  214. expand_members(Entry(entry.name + sfx, array_pos,
  215. byte_size, entry.declpos), members)
  216. # advance
  217. if array_inc(array_loc, array_dim):
  218. break
  219. array_pos += byte_size
  220. return grefs
  221. def decode_dump(path):
  222. fd = open(path, 'r')
  223. if fd is None:
  224. return None
  225. buf_addr = None # starting address
  226. buf_data = None # data
  227. for line in fd:
  228. tokens = line.split(maxsplit=1)
  229. if len(tokens) == 0 or tokens[0] == 'ok':
  230. break
  231. elif len(tokens) < 2 or tokens[0] == 'D2':
  232. continue
  233. addr = int.from_bytes(bytes.fromhex(tokens[0]), 'big')
  234. data = bytes.fromhex(tokens[1])
  235. if buf_addr is None:
  236. buf_addr = addr
  237. buf_data = data
  238. else:
  239. # grow buffer as needed
  240. if addr < buf_addr:
  241. buf_data = FILL_BYTE * (buf_addr - addr)
  242. buf_addr = addr
  243. addr_end = addr + len(data)
  244. buf_end = buf_addr + len(buf_data)
  245. if addr_end > buf_end:
  246. buf_data += FILL_BYTE * (addr_end - buf_end)
  247. # replace new part
  248. rep_start = addr - buf_addr
  249. rep_end = rep_start + len(data)
  250. buf_data = buf_data[:rep_start] + data + buf_data[rep_end:]
  251. return (buf_addr, buf_data)
  252. def annotate_refs(grefs, addr, data, width, gaps=True, overlaps=True):
  253. last_end = None
  254. for entry in grefs:
  255. if entry.loc < addr:
  256. continue
  257. if entry.loc + entry.size > addr + len(data):
  258. continue
  259. pos = entry.loc-addr
  260. end_pos = pos + entry.size
  261. buf = data[pos:end_pos]
  262. buf_repr = ''
  263. if len(buf) in [1, 2, 4]:
  264. # attempt to decode as integers
  265. buf_repr += ' I:' + str(int.from_bytes(buf, 'little')).rjust(10)
  266. if len(buf) in [4, 8]:
  267. # attempt to decode as floats
  268. typ = 'f' if len(buf) == 4 else 'd'
  269. buf_repr += ' F:' + '{:10.3f}'.format(unpack(typ, buf)[0])
  270. if last_end is not None:
  271. if gaps and last_end < pos:
  272. # decode gaps
  273. gap_size = pos - last_end
  274. gap_buf = data[last_end:pos]
  275. print('{:04x} {} {:4} R:{}'.format(addr+last_end, "*UNKNOWN*".ljust(width),
  276. gap_size, gap_buf.hex()))
  277. if overlaps and last_end > pos + 1:
  278. gap_size = pos - last_end
  279. print('{:04x} {} {:4}'.format(addr+last_end, "*OVERLAP*".ljust(width), gap_size))
  280. print('{:04x} {} {:4}{} R:{}'.format(entry.loc, entry.name.ljust(width),
  281. entry.size, buf_repr, buf.hex()))
  282. last_end = end_pos
  283. def print_map(grefs):
  284. print('OFFSET\tSIZE\tNAME\tDECLPOS')
  285. for entry in grefs:
  286. print('{:x}\t{}\t{}\t{}'.format(entry.loc, entry.size, entry.name, entry.declpos))
  287. def print_qdirstat(grefs):
  288. print('[qdirstat 1.0 cache file]')
  289. entries = {}
  290. for entry in grefs:
  291. paths = list(filter(None, re.split(r'[\[\].]', entry.name)))
  292. base = entries
  293. for i in range(len(paths) - 1):
  294. name = paths[i]
  295. if name not in base:
  296. base[name] = {}
  297. base = base[name]
  298. name = paths[-1]
  299. base[name] = entry.size
  300. def walker(root, prefix):
  301. files = []
  302. dirs = []
  303. for name, entries in root.items():
  304. if type(entries) == int:
  305. files.append([name, entries])
  306. else:
  307. dirs.append([name, entries])
  308. # print files
  309. print('D\t{}\t{}\t0x0'.format(prefix, 0))
  310. for name, size in files:
  311. print('F\t{}\t{}\t0x0'.format(name, size))
  312. # recurse directories
  313. for name, entries in dirs:
  314. walker(entries, prefix + '/' + name)
  315. walker(entries, '/')
  316. def main():
  317. ap = argparse.ArgumentParser(description="""
  318. Generate a symbol table map starting directly from an ELF
  319. firmware with DWARF3 debugging information.
  320. When used along with a memory dump obtained from the D2 g-code,
  321. show the value of each symbol which is within the address range.
  322. """)
  323. ap.add_argument('elf', help='ELF file containing DWARF debugging information')
  324. ap.add_argument('--no-gaps', action='store_true',
  325. help='do not dump memory inbetween known symbols')
  326. ap.add_argument('--no-expand-structs', action='store_true',
  327. help='do not decode structure data')
  328. ap.add_argument('--overlaps', action='store_true',
  329. help='annotate overlaps greater than 1 byte')
  330. ap.add_argument('--name-width', type=int, default=50,
  331. help='set name column width')
  332. g = ap.add_mutually_exclusive_group(required=True)
  333. g.add_argument('dump', nargs='?', help='RAM dump obtained from D2 g-code')
  334. g.add_argument('--map', action='store_true', help='dump global memory map')
  335. g.add_argument('--qdirstat', action='store_true',
  336. help='dump qdirstat-compatible size usage map')
  337. args = ap.parse_args()
  338. grefs = get_elf_globals(args.elf, expand_structs=not args.no_expand_structs)
  339. grefs = list(sorted(grefs, key=lambda x: x.loc))
  340. if args.map:
  341. print_map(grefs)
  342. elif args.qdirstat:
  343. print_qdirstat(grefs)
  344. else:
  345. addr, data = decode_dump(args.dump)
  346. annotate_refs(grefs, addr, data,
  347. width=args.name_width,
  348. gaps=not args.no_gaps,
  349. overlaps=args.overlaps)
  350. if __name__ == '__main__':
  351. exit(main())