mke2fs: Bad magic number in super-block

dr_berta · 03-17-2023, 09:18 AM

Hello,
I have created a script to automatically create two partitions on a disk and to create an ext4 filesystem on the disk sda1 and the swap disk on sda2.

The proble I have is that the tool mke2fs, used to create the filesystem on sda1, gives me a "Bad magic number in super-block error" if executed immediately after the disk partitioning.

If I put in the script a read command (see the blue commands below), asking for a enter key pressed, everything goes well and the error is not present.

This is the code showing the issue:

Code:

echo -e "Making new partitions..."
# make partitions
DEVICE="/dev/sda"

sfdisk $DEVICE < /mnt/cdrom/tbs/.sda.sfdisk
sfdisk -qV $DEVICE
RESULT=$?
if [ $RESULT -ne 0 ] ; then
	echo "Error creating partitions!"
	exit
fi

echo "enter return to continue"
read ANS

echo -e "Reloading partition table..."
partprobe $DEVICE
sleep 1

# create filesystems
BIG="16000"
VERYBIG="2000000000"
if [ $PLATFORM = "64bit" ]; then
	mke2fs -t ext4 -L "slash" "${DEVICE}1"
	tune2fs -c $BIG -i $VERYBIG "${DEVICE}1"
	mkswap -L "swap" "${DEVICE}2"
else
	echo "Error!!!"
	exit
fi

Is there anyone that knows the reason of this strange behavior? I would like to remove the "enter key press" request

Thanks
Claudio

xode · 03-18-2023, 02:18 AM

This is a wild guess: without the read command between sfdisk and partprobe, the changes you made with sfdisk haven't yet been written to disk before partprobe again reads the partition, and therefore partprobe would be reading the old partition data. What happens if you put

Code:

sleep 5

between sfdisk and partprobe instead of the read command?

dr_berta · 03-18-2023, 06:42 AM

I tried two different solutions to skip the need of a human interaction:

1)

Code:

sleep 3

2)

Code:

echo "Yes" | read ANS

Both of them have been useless.

To tell the truth, sfdisk, at the and of the disk partitioning, sends a message saying:

Code:

reloading partition table. 
syncing disks.

After this command, I do the check of the new partition checking the result.

Thus I was thinking that after the control the writing of the new disk partition was finished succesfully.

Claudio