Linux debugging

Check our new training course

Linux debugging, tracing, profiling & perf. analysis

Check our new training course
with Creative Commons CC-BY-SA
lecture and lab materials

Bootlin logo

Elixir Cross Referencer

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
#!/usr/bin/env python
# SPDX-License-Identifier: BSD-2-Clause
#
# Copyright (c) 2017, Linaro Limited
#


import argparse
import glob
import os
import re
import subprocess
import sys

CALL_STACK_RE = re.compile('Call stack:')
# This gets the address from lines looking like this:
# E/TC:0  0x001044a8
STACK_ADDR_RE = re.compile(r'[UEIDFM]/T[AC]:.*(?P<addr>0x[0-9a-f]+)')
ABORT_ADDR_RE = re.compile('-abort at address (?P<addr>0x[0-9a-f]+)')
REGION_RE = re.compile('region [0-9]+: va (?P<addr>0x[0-9a-f]+) '
                       'pa 0x[0-9a-f]+ size (?P<size>0x[0-9a-f]+)'
                       '( flags .{6} (\[(?P<elf_idx>[0-9]+)\])?)?')
ELF_LIST_RE = re.compile(r'\[(?P<idx>[0-9]+)\] (?P<uuid>[0-9a-f\-]+)'
                         ' @ (?P<load_addr>0x[0-9a-f\-]+)')

epilog = '''
This scripts reads an OP-TEE abort or panic message from stdin and adds debug
information to the output, such as '<function> at <file>:<line>' next to each
address in the call stack. Any message generated by OP-TEE and containing a
call stack can in principle be processed by this script. This currently
includes aborts and panics from the TEE core as well as from any TA.
The paths provided on the command line are used to locate the appropriate ELF
binary (tee.elf or Trusted Application). The GNU binutils (addr2line, objdump,
nm) are used to extract the debug info.

OP-TEE abort and panic messages are sent to the secure console. They look like
the following:

  E/TC:0 User TA data-abort at address 0xffffdecd (alignment fault)
  ...
  E/TC:0 Call stack:
  E/TC:0  0x4000549e
  E/TC:0  0x40001f4b
  E/TC:0  0x4000273f
  E/TC:0  0x40005da7

Inspired by a script of the same name by the Chromium project.

Sample usage:

  $ scripts/symbolize.py -d out/arm-plat-hikey/core -d ../optee_test/out/ta/*
  <paste whole dump here>
  ^D
'''


def get_args():
    parser = argparse.ArgumentParser(
                formatter_class=argparse.RawDescriptionHelpFormatter,
                description='Symbolizes OP-TEE abort dumps',
                epilog=epilog)
    parser.add_argument('-d', '--dir', action='append', nargs='+',
        help='Search for ELF file in DIR. tee.elf is needed to decode '
             'a TEE Core or pseudo-TA abort, while <TA_uuid>.elf is required '
             'if a user-mode TA has crashed. For convenience, ELF files '
             'may also be given.')
    parser.add_argument('-s', '--strip_path', nargs='?',
        help='Strip STRIP_PATH from file paths (default: current directory, '
             'use -s with no argument to show full paths)',
        default=os.getcwd())

    return parser.parse_args()


class Symbolizer(object):
    def __init__(self, out, dirs, strip_path):
        self._out = out
        self._dirs = dirs
        self._strip_path = strip_path
        self._addr2line = None
        self.reset()

    def get_elf(self, elf_or_uuid):
        if not elf_or_uuid.endswith('.elf'):
            elf_or_uuid += '.elf'
        for d in self._dirs:
            if d.endswith(elf_or_uuid) and os.path.isfile(d):
                return d
            elf = glob.glob(d + '/' + elf_or_uuid)
            if elf:
                return elf[0]

    def set_arch(self):
        if self._arch:
            return
        elf = self.get_elf(self._elfs[0][0])
        if elf is None:
            return
        p = subprocess.Popen(['file', self.get_elf(self._elfs[0][0])],
                             stdout=subprocess.PIPE)
        output = p.stdout.readlines()
        p.terminate()
        if 'ARM aarch64,' in output[0]:
            self._arch = 'aarch64-linux-gnu-'
        elif 'ARM,' in output[0]:
            self._arch = 'arm-linux-gnueabihf-'

    def arch_prefix(self, cmd):
        self.set_arch()
        if self._arch is None:
            return ''
        return self._arch + cmd

    def spawn_addr2line(self, elf_name):
        if elf_name is None:
            return
        if self._addr2line_elf_name is elf_name:
            return
        if self._addr2line:
            self._addr2line.terminate
            self._addr2line = None
        elf = self.get_elf(elf_name)
        if not elf:
            return
        cmd = self.arch_prefix('addr2line')
        if not cmd:
            return
        self._addr2line = subprocess.Popen([cmd, '-f', '-p', '-e', elf],
                                           stdin=subprocess.PIPE,
                                           stdout=subprocess.PIPE)
        self._addr2line_elf_name = elf_name

    # If addr falls into a region that maps a TA ELF file, return the load
    # address of that file.
    def elf_load_addr(self, addr):
        if self._regions:
            for r in self._regions:
                r_addr = int(r[0], 16)
                r_size = int(r[1], 16)
                i_addr = int(addr, 16)
                if (i_addr >= r_addr and i_addr < (r_addr + r_size)):
                    # Found region
                    elf_idx = r[2]
                    if elf_idx is not None:
                        return self._elfs[int(elf_idx)][1]
            return None
        else:
            # tee.elf
            return '0x0'

    def elf_for_addr(self, addr):
        l_addr = self.elf_load_addr(addr)
        if l_addr is None:
            return None
        if l_addr is '0x0':
            return 'tee.elf'
        for k in self._elfs:
            e = self._elfs[k]
            if int(e[1], 16) == int(l_addr, 16):
                return e[0]
        return None

    def subtract_load_addr(self, addr):
        l_addr = self.elf_load_addr(addr)
        if l_addr is None:
            return None
        if int(l_addr, 16) > int(addr, 16):
            return ''
        return '0x{:x}'.format(int(addr, 16) - int(l_addr, 16))

    def resolve(self, addr):
        reladdr = self.subtract_load_addr(addr)
        self.spawn_addr2line(self.elf_for_addr(addr))
        if not reladdr or not self._addr2line:
            return '???'
        try:
            print >> self._addr2line.stdin, reladdr
            ret = self._addr2line.stdout.readline().rstrip('\n')
        except IOError:
            ret = '!!!'
        return ret

    def symbol_plus_offset(self, addr):
        ret = ''
        prevsize = 0
        reladdr = self.subtract_load_addr(addr)
        elf_name = self.elf_for_addr(addr)
        if elf_name is None:
            return ''
        elf = self.get_elf(elf_name)
        cmd = self.arch_prefix('nm')
        if not reladdr or not elf or not cmd:
            return ''
        ireladdr = int(reladdr, 16)
        nm = subprocess.Popen([cmd, '--numeric-sort', '--print-size', elf],
                              stdin=subprocess.PIPE,
                              stdout=subprocess.PIPE)
        for line in iter(nm.stdout.readline, ''):
            try:
                addr, size, _, name = line.split()
            except:
                # Size is missing
                try:
                    addr, _, name = line.split()
                    size = '0'
                except:
                    # E.g., undefined (external) symbols (line = "U symbol")
                    continue
            iaddr = int(addr, 16)
            isize = int(size, 16)
            if iaddr == ireladdr:
                ret = name
                break
            if iaddr < ireladdr and iaddr + isize >= ireladdr:
                offs = ireladdr - iaddr
                ret = name + '+' + str(offs)
                break
            if iaddr > ireladdr and prevsize == 0:
                offs = iaddr + ireladdr
                ret = prevname + '+' + str(offs)
                break
            prevsize = size
            prevname = name
        nm.terminate()
        return ret

    def section_plus_offset(self, addr):
        ret = ''
        reladdr = self.subtract_load_addr(addr)
        elf_name = self.elf_for_addr(addr)
        if elf_name is None:
            return ''
        elf = self.get_elf(elf_name)
        cmd = self.arch_prefix('objdump')
        if not reladdr or not elf or not cmd:
            return ''
        iaddr = int(reladdr, 16)
        objdump = subprocess.Popen([cmd, '--section-headers', elf],
                                   stdin=subprocess.PIPE,
                                   stdout=subprocess.PIPE)
        for line in iter(objdump.stdout.readline, ''):
            try:
                idx, name, size, vma, lma, offs, algn = line.split()
            except:
                continue
            ivma = int(vma, 16)
            isize = int(size, 16)
            if ivma == iaddr:
                ret = name
                break
            if ivma < iaddr and ivma + isize >= iaddr:
                offs = iaddr - ivma
                ret = name + '+' + str(offs)
                break
        objdump.terminate()
        return ret

    def process_abort(self, line):
        ret = ''
        match = re.search(ABORT_ADDR_RE, line)
        addr = match.group('addr')
        pre = match.start('addr')
        post = match.end('addr')
        sym = self.symbol_plus_offset(addr)
        sec = self.section_plus_offset(addr)
        if sym or sec:
            ret += line[:pre]
            ret += addr
            if sym:
                ret += ' ' + sym
            if sec:
                ret += ' ' + sec
            ret += line[post:]
        return ret

    # Return all ELF sections with the ALLOC flag
    def read_sections(self, elf_name):
        if elf_name is None:
            return
        if elf_name in self._sections:
            return
        elf = self.get_elf(elf_name)
        cmd = self.arch_prefix('objdump')
        if not elf or not cmd:
            return
        self._sections[elf_name] = []
        objdump = subprocess.Popen([cmd, '--section-headers', elf],
                                   stdin=subprocess.PIPE,
                                   stdout=subprocess.PIPE)
        for line in iter(objdump.stdout.readline, ''):
            try:
                _, name, size, vma, _, _, _ = line.split()
            except:
                if 'ALLOC' in line:
                    self._sections[elf_name].append([name, int(vma, 16),
                                                     int(size, 16)])

    def overlaps(self, section, addr, size):
        sec_addr = section[1]
        sec_size = section[2]
        if not size or not sec_size:
            return False
        return ((addr <= (sec_addr + sec_size - 1)) and
                ((addr + size - 1) >= sec_addr))

    def sections_in_region(self, addr, size, elf_idx):
        ret = ''
        addr = self.subtract_load_addr(addr)
        if not addr:
            return ''
        iaddr = int(addr, 16)
        isize = int(size, 16)
        elf = self._elfs[int(elf_idx)][0]
        if elf is None:
            return ''
        self.read_sections(elf)
        if elf not in self._sections:
            return ''
        for s in self._sections[elf]:
            if self.overlaps(s, iaddr, isize):
                ret += ' ' + s[0]
        return ret

    def reset(self):
        self._call_stack_found = False
        if self._addr2line:
            self._addr2line.terminate()
            self._addr2line = None
        self._addr2line_elf_name = None
        self._arch = None
        self._saved_abort_line = ''
        self._sections = {}  # {elf_name: [[name, addr, size], ...], ...}
        self._regions = []   # [[addr, size, elf_idx, saved line], ...]
        self._elfs = {0: ["tee.elf", 0]}  # {idx: [uuid, load_addr], ...}


    def pretty_print_path(self, path):
        if self._strip_path:
            return re.sub(re.escape(self._strip_path) + '/*', '', path)
        return path


    def write(self, line):
            if self._call_stack_found:
                match = re.search(STACK_ADDR_RE, line)
                if match:
                    addr = match.group('addr')
                    pre = match.start('addr')
                    post = match.end('addr')
                    self._out.write(line[:pre])
                    self._out.write(addr)
                    res = self.resolve(addr)
                    res = self.pretty_print_path(res)
                    self._out.write(' ' + res)
                    self._out.write(line[post:])
                    return
                else:
                    self.reset()
            match = re.search(REGION_RE, line)
            if match:
                # Region table: save info for later processing once
                # we know which UUID corresponds to which ELF index
                addr = match.group('addr')
                size = match.group('size')
                elf_idx = match.group('elf_idx')
                self._regions.append([addr, size, elf_idx, line])
                return
            match = re.search(ELF_LIST_RE, line)
            if match:
                # ELF list: save info for later. Region table and ELF list
                # will be displayed when the call stack is reached
                i = int(match.group('idx'))
                self._elfs[i] = [match.group('uuid'), match.group('load_addr'),
                                 line]
                return
            match = re.search(CALL_STACK_RE, line)
            if match:
                self._call_stack_found = True
                if self._regions:
                    for r in self._regions:
                        r_addr = r[0]
                        r_size = r[1]
                        elf_idx = r[2]
                        saved_line = r[3]
                        if elf_idx is None:
                            self._out.write(saved_line)
                        else:
                            self._out.write(saved_line.strip() +
                                            self.sections_in_region(r_addr,
                                                                    r_size,
                                                                    elf_idx) +
                                            '\n')
                if self._elfs:
                    for k in self._elfs:
                        e = self._elfs[k]
                        if (len(e) >= 3):
                            # TA executable or library
                            self._out.write(e[2].strip())
                            elf = self.get_elf(e[0])
                            if elf:
                                rpath = os.path.realpath(elf)
                                path = self.pretty_print_path(rpath)
                                self._out.write(' (' + path + ')')
                            self._out.write('\n')
                # Here is a good place to resolve the abort address because we
                # have all the information we need
                if self._saved_abort_line:
                    self._out.write(self.process_abort(self._saved_abort_line))
            match = re.search(ABORT_ADDR_RE, line)
            if match:
                self.reset()
                # At this point the arch and TA load address are unknown.
                # Save the line so We can translate the abort address later.
                self._saved_abort_line = line
            self._out.write(line)

    def flush(self):
        self._out.flush()


def main():
    args = get_args()
    if args.dir:
        # Flatten list in case -d is used several times *and* with multiple
        # arguments
        args.dirs = [item for sublist in args.dir for item in sublist]
    else:
        args.dirs = []
    symbolizer = Symbolizer(sys.stdout, args.dirs, args.strip_path)

    for line in sys.stdin:
        symbolizer.write(line)
    symbolizer.flush()

if __name__ == "__main__":
    main()