Home | History | Annotate | Download | only in scripts 
      1 #!/usr/bin/env python
      2 #===- lib/asan/scripts/asan_symbolize.py -----------------------------------===#
      3 #
      4 #                     The LLVM Compiler Infrastructure
      5 #
      6 # This file is distributed under the University of Illinois Open Source
      7 # License. See LICENSE.TXT for details.
      8 #
      9 #===------------------------------------------------------------------------===#
     10 import os
     11 import re
     12 import sys
     13 import string
     14 import subprocess
     15 
     16 pipes = {}
     17 filetypes = {}
     18 DEBUG=False
     19 
     20 def patch_address(frameno, addr_s):
     21   ''' Subtracts 1 or 2 from the top frame's address.
     22   Top frame is normally the return address from asan_report*
     23   call, which is not expected to return at all. Because of that, this
     24   address often belongs to the next source code line, or even to a different
     25   function. '''
     26   if frameno == '0':
     27     addr = int(addr_s, 16)
     28     if os.uname()[4].startswith('arm'):
     29       # Cancel the Thumb bit
     30       addr = addr & (~1)
     31     addr -= 1
     32     return hex(addr)
     33   return addr_s
     34 
     35 
     36 def fix_filename(file_name):
     37   for path_to_cut in sys.argv[1:]:
     38     file_name = re.sub(".*" + path_to_cut, "", file_name)
     39   file_name = re.sub(".*asan_[a-z_]*.cc:[0-9]*", "_asan_rtl_", file_name)
     40   file_name = re.sub(".*crtstuff.c:0", "???:0", file_name)
     41   return file_name
     42 
     43 
     44 # TODO(glider): need some refactoring here
     45 def symbolize_addr2line(line):
     46   #0 0x7f6e35cf2e45  (/blah/foo.so+0x11fe45)
     47   match = re.match('^( *#([0-9]+) *0x[0-9a-f]+) *\((.*)\+(0x[0-9a-f]+)\)', line)
     48   if match:
     49     frameno = match.group(2)
     50     binary = match.group(3)
     51     addr = match.group(4)
     52     addr = patch_address(frameno, addr)
     53     if not pipes.has_key(binary):
     54       pipes[binary] = subprocess.Popen(["addr2line", "-f", "-e", binary],
     55                          stdin=subprocess.PIPE, stdout=subprocess.PIPE)
     56     p = pipes[binary]
     57     try:
     58       print >>p.stdin, addr
     59       function_name = p.stdout.readline().rstrip()
     60       file_name     = p.stdout.readline().rstrip()
     61     except:
     62       function_name = ""
     63       file_name = ""
     64     file_name = fix_filename(file_name)
     65 
     66     print match.group(1), "in", function_name, file_name
     67   else:
     68     print line.rstrip()
     69 
     70 
     71 def get_macho_filetype(binary):
     72   if not filetypes.has_key(binary):
     73     otool_pipe = subprocess.Popen(["otool", "-Vh",  binary],
     74       stdin=subprocess.PIPE, stdout=subprocess.PIPE)
     75     otool_line = "".join(otool_pipe.stdout.readlines())
     76     for t in ["DYLIB", "EXECUTE"]:
     77       if t in otool_line:
     78         filetypes[binary] = t
     79     otool_pipe.stdin.close()
     80   return filetypes[binary]
     81 
     82 
     83 def symbolize_atos(line):
     84   #0 0x7f6e35cf2e45  (/blah/foo.so+0x11fe45)
     85   match = re.match('^( *#([0-9]+) *)(0x[0-9a-f]+) *\((.*)\+(0x[0-9a-f]+)\)', line)
     86   if match:
     87     #print line
     88     prefix = match.group(1)
     89     frameno = match.group(2)
     90     orig_addr = match.group(3)
     91     binary = match.group(4)
     92     offset = match.group(5)
     93     addr = patch_address(frameno, orig_addr)
     94     load_addr = hex(int(orig_addr, 16) - int(offset, 16))
     95     filetype = get_macho_filetype(binary)
     96 
     97     if not pipes.has_key(binary):
     98       # Guess which arch we're running. 10 = len("0x") + 8 hex digits.
     99       if len(addr) > 10:
    100         arch = "x86_64"
    101       else:
    102         arch = "i386"
    103 
    104     if filetype == "DYLIB":
    105       load_addr = "0x0"
    106     if DEBUG:
    107       print "atos -o %s -arch %s -l %s" % (binary, arch, load_addr)
    108     pipes[binary] = subprocess.Popen(["atos", "-o", binary, "-arch", arch, "-l", load_addr],
    109                          stdin=subprocess.PIPE, stdout=subprocess.PIPE,)
    110     p = pipes[binary]
    111     if filetype == "DYLIB":
    112       print >>p.stdin, "%s" % offset
    113     else:
    114       print >>p.stdin, "%s" % addr
    115     # TODO(glider): it's more efficient to make a batch atos run for each binary.
    116     p.stdin.close()
    117     atos_line = p.stdout.readline().rstrip()
    118     # A well-formed atos response looks like this:
    119     #   foo(type1, type2) (in object.name) (filename.cc:80)
    120     match = re.match('^(.*) \(in (.*)\) \((.*:\d*)\)$', atos_line)
    121     #print "atos_line: ", atos_line
    122     if match:
    123       function_name = match.group(1)
    124       function_name = re.sub("\(.*?\)", "", function_name)
    125       file_name = fix_filename(match.group(3))
    126       print "%s%s in %s %s" % (prefix, addr, function_name, file_name)
    127     else:
    128       print "%s%s in %s" % (prefix, addr, atos_line)
    129     del pipes[binary]
    130   else:
    131     print line.rstrip()
    132 
    133 system = os.uname()[0]
    134 if system in ['Linux', 'Darwin']:
    135   for line in sys.stdin:
    136     if system == 'Linux':
    137       symbolize_addr2line(line)
    138     elif system == 'Darwin':
    139       symbolize_atos(line)
    140 else:
    141   print 'Unknown system: ', system
    142