Wiping Corrupt Drives with dd: The Nuclear Option
Until a couple months ago, I had never actually damaged a drive by unplugging it while it was busy. But boy, when you pull the plug at the wrong time, do things go wrong fast. I bricked a drive so completely that it would not appear in Disk Utility at all, and trying to poke at it with diskutil (including the full range of mount/unmount/repair/verify/partition/erase subcommands) would only ever respond with “resource busy”. It was also impossible to eject and would leave zombie copies of itself around in diskutil list. Fun!
After some sleuthing around the internet and asking people who know more than me, I landed on the following script.
Don't Try This at Home!
Don't run this script without knowing what you're doing. If you destroy your drive/data, it's not my fault.
while [ 1 ]; do
sudo dd if=/dev/zero of=/dev/disk6 bs=1000000 count=1000
done
This script loops forever, writing 1000000 * 1000 = 1000000000 bytes (a gigabyte, in the true sense) of zeroes to the beginning of /dev/disk6. Since partition tables are located near the beginning of the disk, this will trash them and effectively “factory reset” the drive.
To run this script, first determine what disk identifier the corrupted disk will have when you plug it in. I did this by plugging and unplugging it a few times to verify it always got assigned the same identifier (well, when it wasn’t leaving zombie disks behind). Put that identifier into the script in place of /dev/disk6.
Run the modified script before the disk is plugged in. That’ll prompt for the first sudo, then start looping (and failing) forever. Plug in the corrupted disk and wait. A successful run will be a lot quieter than the wall of errors you get while the disk isn’t plugged in, and when it finishes, you can kill the script.
After running this, then unplugging and replugging the drive, Disk Utility found an empty disk and was happy to partition it for me.
A couple notes on the implementation.
First, the infinite loop is used to grab the disk before the higher-level parts of the operating system can. The lower-level parts are demonstrably working (hence the presence of /dev/disk6), but the higher-level parts get stuck forever trying to mount it (or something of the sort), failing, and trying again, thus preventing anything else from accessing the disk.
Second, dd is being used because the disk is unmountable and we specifically want to wipe the entire thing, partition tables included. This is a different, much harsher version of wiping a disk than what one normally does.