xref: /optee_os/scripts/symbolize.py (revision 5f7df50732ee837f1866a39e3ee4bac4d28dc679)
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', nargs='?',
65        help='Strip STRIP_PATH from file paths (default: current directory, '
66             'use -s with no argument to show full paths)',
67        default=os.getcwd())
68
69    return parser.parse_args()
70
71class Symbolizer(object):
72    def __init__(self, out, dirs, strip_path):
73        self._out = out
74        self._dirs = dirs
75        self._strip_path = strip_path
76        self._addr2line = None
77        self._bin = 'tee.elf'
78        self.reset()
79
80    def get_elf(self, elf_or_uuid):
81        if not elf_or_uuid.endswith('.elf'):
82            elf_or_uuid += '.elf'
83        for d in self._dirs:
84            if d.endswith(elf_or_uuid) and os.path.isfile(d):
85                return d
86            elf = glob.glob(d + '/' + elf_or_uuid)
87            if elf:
88                return elf[0]
89
90    def set_arch(self):
91        if self._arch:
92            return
93        if self._bin:
94            p = subprocess.Popen([ 'file', self.get_elf(self._bin) ],
95                                 stdout=subprocess.PIPE)
96            output = p.stdout.readlines()
97            p.terminate()
98            if 'ARM aarch64,' in output[0]:
99                self._arch = 'aarch64-linux-gnu-'
100            elif 'ARM,' in output[0]:
101                self._arch = 'arm-linux-gnueabihf-'
102
103    def arch_prefix(self, cmd):
104        self.set_arch()
105        return self._arch + cmd
106
107    def spawn_addr2line(self):
108        if not self._addr2line:
109            elf = self.get_elf(self._bin)
110            if not elf:
111                return
112            cmd = self.arch_prefix('addr2line')
113            if not cmd:
114                return
115            self._addr2line = subprocess.Popen([cmd, '-f', '-p', '-e', elf],
116                                                stdin = subprocess.PIPE,
117                                                stdout = subprocess.PIPE)
118
119    def subtract_load_addr(self, addr):
120        offs = self._load_addr
121        if int(offs, 16) > int(addr, 16):
122            return ''
123        return '0x{:x}'.format(int(addr, 16) - int(offs, 16))
124
125    def resolve(self, addr):
126        reladdr = self.subtract_load_addr(addr)
127        self.spawn_addr2line()
128        if not reladdr or not self._addr2line:
129            return '???'
130        try:
131            print >> self._addr2line.stdin, reladdr
132            ret = self._addr2line.stdout.readline().rstrip('\n')
133        except IOError:
134            ret = '!!!'
135        return ret
136
137    def symbol_plus_offset(self, addr):
138        ret = ''
139        prevsize = 0
140        reladdr = self.subtract_load_addr(addr)
141        elf = self.get_elf(self._bin)
142        cmd = self.arch_prefix('nm')
143        if not reladdr or not elf or not cmd:
144            return ''
145        ireladdr = int(reladdr, 16)
146        nm = subprocess.Popen([cmd, '--numeric-sort', '--print-size', elf],
147                               stdin = subprocess.PIPE,
148                               stdout = subprocess.PIPE)
149        for line in iter(nm.stdout.readline, ''):
150            try:
151                addr, size, _, name = line.split()
152            except:
153                # Size is missing
154                addr, _, name = line.split()
155                size = '0'
156            iaddr = int(addr, 16)
157            isize = int(size, 16)
158            if iaddr == ireladdr:
159                ret = name
160                break
161            if iaddr < ireladdr and iaddr + isize >= ireladdr:
162                offs = ireladdr - iaddr
163                ret = name + '+' + str(offs)
164                break
165            if iaddr > ireladdr and prevsize == 0:
166                offs = iaddr + ireladdr
167                ret = prevname + '+' + str(offs)
168                break
169            prevsize = size
170            prevname = name
171        nm.terminate()
172        return ret
173
174    def section_plus_offset(self, addr):
175        ret = ''
176        reladdr = self.subtract_load_addr(addr)
177        elf = self.get_elf(self._bin)
178        cmd = self.arch_prefix('objdump')
179        if not reladdr or not elf or not cmd:
180            return ''
181        iaddr = int(reladdr, 16)
182        objdump = subprocess.Popen([cmd, '--section-headers', elf],
183                                    stdin = subprocess.PIPE,
184                                    stdout = subprocess.PIPE)
185        for line in iter(objdump.stdout.readline, ''):
186            try:
187                idx, name, size, vma, lma, offs, algn = line.split()
188            except:
189                continue;
190            ivma = int(vma, 16)
191            isize = int(size, 16)
192            if ivma == iaddr:
193                ret = name
194                break
195            if ivma < iaddr and ivma + isize >= iaddr:
196                offs = iaddr - ivma
197                ret = name + '+' + str(offs)
198                break
199        objdump.terminate()
200        return ret
201
202    def process_abort(self, line):
203        ret = ''
204        match = re.search(ABORT_ADDR_RE, line)
205        addr = match.group('addr')
206        pre = match.start('addr')
207        post = match.end('addr')
208        sym = self.symbol_plus_offset(addr)
209        sec = self.section_plus_offset(addr)
210        if sym or sec:
211            ret += line[:pre]
212            ret += addr
213            if sym:
214                ret += ' ' + sym
215            if sec:
216                ret += ' ' + sec
217            ret += line[post:]
218        return ret
219
220    # Return all ELF sections with the ALLOC flag
221    def read_sections(self):
222        if self._sections:
223            return
224        elf = self.get_elf(self._bin)
225        cmd = self.arch_prefix('objdump')
226        if not elf or not cmd:
227            return
228        objdump = subprocess.Popen([cmd, '--section-headers', elf],
229                                    stdin = subprocess.PIPE,
230                                    stdout = subprocess.PIPE)
231        for line in iter(objdump.stdout.readline, ''):
232            try:
233                _, name, size, vma, _, _, _ = line.split()
234            except:
235                if 'ALLOC' in line:
236                    self._sections.append([name, int(vma, 16), int(size, 16)])
237
238    def overlaps(self, section, addr, size):
239        sec_addr = section[1]
240        sec_size = section[2]
241        if not size or not sec_size:
242            return False
243        return (addr <= (sec_addr + sec_size - 1)) and ((addr + size - 1) >= sec_addr)
244
245    def sections_in_region(self, addr, size):
246        ret = ''
247        addr = self.subtract_load_addr(addr)
248        if not addr:
249            return ''
250        iaddr = int(addr, 16)
251        isize = int(size, 16)
252        self.read_sections()
253        for s in self._sections:
254            if self.overlaps(s, iaddr, isize):
255                ret += ' ' + s[0]
256        return ret
257
258    def reset(self):
259        self._call_stack_found = False
260        self._load_addr = '0'
261        if self._addr2line:
262            self._addr2line.terminate()
263            self._addr2line = None
264        self._arch = None
265        self._saved_abort_line = ''
266        self._sections = []
267        self._bin = "tee.elf"
268
269    def write(self, line):
270            if self._call_stack_found:
271                match = re.search(STACK_ADDR_RE, line)
272                if match:
273                    addr = match.group('addr')
274                    pre = match.start('addr')
275                    post = match.end('addr')
276                    self._out.write(line[:pre])
277                    self._out.write(addr)
278                    res = self.resolve(addr)
279                    if self._strip_path:
280                        res = re.sub(re.escape(self._strip_path) + '/*', '',
281                              res)
282                    self._out.write(' ' + res)
283                    self._out.write(line[post:])
284                    return
285                else:
286                    self.reset()
287            match = re.search(REGION_RE, line)
288            if match:
289                addr = match.group('addr')
290                size = match.group('size')
291                self._out.write(line.strip() +
292                                self.sections_in_region(addr, size) + '\n');
293                return
294            match = re.search(CALL_STACK_RE, line)
295            if match:
296                self._call_stack_found = True
297                # Here is a good place to resolve the abort address because we
298                # have all the information we need
299                if self._saved_abort_line:
300                    self._out.write(self.process_abort(self._saved_abort_line))
301            match = re.search(TA_UUID_RE, line)
302            if match:
303                self._bin = match.group('uuid')
304            match = re.search(TA_INFO_RE, line)
305            if match:
306                self._load_addr = match.group('load_addr')
307            match = re.search(ABORT_ADDR_RE, line)
308            if match:
309                self.reset()
310                # At this point the arch and TA load address are unknown.
311                # Save the line so We can translate the abort address later.
312                self._saved_abort_line = line
313            self._out.write(line)
314
315    def flush(self):
316        self._out.flush()
317
318def main():
319    args = get_args()
320    if args.dir:
321        # Flatten list in case -d is used several times *and* with multiple
322        # arguments
323        args.dirs = [item for sublist in args.dir for item in sublist]
324    else:
325        args.dirs = []
326    symbolizer = Symbolizer(sys.stdout, args.dirs, args.strip_path)
327
328    for line in sys.stdin:
329        symbolizer.write(line)
330    symbolizer.flush()
331
332if __name__ == "__main__":
333    main()
334