Linux @lemmy.ml OwlPaste @lemmy.world 6d ago

What are common tools on Linux to wipe an HDD pre RMA?

A Qnap NAS has a drive with some bad sectors, I want to RMA it, but before just want to figure out how to prepare a drive? It's part of a raid 5 setup of 4 drives unencrypted. So I want to remove it and wipe it. Got a Linux machine I can use, but never done this before.

What are common Linux tools to do that sensibly?

38 comments

Since dipshits are down voting me for being right, I'll just leave this here

https://skeptics.stackexchange.com/questions/13674/is-it-possible-to-recover-data-on-a-zeroed-hard-drive
- Thank you thats useful to know!
shred

Can also be used from KDE Partition Manager.

Use LUKS encryption in the future.
- I assume in this context the benefit using LUKS is that even if someone does recover data from the drive, they'd just get encrypted data which is unreadable? If so, I hadn't considered that before!
sudo dd if=/dev/urandom of=/dev/[sdx] bs=4096K status=progress

Or for multiple passes:

sudo shred -fzv /dev/[sdx]

Change [sdx] to the drive you want to wipe, make sure you double check it's the right one.
shred or alternatively you can zero out all the bytes in a drive with dd if=/dev/zero of=/dev/<DRIVE>
dd if=/dev/zero of=/dev/(your drive)

You can do status=progress if you want like someone else posted and if you pick a block size go with either the physical block size reported by the disk in smartcontrol or some multiple of it that coincides with a big even division of your controllers memory. The drives physical block size will be “easy” for the drive, bigger blocks are faster.

People saying physical destruction are operating in a different world than you and people saying urandom or shred are operating off old (>30 years) information. The same technology that makes ssds unrecoverable black boxes was originally developed and deployed in spinning drives to eek out speed gains because the disk itself can be expected to know better than the operating system where to put shit and makes techniques (which were postulated but never actually implemented successfully in the wild) to recover overwritten data infeasible.

Alternately just reformat it and don’t worry. No one doing drive rmas cares about your data. They’re already on the razors edge with feedback and customer trust, you think they’re gonna burn their above board bread and butter to run a harvesting operation for a few bucks on the side? That’s usually the purview of your local pc repair shop…
dd if=/dev/zero of=/dev/disk
- I’d recommend /dev/urandom instead of /dev/zero
  
  Use shred , it will automate multiple random passes, and finish with a zero pass.
  
  Yeh, you're right
- This is what I was gonna say
no help to you, but a heads-up to anybody yet to deploy disks in such a scenario: always use encryption by way of LUKS2. you can set it up easily to unlock it on boot by a key file on the boot drive, thumb drive, TPM and such. so when a drive gets sold, RMA'd, etc., you got none of these issues.

source: sold my old drives recently and the shred procedure took ages. the new ones are encrypted so none of that shit no more.
https://linux.die.net/man/1/wipe

https://dban.org/
Assuming the drive writes normally a simple command like

dd if=/dev/random of=/dev/sdX

Where sdX is the location of the drive should do the trick. Depending on drive time this may take a bit.
- Yeah my first thought was just keep running dd commands, and sooner or later you'll have the hdd wiped.
  
  Instructions unclear, accidentally deleted 200 EB of irrecoverable NASA data.
- Just keep in mind that you can't wipe the bad sectors that have been remapped. That's unlikely to be an issue for a personal drive, but something to consider if it held particularly sensitive information.
  
  Sectors are what, 4kb? The odds of something sensitive being in those each of those 4kb is low (but never zero).
  
  If the drive implements the ATA secure erase command properly, it should also clear all the remapped sectors too. But I don't know if I trust them to implement it properly. Maybe try it and inspect the disk with a hex editor.
- Always use /dev/urandom for this purposes. /dev/random will be locked if it doesn't have enough entropy. It is good for getting some random kilobytes for cryptography but not 2 TB of random data for disk wipe.
  
  Thanks for the heads up!
- dd if=/dev/zero of=/dev/sdX bs=1M
  
  This command is much faster. Instead of random bits, it just marks everything zero (dude). Is good enough.
I like badblocks in destructive mode. It can also do multiple rounds of overwriting. It is also a good tool to burn in a new hdd or test a used one. just check smartctl bad sector count, run it, check again if it increased.
When I sold my drives, I used veracrypt with a 128 character password and PIM of 800+.

Isn't that the same thing as shredding?
- No.
  
  Can you elaborate?
If the drive has bad sectors that it can't read right now, it likely had other sectors that were marginal and got copied (remapped) to new spare sectors before they became unreadable. So there is still potentially recoverable data in the remapped sectors, and not much you can do about it.

Basically, writing zeros to the disk is about as good as you can hope for. If your data is s00per seekrit to the point where you can't stand the possibility of any bits at all being recovered, you basically have to melt the drive. Otherwise, zero it and send it in.

Next time, set up encryption ahead of time, so your new drives never see plaintext. Some drives have a "secure erase" feature that is basically a crappy version of this built into the drive.
The day job is Private Possum territory, so our agreement is

we won't send in drives

we will send proof of destruction

and I'm sure it costs more, but we don't have to worry about recovered user data.

Our stuff is also in those crazy-secure datacenters, too, so we don't need to mess with crypto-at-rest.
DNAM. Is or used to be on the UBCD.

For the future remember, encryption helps when the disk is no longer operational.

38 comments