xref: /optee_os/scripts/symbolize.py (revision b1d7375c01ec8bcbf3561d27425d320afed23bce)
1#!/usr/bin/env python
2# SPDX-License-Identifier: BSD-2-Clause
3#
4# Copyright (c) 2017, Linaro Limited
5#
6
7
8import argparse
9import glob
10import os
11import re
12import subprocess
13import sys
14
15TA_UUID_RE = re.compile(r'Status of TA (?P<uuid>[0-9a-f\-]+)')
16TA_INFO_RE = re.compile('  arch: (?P<arch>\w+)  '
17                        'load address: (?P<load_addr>0x[0-9a-f]+)')
18CALL_STACK_RE = re.compile('Call stack:')
19
20# This gets the address from lines looking like this:
21# E/TC:0  0x001044a8
22STACK_ADDR_RE = re.compile(r'[UEIDFM]/T[AC]:.*(?P<addr>0x[0-9a-f]+)')
23ABORT_ADDR_RE = re.compile('-abort at address (?P<addr>0x[0-9a-f]+)')
24REGION_RE = re.compile('region [0-9]+: va (?P<addr>0x[0-9a-f]+) '
25                       'pa 0x[0-9a-f]+ size (?P<size>0x[0-9a-f]+)')
26
27epilog = '''
28This scripts reads an OP-TEE abort message from stdin and adds debug
29information ('function at file:line') next to each address in the call stack.
30It uses the paths provided on the command line to locate the appropriate ELF
31binary (tee.elf or Trusted Application) and runs arm-linux-gnueabihf-addr2line
32or aarch64-linux-gnu-addr2line to process the addresses.
33
34OP-TEE abort messages are sent to the secure console. They look like the
35following:
36
37  ERROR:   TEE-CORE: User TA data-abort at address 0xffffdecd (alignment fault)
38  ...
39  ERROR:   TEE-CORE: Call stack:
40  ERROR:   TEE-CORE:  0x4000549e
41  ERROR:   TEE-CORE:  0x40001f4b
42  ERROR:   TEE-CORE:  0x4000273f
43  ERROR:   TEE-CORE:  0x40005da7
44
45Inspired by a script of the same name by the Chromium project.
46
47Sample usage:
48
49  $ scripts/symbolize.py -d out/arm-plat-hikey/core -d ../optee_test/out/ta/*
50  <paste whole dump here>
51  ^D
52'''
53
54def get_args():
55    parser = argparse.ArgumentParser(
56                formatter_class=argparse.RawDescriptionHelpFormatter,
57                description='Symbolizes OP-TEE abort dumps',
58                epilog=epilog)
59    parser.add_argument('-d', '--dir', action='append', nargs='+',
60        help='Search for ELF file in DIR. tee.elf is needed to decode '
61             'a TEE Core or pseudo-TA abort, while <TA_uuid>.elf is required '
62             'if a user-mode TA has crashed. For convenience, ELF files '
63             'may also be given.')
64    parser.add_argument('-s', '--strip_path',
65        help='Strip STRIP_PATH from file paths')
66
67    return parser.parse_args()
68
69class Symbolizer(object):
70    def __init__(self, out, dirs, strip_path):
71        self._out = out
72        self._dirs = dirs
73        self._strip_path = strip_path
74        self._addr2line = None
75        self._bin = 'tee.elf'
76        self.reset()
77
78    def get_elf(self, elf_or_uuid):
79        if not elf_or_uuid.endswith('.elf'):
80            elf_or_uuid += '.elf'
81        for d in self._dirs:
82            if d.endswith(elf_or_uuid) and os.path.isfile(d):
83                return d
84            elf = glob.glob(d + '/' + elf_or_uuid)
85            if elf:
86                return elf[0]
87
88    def set_arch(self):
89        if self._arch:
90            return
91        if self._bin:
92            p = subprocess.Popen([ 'file', self.get_elf(self._bin) ],
93                                 stdout=subprocess.PIPE)
94            output = p.stdout.readlines()
95            p.terminate()
96            if 'ARM aarch64,' in output[0]:
97                self._arch = 'aarch64-linux-gnu-'
98            elif 'ARM,' in output[0]:
99                self._arch = 'arm-linux-gnueabihf-'
100
101    def arch_prefix(self, cmd):
102        self.set_arch()
103        return self._arch + cmd
104
105    def spawn_addr2line(self):
106        if not self._addr2line:
107            elf = self.get_elf(self._bin)
108            if not elf:
109                return
110            cmd = self.arch_prefix('addr2line')
111            if not cmd:
112                return
113            self._addr2line = subprocess.Popen([cmd, '-f', '-p', '-e', elf],
114                                                stdin = subprocess.PIPE,
115                                                stdout = subprocess.PIPE)
116
117    def subtract_load_addr(self, addr):
118        offs = self._load_addr
119        if int(offs, 16) > int(addr, 16):
120            return ''
121        return '0x{:x}'.format(int(addr, 16) - int(offs, 16))
122
123    def resolve(self, addr):
124        reladdr = self.subtract_load_addr(addr)
125        self.spawn_addr2line()
126        if not reladdr or not self._addr2line:
127            return '???'
128        try:
129            print >> self._addr2line.stdin, reladdr
130            ret = self._addr2line.stdout.readline().rstrip('\n')
131        except IOError:
132            ret = '!!!'
133        return ret
134
135    def symbol_plus_offset(self, addr):
136        ret = ''
137        prevsize = 0
138        reladdr = self.subtract_load_addr(addr)
139        elf = self.get_elf(self._bin)
140        cmd = self.arch_prefix('nm')
141        if not reladdr or not elf or not cmd:
142            return ''
143        ireladdr = int(reladdr, 16)
144        nm = subprocess.Popen([cmd, '--numeric-sort', '--print-size', elf],
145                               stdin = subprocess.PIPE,
146                               stdout = subprocess.PIPE)
147        for line in iter(nm.stdout.readline, ''):
148            try:
149                addr, size, _, name = line.split()
150            except:
151                # Size is missing
152                addr, _, name = line.split()
153                size = '0'
154            iaddr = int(addr, 16)
155            isize = int(size, 16)
156            if iaddr == ireladdr:
157                ret = name
158                break
159            if iaddr < ireladdr and iaddr + isize >= ireladdr:
160                offs = ireladdr - iaddr
161                ret = name + '+' + str(offs)
162                break
163            if iaddr > ireladdr and prevsize == 0:
164                offs = iaddr + ireladdr
165                ret = prevname + '+' + str(offs)
166                break
167            prevsize = size
168            prevname = name
169        nm.terminate()
170        return ret
171
172    def section_plus_offset(self, addr):
173        ret = ''
174        reladdr = self.subtract_load_addr(addr)
175        elf = self.get_elf(self._bin)
176        cmd = self.arch_prefix('objdump')
177        if not reladdr or not elf or not cmd:
178            return ''
179        iaddr = int(reladdr, 16)
180        objdump = subprocess.Popen([cmd, '--section-headers', elf],
181                                    stdin = subprocess.PIPE,
182                                    stdout = subprocess.PIPE)
183        for line in iter(objdump.stdout.readline, ''):
184            try:
185                idx, name, size, vma, lma, offs, algn = line.split()
186            except:
187                continue;
188            ivma = int(vma, 16)
189            isize = int(size, 16)
190            if ivma == iaddr:
191                ret = name
192                break
193            if ivma < iaddr and ivma + isize >= iaddr:
194                offs = iaddr - ivma
195                ret = name + '+' + str(offs)
196                break
197        objdump.terminate()
198        return ret
199
200    def process_abort(self, line):
201        ret = ''
202        match = re.search(ABORT_ADDR_RE, line)
203        addr = match.group('addr')
204        pre = match.start('addr')
205        post = match.end('addr')
206        sym = self.symbol_plus_offset(addr)
207        sec = self.section_plus_offset(addr)
208        if sym or sec:
209            ret += line[:pre]
210            ret += addr
211            if sym:
212                ret += ' ' + sym
213            if sec:
214                ret += ' ' + sec
215            ret += line[post:]
216        return ret
217
218    # Return all ELF sections with the ALLOC flag
219    def read_sections(self):
220        if self._sections:
221            return
222        elf = self.get_elf(self._bin)
223        cmd = self.arch_prefix('objdump')
224        if not elf or not cmd:
225            return
226        objdump = subprocess.Popen([cmd, '--section-headers', elf],
227                                    stdin = subprocess.PIPE,
228                                    stdout = subprocess.PIPE)
229        for line in iter(objdump.stdout.readline, ''):
230            try:
231                _, name, size, vma, _, _, _ = line.split()
232            except:
233                if 'ALLOC' in line:
234                    self._sections.append([name, int(vma, 16), int(size, 16)])
235
236    def overlaps(self, section, addr, size):
237        sec_addr = section[1]
238        sec_size = section[2]
239        if not size or not sec_size:
240            return False
241        return (addr <= (sec_addr + sec_size - 1)) and ((addr + size - 1) >= sec_addr)
242
243    def sections_in_region(self, addr, size):
244        ret = ''
245        addr = self.subtract_load_addr(addr)
246        if not addr:
247            return ''
248        iaddr = int(addr, 16)
249        isize = int(size, 16)
250        self.read_sections()
251        for s in self._sections:
252            if self.overlaps(s, iaddr, isize):
253                ret += ' ' + s[0]
254        return ret
255
256    def reset(self):
257        self._call_stack_found = False
258        self._load_addr = '0'
259        if self._addr2line:
260            self._addr2line.terminate()
261            self._addr2line = None
262        self._arch = None
263        self._saved_abort_line = ''
264        self._sections = []
265        self._bin = "tee.elf"
266
267    def write(self, line):
268            if self._call_stack_found:
269                match = re.search(STACK_ADDR_RE, line)
270                if match:
271                    addr = match.group('addr')
272                    pre = match.start('addr')
273                    post = match.end('addr')
274                    self._out.write(line[:pre])
275                    self._out.write(addr)
276                    res = self.resolve(addr)
277                    if self._strip_path:
278                        res = re.sub(re.escape(self._strip_path) + '/*', '',
279                              res)
280                    self._out.write(' ' + res)
281                    self._out.write(line[post:])
282                    return
283                else:
284                    self.reset()
285            match = re.search(REGION_RE, line)
286            if match:
287                addr = match.group('addr')
288                size = match.group('size')
289                self._out.write(line.strip() +
290                                self.sections_in_region(addr, size) + '\n');
291                return
292            match = re.search(CALL_STACK_RE, line)
293            if match:
294                self._call_stack_found = True
295                # Here is a good place to resolve the abort address because we
296                # have all the information we need
297                if self._saved_abort_line:
298                    self._out.write(self.process_abort(self._saved_abort_line))
299            match = re.search(TA_UUID_RE, line)
300            if match:
301                self._bin = match.group('uuid')
302            match = re.search(TA_INFO_RE, line)
303            if match:
304                self._load_addr = match.group('load_addr')
305            match = re.search(ABORT_ADDR_RE, line)
306            if match:
307                self.reset()
308                # At this point the arch and TA load address are unknown.
309                # Save the line so We can translate the abort address later.
310                self._saved_abort_line = line
311            self._out.write(line)
312
313    def flush(self):
314        self._out.flush()
315
316def main():
317    args = get_args()
318    if args.dir:
319        # Flatten list in case -d is used several times *and* with multiple
320        # arguments
321        args.dirs = [item for sublist in args.dir for item in sublist]
322    else:
323        args.dirs = []
324    symbolizer = Symbolizer(sys.stdout, args.dirs, args.strip_path)
325
326    for line in sys.stdin:
327        symbolizer.write(line)
328    symbolizer.flush()
329
330if __name__ == "__main__":
331    main()
332