tools: Split up large binaries into smaller chunks in the DFU binary

ROM will erase the region a partition is in as soon as it receives the
first bit of the data that is in the partition. For large partitions it
takes more than 5 seconds to erase which is a hard-coded limit in
dfu-utils.

This splits large binaries and adds them by chunks which should avoid
timing-out during flashing.

Closes https://github.com/espressif/esp-idf/issues/6999
This commit is contained in:
Roland Dobai 2021-05-14 14:22:54 +02:00
parent 71cf821659
commit 7b70154566
6 changed files with 175 additions and 107 deletions

View File

@ -6,9 +6,9 @@ def action_extensions(base_actions, project_path):
SUPPORTED_TARGETS = ['esp32s2']
def dfu_target(target_name, ctx, args):
def dfu_target(target_name, ctx, args, part_size):
ensure_build_directory(args, ctx.info_name)
run_target(target_name, args)
run_target(target_name, args, {'ESP_DFU_PART_SIZE': part_size} if part_size else {})
def dfu_flash_target(target_name, ctx, args):
ensure_build_directory(args, ctx.info_name)
@ -22,16 +22,24 @@ def action_extensions(base_actions, project_path):
raise
dfu_actions = {
"actions": {
"dfu": {
"callback": dfu_target,
"short_help": "Build the DFU binary",
"dependencies": ["all"],
'actions': {
'dfu': {
'callback': dfu_target,
'short_help': 'Build the DFU binary',
'dependencies': ['all'],
'options': [
{
'names': ['--part-size'],
'help': 'Large files are split up into smaller partitions in order to avoid timeout during '
'erasing flash. This option allows to overwrite the default partition size of '
'mkdfu.py.'
}
],
},
"dfu-flash": {
"callback": dfu_flash_target,
"short_help": "Flash the DFU binary",
"order_dependencies": ["dfu"],
'dfu-flash': {
'callback': dfu_flash_target,
'short_help': 'Flash the DFU binary',
'order_dependencies': ['dfu'],
},
}
}

View File

@ -1,6 +1,6 @@
#!/usr/bin/env python
#
# Copyright 2020 Espressif Systems (Shanghai) PTE LTD
# Copyright 2020-2021 Espressif Systems (Shanghai) CO LTD
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -21,14 +21,18 @@
# This file must be the first one in the archive. It contains binary structures describing each
# subsequent file (for example, where the file needs to be flashed/loaded).
from collections import namedtuple
from future.utils import iteritems
from __future__ import print_function, unicode_literals
import argparse
import hashlib
import json
import os
import struct
import zlib
from collections import namedtuple
from functools import partial
from future.utils import iteritems
try:
import typing
@ -43,28 +47,28 @@ except ImportError:
pass
# CPIO ("new ASCII") format related things
CPIO_MAGIC = b"070701"
CPIO_STRUCT = b"=6s" + b"8s" * 13
CPIO_MAGIC = b'070701'
CPIO_STRUCT = b'=6s' + b'8s' * 13
CPIOHeader = namedtuple(
"CPIOHeader",
'CPIOHeader',
[
"magic",
"ino",
"mode",
"uid",
"gid",
"nlink",
"mtime",
"filesize",
"devmajor",
"devminor",
"rdevmajor",
"rdevminor",
"namesize",
"check",
'magic',
'ino',
'mode',
'uid',
'gid',
'nlink',
'mtime',
'filesize',
'devmajor',
'devminor',
'rdevmajor',
'rdevminor',
'namesize',
'check',
],
)
CPIO_TRAILER = "TRAILER!!!"
CPIO_TRAILER = 'TRAILER!!!'
def make_cpio_header(
@ -73,7 +77,7 @@ def make_cpio_header(
""" Returns CPIOHeader for the given file name and file size """
def as_hex(val): # type: (int) -> bytes
return "{:08x}".format(val).encode("ascii")
return '{:08x}'.format(val).encode('ascii')
hex_0 = as_hex(0)
mode = hex_0 if is_trailer else as_hex(0o0100644)
@ -98,19 +102,17 @@ def make_cpio_header(
# DFU format related things
# Structure of one entry in dfuinfo0.dat
DFUINFO_STRUCT = b"<I I 64s 16s"
DFUInfo = namedtuple("DFUInfo", ["address", "flags", "name", "md5"])
DFUINFO_FILE = "dfuinfo0.dat"
DFUINFO_STRUCT = b'<I I 64s 16s'
DFUInfo = namedtuple('DFUInfo', ['address', 'flags', 'name', 'md5'])
DFUINFO_FILE = 'dfuinfo0.dat'
# Structure which gets added at the end of the entire DFU file
DFUSUFFIX_STRUCT = b"<H H H H 3s B"
DFUSUFFIX_STRUCT = b'<H H H H 3s B'
DFUSuffix = namedtuple(
"DFUSuffix", ["bcd_device", "pid", "vid", "bcd_dfu", "sig", "len"]
'DFUSuffix', ['bcd_device', 'pid', 'vid', 'bcd_dfu', 'sig', 'len']
)
ESPRESSIF_VID = 12346
# TODO: set PID based on the chip type (add a command line argument)
DFUSUFFIX_DEFAULT = DFUSuffix(0xFFFF, 0xFFFF, ESPRESSIF_VID, 0x0100, b"UFD", 16)
# This CRC32 gets added after DFUSUFFIX_STRUCT
DFUCRC_STRUCT = b"<I"
DFUCRC_STRUCT = b'<I'
def dfu_crc(data, crc=0): # type: (bytes, int) -> int
@ -119,39 +121,56 @@ def dfu_crc(data, crc=0): # type: (bytes, int) -> int
return uint32_max - (zlib.crc32(data, crc) & uint32_max)
def pad_bytes(b, multiple, padding=b"\x00"): # type: (bytes, int, bytes) -> bytes
def pad_bytes(b, multiple, padding=b'\x00'): # type: (bytes, int, bytes) -> bytes
""" Pad 'b' to a length divisible by 'multiple' """
padded_len = (len(b) + multiple - 1) // multiple * multiple
return b + padding * (padded_len - len(b))
class EspDfuWriter(object):
def __init__(self, dest_file): # type: (typing.BinaryIO) -> None
def __init__(self, dest_file, pid, part_size): # type: (typing.BinaryIO, int, int) -> None
self.dest = dest_file
self.pid = pid
self.part_size = part_size
self.entries = [] # type: typing.List[bytes]
self.index = [] # type: typing.List[DFUInfo]
def add_file(self, flash_addr, path): # type: (int, str) -> None
""" Add file to be written into flash at given address """
with open(path, "rb") as f:
self._add_cpio_flash_entry(os.path.basename(path), flash_addr, f.read())
"""
Add file to be written into flash at given address
Files are split up into chunks in order avoid timing-out during erasing large regions. Instead of adding
"app.bin" at flash_addr it will add:
1. app.bin at flash_addr # sizeof(app.bin) == self.part_size
2. app.bin.1 at flash_addr + self.part_size
3. app.bin.2 at flash_addr + 2 * self.part_size
...
"""
f_name = os.path.basename(path)
with open(path, 'rb') as f:
for i, chunk in enumerate(iter(partial(f.read, self.part_size), b'')):
n = f_name if i == 0 else '.'.join([f_name, str(i)])
self._add_cpio_flash_entry(n, flash_addr, chunk)
flash_addr += len(chunk)
def finish(self): # type: () -> None
""" Write DFU file """
# Prepare and add dfuinfo0.dat file
dfuinfo = b"".join([struct.pack(DFUINFO_STRUCT, *item) for item in self.index])
dfuinfo = b''.join([struct.pack(DFUINFO_STRUCT, *item) for item in self.index])
self._add_cpio_entry(DFUINFO_FILE, dfuinfo, first=True)
# Add CPIO archive trailer
self._add_cpio_entry(CPIO_TRAILER, b"", trailer=True)
self._add_cpio_entry(CPIO_TRAILER, b'', trailer=True)
# Combine all the entries and pad the file
out_data = b"".join(self.entries)
out_data = b''.join(self.entries)
cpio_block_size = 10240
out_data = pad_bytes(out_data, cpio_block_size)
# Add DFU suffix and CRC
out_data += struct.pack(DFUSUFFIX_STRUCT, *DFUSUFFIX_DEFAULT)
dfu_suffix = DFUSuffix(0xFFFF, self.pid, ESPRESSIF_VID, 0x0100, b'UFD', 16)
out_data += struct.pack(DFUSUFFIX_STRUCT, *dfu_suffix)
out_data += struct.pack(DFUCRC_STRUCT, dfu_crc(out_data))
# Finally write the entire binary
@ -166,7 +185,7 @@ class EspDfuWriter(object):
DFUInfo(
address=flash_addr,
flags=0,
name=filename.encode("utf-8"),
name=filename.encode('utf-8'),
md5=md5.digest(),
)
)
@ -175,7 +194,7 @@ class EspDfuWriter(object):
def _add_cpio_entry(
self, filename, data, first=False, trailer=False
): # type: (str, bytes, bool, bool) -> None
filename_b = filename.encode("utf-8") + b"\x00"
filename_b = filename.encode('utf-8') + b'\x00'
cpio_header = make_cpio_header(len(filename_b), len(data), is_trailer=trailer)
entry = pad_bytes(
struct.pack(CPIO_STRUCT, *cpio_header) + filename_b, 4
@ -186,30 +205,41 @@ class EspDfuWriter(object):
self.entries.insert(0, entry)
def action_write(args):
writer = EspDfuWriter(args['output_file'])
def action_write(args): # type: (typing.Mapping[str, typing.Any]) -> None
writer = EspDfuWriter(args['output_file'], args['pid'], args['part_size'])
for addr, f in args['files']:
print('Adding {} at {:#x}'.format(f, addr))
writer.add_file(addr, f)
writer.finish()
print('"{}" has been written. You may proceed with DFU flashing.'.format(args['output_file'].name))
if args['part_size'] % (4 * 1024) != 0:
print('WARNING: Partition size of DFU is not multiple of 4k (4096). You might get unexpected behavior.')
def main():
parser = argparse.ArgumentParser()
# Provision to add "info" command
subparsers = parser.add_subparsers(dest="command")
write_parser = subparsers.add_parser("write")
write_parser.add_argument("-o", "--output-file",
subparsers = parser.add_subparsers(dest='command')
write_parser = subparsers.add_parser('write')
write_parser.add_argument('-o', '--output-file',
help='Filename for storing the output DFU image',
required=True,
type=argparse.FileType("wb"))
write_parser.add_argument("--json",
type=argparse.FileType('wb'))
write_parser.add_argument('--pid',
required=False, # This ESP-IDF release supports one compatible target only
default=2, # ESP32-S2
type=lambda h: int(h, 16),
help='Hexa-decimal product indentificator')
write_parser.add_argument('--json',
help='Optional file for loading "flash_files" dictionary with <address> <file> items')
write_parser.add_argument("files",
metavar="<address> <file>", help='Add <file> at <address>',
nargs="*")
write_parser.add_argument('--part-size',
default=os.environ.get('ESP_DFU_PART_SIZE', 512 * 1024),
type=lambda x: int(x, 0),
help='Larger files are split-up into smaller partitions of this size')
write_parser.add_argument('files',
metavar='<address> <file>', help='Add <file> at <address>',
nargs='*')
args = parser.parse_args()
@ -236,16 +266,18 @@ def main():
files += [(int(addr, 0),
process_json_file(f_name)) for addr, f_name in iteritems(json.load(f)['flash_files'])]
files = sorted([(addr, f_name) for addr, f_name in iteritems(dict(files))],
files = sorted([(addr, f_name.decode('utf-8') if isinstance(f_name, type(b'')) else f_name) for addr, f_name in iteritems(dict(files))],
key=lambda x: x[0]) # remove possible duplicates and sort based on the address
cmd_args = {'output_file': args.output_file,
'files': files,
'pid': args.pid,
'part_size': args.part_size,
}
{'write': action_write
}[args.command](cmd_args)
if __name__ == "__main__":
if __name__ == '__main__':
main()

Binary file not shown.

View File

@ -1,7 +0,0 @@
{
"flash_files" : {
"0x8000" : "2.bin",
"0x1000" : "1.bin",
"0x10000" : "3.bin"
}
}

BIN
tools/test_mkdfu/2/dfu.bin Normal file

Binary file not shown.

View File

@ -1,7 +1,7 @@
#!/usr/bin/env python
# -*- coding: utf-8 -*-
#
# Copyright 2020 Espressif Systems (Shanghai) CO LTD
# Copyright 2020-2021 Espressif Systems (Shanghai) CO LTD
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -16,33 +16,49 @@
# limitations under the License.
from __future__ import unicode_literals
import collections
import filecmp
import json
import os
import pexpect
import shutil
import sys
import tempfile
import time
import unittest
import pexpect
current_dir = os.path.dirname(os.path.realpath(__file__))
mkdfu_path = os.path.join(current_dir, '..', 'mkdfu.py')
class TestHelloWorldExample(unittest.TestCase):
def common_test(self, add_args):
with tempfile.NamedTemporaryFile(delete=False) as f:
self.addCleanup(os.unlink, f.name)
cmd = ' '.join([sys.executable, mkdfu_path, 'write',
'-o', f.name,
add_args])
p = pexpect.spawn(cmd, timeout=10)
class TestMkDFU(unittest.TestCase):
def common_test(self, json_input=None, file_args=[], output_to_compare=None, part_size=None):
'''
- json_input - input JSON file compatible with mkdfu.py - used when not None
- file_args - list of (address, path_to_file) tuples
- output_to_compare - path to the file containing the expected output - tested when not None
- part_size - partition size - used when not None
'''
with tempfile.NamedTemporaryFile(delete=False) as f_out:
self.addCleanup(os.unlink, f_out.name)
args = [mkdfu_path, 'write',
'-o', f_out.name,
'--pid', '2']
if part_size:
args += ['--part-size', str(part_size)]
if json_input:
args += ['--json', json_input]
for addr, f_path in file_args:
args += [str(addr), f_path]
p = pexpect.spawn(sys.executable, args, timeout=10, encoding='utf-8')
self.addCleanup(p.terminate, force=True)
p.expect_exact(['Adding 1/bootloader.bin at 0x1000',
'Adding 1/partition-table.bin at 0x8000',
'Adding 1/hello-world.bin at 0x10000',
'"{}" has been written. You may proceed with DFU flashing.'.format(f.name)])
for addr, f_path in sorted(file_args, key=lambda e: e[0]):
p.expect_exact('Adding {} at {}'.format(f_path, hex(addr)))
p.expect_exact('"{}" has been written. You may proceed with DFU flashing.'.format(f_out.name))
# Need to wait for the process to end because the output file is closed when mkdfu exits.
# Do non-blocking wait instead of the blocking p.wait():
@ -53,25 +69,34 @@ class TestHelloWorldExample(unittest.TestCase):
else:
p.terminate()
self.assertTrue(filecmp.cmp(f.name, os.path.join(current_dir, '1','dfu.bin')), 'Output files are different')
if output_to_compare:
self.assertTrue(filecmp.cmp(f_out.name, os.path.join(current_dir, output_to_compare)), 'Output files are different')
class TestHelloWorldExample(TestMkDFU):
'''
tests with images prepared in the "1" subdirectory
'''
def test_with_json(self):
self.common_test(' '.join(['--json', os.path.join(current_dir, '1', 'flasher_args.json')]))
with tempfile.NamedTemporaryFile(mode='w', dir=os.path.join(current_dir, '1'), delete=False) as f:
self.addCleanup(os.unlink, f.name)
bins = [('0x1000', '1.bin'), ('0x8000', '2.bin'), ('0x10000', '3.bin')]
json.dump({'flash_files': collections.OrderedDict(bins)}, f)
self.common_test(json_input=f.name, output_to_compare='1/dfu.bin')
def test_without_json(self):
self.common_test(' '.join(['0x1000', os.path.join(current_dir, '1', '1.bin'),
'0x8000', os.path.join(current_dir, '1', '2.bin'),
'0x10000', os.path.join(current_dir, '1', '3.bin')
]))
self.common_test(file_args=[(0x1000, '1/1.bin'),
(0x8000, '1/2.bin'),
(0x10000, '1/3.bin')],
output_to_compare='1/dfu.bin')
def test_filenames(self):
temp_dir = tempfile.mkdtemp(prefix='very_long_directory_name' * 8)
self.addCleanup(shutil.rmtree, temp_dir, ignore_errors=True)
with tempfile.NamedTemporaryFile(dir=temp_dir, delete=False) as f:
output = f.name
with tempfile.NamedTemporaryFile(prefix='ľščťžýáíéěř\u0420\u043e\u0441\u0441\u0438\u044f',
dir=temp_dir,
delete=False) as f:
@ -79,20 +104,30 @@ class TestHelloWorldExample(unittest.TestCase):
shutil.copyfile(os.path.join(current_dir, '1', '1.bin'), bootloader)
cmd = ' '.join([sys.executable, mkdfu_path, 'write',
'-o', output,
' '.join(['0x1000', bootloader,
'0x8000', os.path.join(current_dir, '1', '2.bin'),
'0x10000', os.path.join(current_dir, '1', '3.bin')
])
])
p = pexpect.spawn(cmd, timeout=10, encoding='utf-8')
self.addCleanup(p.terminate, force=True)
self.common_test(file_args=[(0x1000, bootloader),
(0x8000, os.path.join(current_dir, '1', '2.bin')),
(0x10000, os.path.join(current_dir, '1', '3.bin'))])
p.expect_exact(['Adding {} at 0x1000'.format(bootloader),
'Adding 1/2.bin at 0x8000',
'Adding 1/3.bin at 0x10000',
'"{}" has been written. You may proceed with DFU flashing.'.format(output)])
class TestSplit(TestMkDFU):
'''
tests with images prepared in the "2" subdirectory
"2/dfu.bin" was prepared with:
mkdfu.py write --part-size 5 --pid 2 -o 2/dfu.bin 0 bin
where the content of "bin" is b"\xce" * 10
'''
def test_split(self):
temp_dir = tempfile.mkdtemp(dir=current_dir)
self.addCleanup(shutil.rmtree, temp_dir, ignore_errors=True)
with open(os.path.join(temp_dir, 'bin'), 'wb') as f:
self.addCleanup(os.unlink, f.name)
f.write(b'\xce' * 10)
self.common_test(file_args=[(0, f.name)],
part_size=5,
output_to_compare='2/dfu.bin')
if __name__ == '__main__':