Hi guys,

I have a file that has few info in it that I need to execute in bash... Here's the details:

I am using sfdisk to automate partitioning of hard disk. To do this, I extract the information I need (start and sizes of each existing partition) from sfdisk -l and write them in a .txt file in the form of:

start,size,L

which is easy input for sfdisk. I can't use sfdisk -d output file for many reasons which are irrelevant for this topic.
What I want to do is, to get the inputs to sfdisk from this .txt file and execute it in bash after I call sfdisk /dev/hda << EOF and before EOF

Here's the sample contents of my .txt file:

and so, what I want is to execute:

can someone help me figure out how to do this please?

Thanks

Hi,

Try this: sfdisk /dev/hda < infile.txt

Using a hear-document structure when your input is in a file is not the way to go. You can probably get it to work, but it would be the hard way.

I hope you have backups, experimenting with sfdisk can have 'funny' results and typos are made by the best of us ;-)

Alright, that scared me! You have any other suggestions?

Hi,

Not really, except the 'obvious' stuff:

- Be familiar with the commands you are using (especially if you start automating them).

- Have backups. In this case you're targeting a complete disk, everything on it will be gone. Check out the sfdisk manpage and look for the -O and -I options. Nonetheless, making a typo (hdb instead of hda) could destroy more important stuff.

- If at all possible test the script(s) in a safe environment (or as safe as possible). Most of the time you encounter one or more errors the first time you run a (larger) script. These errors can potentionally be harmful (from a deleted file to wiping a harddisk). In this case you need to be 'extra' careful, sfdisk is a command that needs to be run as root.

If you have a 'spare' device around that can be partitioned (usb stick/disk?), use that as your test harddisk while you write/test/debug the script and make very sure sfdisk is given the correct device!

Great! Thanks for the info! Very useful. I actually, am testing my script on a virtual machine running Knoppix 5.1.1 . I figured, using virtual machine and virtual disk, I can do anything without fearing of these side effects of buggy scripts...

Thanks for the comments!

A HERE document is used for documents embedded in your script. Suppose that you want to write an xml file from a template. You would enter the document into the bash script after the line with the <<EOF. The lines can include variables which are expanded. Before the package systems we use today, software was distributed in a shell script that contained a here document for each file to produce.

If I were to use fdisk in a script, I would use the -u option so that the begin and end values are based on 512 byte blocks. This eliminates rounding errors when you read the disk. For example, if I had an image file of a disk, I would use the fdisk start value to determine the offset for losetup to be able to mount the file system.

I'm not at my Linux machine right now, so I can't bring up the cfdisk manpage to see if there is a similar problem.

One thing you might want to look at is the set command to debug your script. After assigning values to variables from a line of data, instead of calling cfdisk, use something like "set $OPTION $START $END $DEVICE" instead and examine the values of $1, $2, $3 & $4 to make sure they are what you expected. The set command will setup the argc and argv array that the cfdisk command would see.

sfdisk not targeting complete disk it's change only MBR (or more precisely partition table in the MBR) so dd if=/dev/sda of=backup bs=512 count=1 will be enough.

you could add
#!sfdisk /dev/hda
as the first line then chmod +x the file and run it as a script
but for some thing like this I would do it by hand