2020-04-02 01:32:47 +00:00
|
|
|
#!/usr/bin/env python
|
|
|
|
#
|
2020-04-06 14:41:44 +00:00
|
|
|
# Copyright 2020 Espressif Systems (Shanghai) PTE LTD
|
|
|
|
#
|
|
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
|
|
# you may not use this file except in compliance with the License.
|
|
|
|
# You may obtain a copy of the License at
|
|
|
|
#
|
|
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
#
|
|
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
# See the License for the specific language governing permissions and
|
|
|
|
# limitations under the License.
|
|
|
|
#
|
2020-04-02 01:32:47 +00:00
|
|
|
# This program creates archives compatible with ESP32-S* ROM DFU implementation.
|
|
|
|
#
|
|
|
|
# The archives are in CPIO format. Each file which needs to be flashed is added to the archive
|
|
|
|
# as a separate file. In addition to that, a special index file, 'dfuinfo0.dat', is created.
|
|
|
|
# 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
|
2020-04-06 14:41:44 +00:00
|
|
|
from future.utils import iteritems
|
|
|
|
import argparse
|
2020-04-02 01:32:47 +00:00
|
|
|
import hashlib
|
2020-04-06 14:41:44 +00:00
|
|
|
import json
|
2020-04-02 01:32:47 +00:00
|
|
|
import os
|
|
|
|
import struct
|
|
|
|
import zlib
|
|
|
|
|
|
|
|
try:
|
|
|
|
import typing
|
|
|
|
except ImportError:
|
|
|
|
# Only used for type annotations
|
|
|
|
pass
|
|
|
|
|
2020-04-06 14:41:44 +00:00
|
|
|
try:
|
|
|
|
from itertools import izip as zip
|
|
|
|
except ImportError:
|
|
|
|
# Python 3
|
|
|
|
pass
|
|
|
|
|
2020-04-02 01:32:47 +00:00
|
|
|
# CPIO ("new ASCII") format related things
|
|
|
|
CPIO_MAGIC = b"070701"
|
|
|
|
CPIO_STRUCT = b"=6s" + b"8s" * 13
|
|
|
|
CPIOHeader = namedtuple(
|
|
|
|
"CPIOHeader",
|
|
|
|
[
|
|
|
|
"magic",
|
|
|
|
"ino",
|
|
|
|
"mode",
|
|
|
|
"uid",
|
|
|
|
"gid",
|
|
|
|
"nlink",
|
|
|
|
"mtime",
|
|
|
|
"filesize",
|
|
|
|
"devmajor",
|
|
|
|
"devminor",
|
|
|
|
"rdevmajor",
|
|
|
|
"rdevminor",
|
|
|
|
"namesize",
|
|
|
|
"check",
|
|
|
|
],
|
|
|
|
)
|
|
|
|
CPIO_TRAILER = "TRAILER!!!"
|
|
|
|
|
|
|
|
|
|
|
|
def make_cpio_header(
|
|
|
|
filename_len, file_len, is_trailer=False
|
|
|
|
): # type: (int, int, bool) -> CPIOHeader
|
|
|
|
""" Returns CPIOHeader for the given file name and file size """
|
|
|
|
|
|
|
|
def as_hex(val): # type: (int) -> bytes
|
|
|
|
return "{:08x}".format(val).encode("ascii")
|
|
|
|
|
|
|
|
hex_0 = as_hex(0)
|
|
|
|
mode = hex_0 if is_trailer else as_hex(0o0100644)
|
|
|
|
nlink = as_hex(1) if is_trailer else hex_0
|
|
|
|
return CPIOHeader(
|
|
|
|
magic=CPIO_MAGIC,
|
|
|
|
ino=hex_0,
|
|
|
|
mode=mode,
|
|
|
|
uid=hex_0,
|
|
|
|
gid=hex_0,
|
|
|
|
nlink=nlink,
|
|
|
|
mtime=hex_0,
|
|
|
|
filesize=as_hex(file_len),
|
|
|
|
devmajor=hex_0,
|
|
|
|
devminor=hex_0,
|
|
|
|
rdevmajor=hex_0,
|
|
|
|
rdevminor=hex_0,
|
|
|
|
namesize=as_hex(filename_len),
|
|
|
|
check=hex_0,
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
# 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"
|
|
|
|
# Structure which gets added at the end of the entire DFU file
|
|
|
|
DFUSUFFIX_STRUCT = b"<H H H H 3s B"
|
|
|
|
DFUSuffix = namedtuple(
|
|
|
|
"DFUSuffix", ["bcd_device", "pid", "vid", "bcd_dfu", "sig", "len"]
|
|
|
|
)
|
|
|
|
ESPRESSIF_VID = 12346
|
|
|
|
# This CRC32 gets added after DFUSUFFIX_STRUCT
|
|
|
|
DFUCRC_STRUCT = b"<I"
|
|
|
|
|
|
|
|
|
|
|
|
def dfu_crc(data, crc=0): # type: (bytes, int) -> int
|
|
|
|
""" Calculate CRC32/JAMCRC of data, with an optional initial value """
|
|
|
|
uint32_max = 0xFFFFFFFF
|
|
|
|
return uint32_max - (zlib.crc32(data, crc) & uint32_max)
|
|
|
|
|
|
|
|
|
|
|
|
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):
|
2020-05-26 15:35:40 +00:00
|
|
|
def __init__(self, dest_file, pid): # type: (typing.BinaryIO) -> None
|
2020-04-02 01:32:47 +00:00
|
|
|
self.dest = dest_file
|
2020-05-26 15:35:40 +00:00
|
|
|
self.pid = pid
|
2020-04-02 01:32:47 +00:00
|
|
|
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())
|
|
|
|
|
|
|
|
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])
|
|
|
|
self._add_cpio_entry(DFUINFO_FILE, dfuinfo, first=True)
|
|
|
|
|
|
|
|
# Add CPIO archive trailer
|
|
|
|
self._add_cpio_entry(CPIO_TRAILER, b"", trailer=True)
|
|
|
|
|
|
|
|
# Combine all the entries and pad the file
|
|
|
|
out_data = b"".join(self.entries)
|
|
|
|
cpio_block_size = 10240
|
|
|
|
out_data = pad_bytes(out_data, cpio_block_size)
|
|
|
|
|
|
|
|
# Add DFU suffix and CRC
|
2020-05-26 15:35:40 +00:00
|
|
|
dfu_suffix = DFUSuffix(0xFFFF, self.pid, ESPRESSIF_VID, 0x0100, b"UFD", 16)
|
|
|
|
out_data += struct.pack(DFUSUFFIX_STRUCT, *dfu_suffix)
|
2020-04-02 01:32:47 +00:00
|
|
|
out_data += struct.pack(DFUCRC_STRUCT, dfu_crc(out_data))
|
|
|
|
|
|
|
|
# Finally write the entire binary
|
|
|
|
self.dest.write(out_data)
|
|
|
|
|
|
|
|
def _add_cpio_flash_entry(
|
|
|
|
self, filename, flash_addr, data
|
|
|
|
): # type: (str, int, bytes) -> None
|
|
|
|
md5 = hashlib.md5()
|
|
|
|
md5.update(data)
|
|
|
|
self.index.append(
|
|
|
|
DFUInfo(
|
|
|
|
address=flash_addr,
|
|
|
|
flags=0,
|
|
|
|
name=filename.encode("utf-8"),
|
|
|
|
md5=md5.digest(),
|
|
|
|
)
|
|
|
|
)
|
|
|
|
self._add_cpio_entry(filename, data)
|
|
|
|
|
|
|
|
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"
|
|
|
|
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
|
|
|
|
) + pad_bytes(data, 4)
|
|
|
|
if not first:
|
|
|
|
self.entries.append(entry)
|
|
|
|
else:
|
|
|
|
self.entries.insert(0, entry)
|
|
|
|
|
|
|
|
|
|
|
|
def action_write(args):
|
2020-05-26 15:35:40 +00:00
|
|
|
writer = EspDfuWriter(args['output_file'], args['pid'])
|
2020-04-06 14:41:44 +00:00
|
|
|
for addr, f in args['files']:
|
|
|
|
print('Adding {} at {:#x}'.format(f, addr))
|
|
|
|
writer.add_file(addr, f)
|
2020-04-02 01:32:47 +00:00
|
|
|
writer.finish()
|
2020-04-06 14:41:44 +00:00
|
|
|
print('"{}" has been written. You may proceed with DFU flashing.'.format(args['output_file'].name))
|
2020-04-02 01:32:47 +00:00
|
|
|
|
|
|
|
|
|
|
|
def main():
|
2020-04-06 14:41:44 +00:00
|
|
|
parser = argparse.ArgumentParser()
|
2020-04-02 01:32:47 +00:00
|
|
|
|
|
|
|
# Provision to add "info" command
|
|
|
|
subparsers = parser.add_subparsers(dest="command")
|
|
|
|
write_parser = subparsers.add_parser("write")
|
2020-04-06 14:41:44 +00:00
|
|
|
write_parser.add_argument("-o", "--output-file",
|
|
|
|
help='Filename for storing the output DFU image',
|
|
|
|
required=True,
|
|
|
|
type=argparse.FileType("wb"))
|
2020-05-26 15:35:40 +00:00
|
|
|
write_parser.add_argument("--pid",
|
|
|
|
required=True,
|
|
|
|
type=lambda h: int(h, 16),
|
|
|
|
help='Hexa-decimal product indentificator')
|
2020-04-06 14:41:44 +00:00
|
|
|
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="*")
|
2020-04-02 01:32:47 +00:00
|
|
|
|
|
|
|
args = parser.parse_args()
|
2020-04-06 14:41:44 +00:00
|
|
|
|
|
|
|
def check_file(file_name):
|
|
|
|
if not os.path.isfile(file_name):
|
|
|
|
raise RuntimeError('{} is not a regular file!'.format(file_name))
|
|
|
|
return file_name
|
|
|
|
|
|
|
|
files = []
|
|
|
|
if args.files:
|
|
|
|
files += [(int(addr, 0), check_file(f_name)) for addr, f_name in zip(args.files[::2], args.files[1::2])]
|
|
|
|
|
|
|
|
if args.json:
|
|
|
|
json_dir = os.path.dirname(os.path.abspath(args.json))
|
|
|
|
|
|
|
|
def process_json_file(path):
|
|
|
|
'''
|
|
|
|
The input path is relative to json_dir. This function makes it relative to the current working
|
|
|
|
directory.
|
|
|
|
'''
|
|
|
|
return check_file(os.path.relpath(os.path.join(json_dir, path), start=os.curdir))
|
|
|
|
|
|
|
|
with open(args.json) as f:
|
|
|
|
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))],
|
|
|
|
key=lambda x: x[0]) # remove possible duplicates and sort based on the address
|
|
|
|
|
|
|
|
cmd_args = {'output_file': args.output_file,
|
|
|
|
'files': files,
|
2020-05-26 15:35:40 +00:00
|
|
|
'pid': args.pid,
|
2020-04-06 14:41:44 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
{'write': action_write
|
|
|
|
}[args.command](cmd_args)
|
2020-04-02 01:32:47 +00:00
|
|
|
|
|
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
|
main()
|