esp-idf/components/fatfs/fatfsgen_utils/utils.py

146 lines
4.9 KiB
Python
Raw Normal View History

# SPDX-FileCopyrightText: 2021-2022 Espressif Systems (Shanghai) CO LTD
2021-09-21 18:32:54 -04:00
# SPDX-License-Identifier: Apache-2.0
import argparse
import binascii
2021-09-21 18:32:54 -04:00
import os
import uuid
from typing import List, Optional, Tuple
2021-09-21 18:32:54 -04:00
from construct import Int16ul
FAT12_MAX_CLUSTERS: int = 4085
FAT16_MAX_CLUSTERS: int = 65525
FAT12: int = 12
FAT16: int = 16
FAT32: int = 32
BYTES_PER_DIRECTORY_ENTRY = 32
2021-09-21 18:32:54 -04:00
def crc32(input_values: List[int], crc: int) -> int:
"""
Name Polynomial Reversed? Init-value XOR-out
crc32 0x104C11DB7 True 4294967295 (UINT32_MAX) 0xFFFFFFFF
"""
return binascii.crc32(bytearray(input_values), crc)
def number_of_clusters(number_of_sectors: int, sectors_per_cluster: int) -> int:
return number_of_sectors // sectors_per_cluster
def get_non_data_sectors_cnt(reserved_sectors_cnt: int, sectors_per_fat_cnt: int, root_dir_sectors_cnt: int) -> int:
return reserved_sectors_cnt + sectors_per_fat_cnt + root_dir_sectors_cnt
def get_fatfs_type(clusters_count: int) -> int:
if clusters_count < FAT12_MAX_CLUSTERS:
return FAT12
if clusters_count < FAT16_MAX_CLUSTERS:
return FAT16
return FAT32
def required_clusters_count(cluster_size: int, content: bytes) -> int:
2021-09-21 18:32:54 -04:00
# compute number of required clusters for file text
return (len(content) + cluster_size - 1) // cluster_size
def generate_4bytes_random() -> int:
return uuid.uuid4().int & 0xFFFFFFFF
def pad_string(content: str, size: Optional[int] = None, pad: int = 0x20) -> str:
2021-09-21 18:32:54 -04:00
# cut string if longer and fill with pad character if shorter than size
return content.ljust(size or len(content), chr(pad))[:size]
def split_to_name_and_extension(full_name: str) -> Tuple[str, str]:
2021-09-21 18:32:54 -04:00
name, extension = os.path.splitext(full_name)
return name, extension.replace('.', '')
def is_valid_fatfs_name(string: str) -> bool:
return string == string.upper()
def split_by_half_byte_12_bit_little_endian(value: int) -> Tuple[int, int, int]:
2021-09-21 18:32:54 -04:00
value_as_bytes = Int16ul.build(value)
return value_as_bytes[0] & 0x0f, value_as_bytes[0] >> 4, value_as_bytes[1] & 0x0f
def build_byte(first_half: int, second_half: int) -> int:
return (first_half << 4) | second_half
def clean_first_half_byte(bytes_array: bytearray, address: int) -> None:
"""
the function sets to zero first four bits of the byte.
E.g. 10111100 -> 10110000
"""
bytes_array[address] &= 0xf0
def clean_second_half_byte(bytes_array: bytearray, address: int) -> None:
"""
the function sets to zero last four bits of the byte.
E.g. 10111100 -> 00001100
"""
bytes_array[address] &= 0x0f
def split_content_into_sectors(content: bytes, sector_size: int) -> List[bytes]:
2021-09-21 18:32:54 -04:00
result = []
clusters_cnt = required_clusters_count(cluster_size=sector_size, content=content)
for i in range(clusters_cnt):
result.append(content[sector_size * i:(i + 1) * sector_size])
return result
def get_args_for_partition_generator(desc: str) -> argparse.Namespace:
parser = argparse.ArgumentParser(
description=desc)
parser.add_argument('input_directory',
help='Path to the directory that will be encoded into fatfs image')
parser.add_argument('--output_file',
default='fatfs_image.img',
help='Filename of the generated fatfs image')
parser.add_argument('--partition_size',
default=1024 * 1024,
help='Size of the partition in bytes')
parser.add_argument('--sector_size',
default=4096,
type=int,
choices=[512, 1024, 2048, 4096],
help='Size of the partition in bytes')
parser.add_argument('--sectors_per_cluster',
default=1,
type=int,
choices=[1, 2, 4, 8, 16, 32, 64, 128],
help='Number of sectors per cluster')
parser.add_argument('--root_entry_count',
default=512,
help='Number of entries in the root directory')
parser.add_argument('--fat_type',
default=0,
type=int,
choices=[12, 16, 0],
help="""
Type of fat. Select 12 for fat12, 16 for fat16. Don't set, or set to 0 for automatic
calculation using cluster size and partition size.
""")
args = parser.parse_args()
if args.fat_type == 0:
args.fat_type = None
args.partition_size = int(str(args.partition_size), 0)
if not os.path.isdir(args.input_directory):
raise NotADirectoryError(f'The target directory `{args.input_directory}` does not exist!')
return args
def read_filesystem(path: str) -> bytearray:
with open(path, 'rb') as fs_file:
return bytearray(fs_file.read())