2 # Merge or print the coverage data collected by asan's coverage.
3 # Input files are sequences of 4-byte integers.
4 # We need to merge these integers into a set and then
5 # either print them (as hex) or dump them into another file.
17 print >> sys.stderr, "Usage: \n" + \
18 " " + prog_name + " merge FILE [FILE...] > OUTPUT\n" \
19 " " + prog_name + " print FILE [FILE...]\n" \
20 " " + prog_name + " unpack FILE [FILE...]\n" \
21 " " + prog_name + " rawunpack FILE [FILE ...]\n" \
22 " " + prog_name + " missing BINARY < LIST_OF_PCS\n"
26 if bits != 32 and bits != 64:
27 raise Exception("Wrong bitness: %d" % bits)
29 def TypeCodeForBits(bits):
31 return 'L' if bits == 64 else 'I'
33 kMagic32SecondHalf = 0xFFFFFF32;
34 kMagic64SecondHalf = 0xFFFFFF64;
35 kMagicFirstHalf = 0xC0BFFFFF;
37 def MagicForBits(bits):
39 if sys.byteorder == 'little':
40 return [kMagic64SecondHalf if bits == 64 else kMagic32SecondHalf, kMagicFirstHalf]
42 return [kMagicFirstHalf, kMagic64SecondHalf if bits == 64 else kMagic32SecondHalf]
44 def ReadMagicAndReturnBitness(f, path):
45 magic_bytes = f.read(8)
46 magic_words = struct.unpack('II', magic_bytes);
48 idx = 1 if sys.byteorder == 'little' else 0
49 if magic_words[idx] == kMagicFirstHalf:
50 if magic_words[1-idx] == kMagic64SecondHalf:
52 elif magic_words[1-idx] == kMagic32SecondHalf:
55 raise Exception('Bad magic word in %s' % path)
58 def ReadOneFile(path):
59 with open(path, mode="rb") as f:
64 raise Exception('File %s is short (< 8 bytes)' % path)
65 bits = ReadMagicAndReturnBitness(f, path)
67 s = array.array(TypeCodeForBits(bits), f.read(size))
68 print >>sys.stderr, "%s: read %d %d-bit PCs from %s" % (prog_name, size * 8 / bits, bits, path)
74 s = s.union(set(ReadOneFile(f)))
75 print >> sys.stderr, "%s: %d files merged; %d PCs total" % \
76 (prog_name, len(files), len(s))
79 def PrintFiles(files):
82 else: # If there is just on file, print the PCs in order.
83 s = ReadOneFile(files[0])
84 print >> sys.stderr, "%s: 1 file merged; %d PCs total" % \
89 def MergeAndPrint(files):
90 if sys.stdout.isatty():
94 if max(s) > 0xFFFFFFFF:
96 array.array('I', MagicForBits(bits)).tofile(sys.stdout)
97 a = array.array(TypeCodeForBits(bits), s)
101 def UnpackOneFile(path):
102 with open(path, mode="rb") as f:
103 print >> sys.stderr, "%s: unpacking %s" % (prog_name, path)
106 if not header: return
109 pid, module_length, blob_size = struct.unpack('iII', header)
110 module = f.read(module_length)
111 blob = f.read(blob_size)
112 assert(len(module) == module_length)
113 assert(len(blob) == blob_size)
114 extracted_file = "%s.%d.sancov" % (module, pid)
115 print >> sys.stderr, "%s: extracting %s" % \
116 (prog_name, extracted_file)
117 # The packed file may contain multiple blobs for the same pid/module
118 # pair. Append to the end of the file instead of overwriting.
119 with open(extracted_file, 'ab') as f2:
122 raise Exception('Error reading file %s' % path)
129 def UnpackOneRawFile(path, map_path):
131 with open(map_path, mode="rt") as f_map:
132 print >> sys.stderr, "%s: reading map %s" % (prog_name, map_path)
133 bits = int(f_map.readline())
134 if bits != 32 and bits != 64:
135 raise Exception('Wrong bits size in the map')
137 parts = line.rstrip().split()
138 mem_map.append((int(parts[0], 16),
141 ' '.join(parts[3:])))
142 mem_map.sort(key=lambda m : m[0])
143 mem_map_keys = [m[0] for m in mem_map]
145 with open(path, mode="rb") as f:
146 print >> sys.stderr, "%s: unpacking %s" % (prog_name, path)
151 pcs = array.array(TypeCodeForBits(bits), f.read(size))
152 mem_map_pcs = [[] for i in range(0, len(mem_map))]
156 map_idx = bisect.bisect(mem_map_keys, pc) - 1
157 (start, end, base, module_path) = mem_map[map_idx]
160 print >> sys.stderr, "warning: %s: pc %x outside of any known mapping" % (prog_name, pc)
162 mem_map_pcs[map_idx].append(pc - base)
164 for ((start, end, base, module_path), pc_list) in zip(mem_map, mem_map_pcs):
165 if len(pc_list) == 0: continue
166 assert path.endswith('.sancov.raw')
167 dst_path = module_path + '.' + os.path.basename(path)[:-4]
168 print >> sys.stderr, "%s: writing %d PCs to %s" % (prog_name, len(pc_list), dst_path)
169 arr = array.array(TypeCodeForBits(bits))
170 arr.fromlist(sorted(pc_list))
171 with open(dst_path, 'ab') as f2:
172 array.array('I', MagicForBits(bits)).tofile(f2)
175 def RawUnpack(files):
177 if not f.endswith('.sancov.raw'):
178 raise Exception('Unexpected raw file name %s' % f)
179 f_map = f[:-3] + 'map'
180 UnpackOneRawFile(f, f_map)
182 def GetInstrumentedPCs(binary):
183 # This looks scary, but all it does is extract all offsets where we call:
184 # - __sanitizer_cov() or __sanitizer_cov_with_check(),
185 # - with call or callq,
186 # - directly or via PLT.
187 cmd = "objdump -d %s | " \
188 "grep '^\s\+[0-9a-f]\+:.*\scall\(q\|\)\s\+[0-9a-f]\+ <__sanitizer_cov\(_with_check\|\)\(@plt\|\)>' | " \
189 "grep '^\s\+[0-9a-f]\+' -o" % binary
190 proc = subprocess.Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE,
193 # The PCs we get from objdump are off by 4 bytes, as they point to the
194 # beginning of the callq instruction. Empirically this is true on x86 and
196 return set(int(line.strip(), 16) + 4 for line in proc.stdout)
198 def PrintMissing(binary):
199 if not os.path.isfile(binary):
200 raise Exception('File not found: %s' % binary)
201 instrumented = GetInstrumentedPCs(binary)
202 print >> sys.stderr, "%s: found %d instrumented PCs in %s" % (prog_name,
205 covered = set(int(line, 16) for line in sys.stdin)
206 print >> sys.stderr, "%s: read %d PCs from stdin" % (prog_name, len(covered))
207 missing = instrumented - covered
208 print >> sys.stderr, "%s: %d PCs missing from coverage" % (prog_name, len(missing))
209 if (len(missing) > len(instrumented) - len(covered)):
210 print >> sys.stderr, \
211 "%s: WARNING: stdin contains PCs not found in binary" % prog_name
212 for pc in sorted(missing):
215 if __name__ == '__main__':
216 prog_name = sys.argv[0]
217 if len(sys.argv) <= 2:
220 if sys.argv[1] == "missing":
221 if len(sys.argv) != 3:
223 PrintMissing(sys.argv[2])
227 for f in sys.argv[2:]:
228 file_list += glob.glob(f)
232 if sys.argv[1] == "print":
233 PrintFiles(file_list)
234 elif sys.argv[1] == "merge":
235 MergeAndPrint(file_list)
236 elif sys.argv[1] == "unpack":
238 elif sys.argv[1] == "rawunpack":