By Gomagoti
(License CC: Some rights reserved)  

I have three of these SSD
drives and they work nicely!
I recommend them, especially
if you have a low budget.
(I'm an Amazon.com affiliate)
 

Secure Erase

Modern hard drive controllers and drivers on Linux can make use of the special feature called SECURE ERASE.

To see whether your drive supports that feature, use the hdparam command:

sudo hdparm -I /dev/sdX

where X is the partition you would like to erase (i.e. f3). If you'd like to erase the entire drive, you may want to repartition the entire drive in a single partition (see the fdisk/parted talk in the Old Method below.)

What you are interested in is the Security section. Here is that section for one of my 10Tb hard drives:

Security:
    Master password revision code = 65534
        supported
    not    enabled
    not    locked
    not    frozen
    not    expired: security count
    not    supported: enhanced erase
    1072min for SECURITY ERASE UNIT.

As we can see, the security erase would take about 1072 minutes (over 17 hours). Another important aspect, the drive must be "not frozen". If currently frozen, the security erase command will not work. A trick to have the freeze removed is to suspend the computer and wake it back up. If that doesn't work, you'll have to check your BIOS settings. The BIOS is what freezes a drive in this way.

Note: the "not enabled" means that there is no password. See below for details about that one.

The command to do a security erase on your partition is as follow:

sudo hdparm --user-master u --security-erase "password1" /dev/sdX

This command will work properly on SSD drives as well since it sends an ATA command to the controller which can then do what is required to erase all the concerned chips available on the drive.

Note that you MUST be sure that /dev/sdX is the partition you really want to erase. If you use the wrong one... you know, that's a one way street. After that erase the data is 100% gone.

"password1" is required. This is a form of security check that you're not erasing the wrong partition. By default, your partition probably has no password. You should be able to setup a password using your BIOS, but that would require rebooting (and actually some BIOS may have a hard time to boot in that situation). Instead you can use the hdparm command like so:

sudo hdparm --user-master u --security-set-pass password1 /dev/sdX

The erase will remove the password so you won't have to do anything more. Any password will do, "password1" is easy to remember.

Source: Solid State Drive Clearing

Old Method

I have three of these HDD.
That gives me about 18Tb of
space after formatting and
because I use a RAID 1 with
a couple of them. Oh and
they work as expected.
(I'm an Amazon.com affiliate)
 

Quick Reset

To low level format a hard drive, the best is to clear it with all zeroes or some random value.

One of the easiest way is to use the cat command line instruction with the zero device:

cat /dev/zero >/dev/sdb

The zero device is faster than the random device. Also we clear the entire device (/dev/sdb) and not just a partition (i.e. /dev/sdb1).

Magnetic Data is Sticky

Although some people will be able to restore the data, even after you cleared it with zeroes, they'll need some $100,000 gear to be able to do it. They better need your data really bad (as in: make at least $200,000 in return, although they may get many hard drives... but crooks don't think that way.)

To fix the problem, I suggest you use the random device instead:

cat /dev/urandom >/dev/sdb

This is neat. However, the cat instruction will continue to copy the data as long as possible, past the end of the /dev/sdb device. In other words, it will generate errors and in some case make the device unusable afterward (you'll need a reboot to access the device again.)

Reset with the Exact Size

A better tool to copy data with a known size is dd. First we need to determine the number of blocks. Using fdisk you can wipe out the existing partition table and then create one entry representing the entire hard drive (although if your drive is more than 2Tb, fdisk wouldn't work too well. Newer versions seem to have been enhanced to support very large disks properly. Otherwise consider using parted instead...) fdisk commands are one letter. It would be something like this:

fdisk /dev/sdb
m
p
d
15 to 1
[repeat d until all partitions are gone, use p to verify]
n
1
<enter>
<enter>
p
w

The m command gives you help information about all the available commands.

The p command prints the existing partition table.

The d command is used to delete a partition.

The n command is to create a new partition. It is followed by 1 (partition number) then <enter> twice to accept the default block numbers (first and last by default.)

Use p again to have a good view of the data you just saved in the partition.

The w command writes the new partition to the hard drive.

Now we're ready to use the dd command:

dd status=progress count=<see below> if=/dev/urandom of=/dev/sdb1

The count option is the number of blocks as shown by the p command. In the following partition example, it would be 97659103:

   Device   Boot  Start   End      Blocks     Id  System
/dev/sda1         1       12158    97659103+  8e  Linux

The if option defines the input file. Here I used the urandom device, although you could use /dev/zero too.

The of option defines the output "file" or block device. Notice that here I used /dev/sdb1 instead of /dev/sdb. Since the partition is the entire drive, it will delete everything and the partition itself we just destroyed it anyway.

I've read that to really clear everything what you want to do is repeat the process a few times (3 to 5 times.) If you have a very slow drive, think about it twice... Of course, it can run in the background while you do other work. In my case, I did that on a 120Gb laptop hard drive which had a mega speed of 11Mb/s. So... 120,000 / 11 = 10,909 seconds, or a little over 3 hours to clear the drive. Not only that, when you use the /dev/urandom, the computer has to come up with random data. That may be even slower than the 11Mb/s output! It took more than 5h to reset that drive to random numbers!

Current Status

dd status=progress ...

Now, the neat thing with dd is that it understands the USR1 signal and prints out progress statistics when you send it to the tool. This way  you can see whether the process is stuck.

First you need to find the pid (you could start dd in the background and use $$ but I don't recommand it.)

ps -ef | grep dd

The list that ps outputs will include dd. The first number is the process identifier (PID).

On modern Linux, you can also use one of the following:

pidof dd
pgrep -x dd

The -x command line option is important to make sure you only get information about dd and not any process that happen to include the letters dd in their name.

Using that PID number, run the kill command as in:

kill -USR1 1234

At that point you get some output from dd telling you where it's at. The first number between parenthesis is the total number of kilobytes or gigabytes. That's probably the easiest way to see how much is already done (i.e. if it says 40Gb and you have a 120Gb hard drive, then 1/3rd is done.) There is also the number of seconds it has been working on the process and the speed at which data is flowing. At the beginning it may look like it goes really fast, that's because it will bufferize using your entire memory (i.e. if you have 96Gb of RAM, then it will bufferize 90+Gb before really starting to write to the output, that will go really fast and thus give you a crazy speed...)

What's up with SSDs?

It has been a while now. To make things faster and especially to extend the lifetime of the drives (i.e. the number of writes are limited), the SSDs controllers have implemented some special behaviors to handle where data gets wrtten.

When you ask the controller to write data on block N, the value of N does not change for you, but as far as the controller is concerned it changes on each iteration. The idea is quite simple: an SSD is composed of many chips. You can do a write on any one chip. As far as the chip is concerned, this counts as aging by one more day. (this is, of course, a metaphore)

Any one chip has a maximum age; say 100,000 days. In other words, we can write to that one chip a maximum of 100,000 times before the chip is much more likely to fail (in general, manufacturers use a 20% gap. In other words, if they tell you that it can survive 100,000 writes, it is likely that the chip supports 120,000. This makes sure that the 100,000th write works as expected and you will be able to re-read your data.)

Once all the chips were written 100,000 times, your drive should only be used in read-only mode.

Of course, as you use your drive some blocks get written once and never get modified (i.e. you install your OS once and never upgrade). So those may get a very small age and sit there. On the other hand, some other blocks may be overwritten all the time. Say you have many services writing logs, those are going to be written, deleted, overwritten, deleted, etc. In that case, a new block is going to be used for each write and it will rotate through your drive. However, as far as the OS is concerned, the blocks have the same number, yet the controller uses a different chip each time.

As a result the OS has no means to say "I want to write data on this very chip". It can only say "overwrite block N with this new data". So for the purpose of clearing a Drive, the controller will prevent us from doing so. If many of the chips have an age of 50,000 and one has an age of 10 or so, most of your write are going to happen on the one age 10 if the controller is given the chance of reusing that one chip. It is very likely that cat or dd will not be able to overwrite the entire disk and it will just make some of the chips look much older for no good reasons.