Quicker Backup of SD Cards

Introduction

This article shows a script that copies an SD card to hard drive using efficient reading and minimizing the amount of data read.

The ideas are two: only copy the area that is used (i.e. partitioned with non-swap partitions), and do the reads in larger chunks than the typical erase block.

Using fdisk -lcu /dev/sdX the partitions are listed in number of sectors of 512 bytes. Naively reading a large count of 512 byte blocks in a dd command could lead to a many small reads and be potentially very inefficient. Instead the script uses the following observation:
reading N blocks of 512 bytes, equals reading 512 blocks of N bytes. It is perfectly possible to swap the block size and the count parameters in a dd command.

Script

The script is as below. It takes two arguments: block device of the sd card and the filename for the copied SD card image.

    #!/bin/sh
     
    dev=$1
    outfile=$2
    if [ X$dev = X ] || [ Y$outfile = Y ]
    then
            echo Usage : $0 device filename
            exit
    fi
     
    devminor=`basename $dev`
    bs=`echo pq | fdisk -u $dev | grep -v 'swap' | grep $devminor | tail -1 | awk '{print $3 }'`
    bs=`expr $bs + 1`
    dd if=$dev of=$outfile bs=$bs count=512 2>&1 > /dev/null