Do you ever try to reformat / repartition disks and Linux complains the kernel is still using the old table and you have to REBOOT (a sin on any *NIX) to fix this?
The error is “WARNING: Re-reading the partition table failed with error 16: Device or resource busy.
The kernel still uses the old table.
The new table will be used at the next reboot.”

This annoys me, what was happening on my Ubuntu 7.10 computer was I had a disk that used to be Linux and I put it into a USB / Firewire enclosure and fdisk’d it and then formatted it, but the automounter sees the partition exists as soon as the kernel rereads the partition table and mounts it. This isn’t safe or a good idea really, so, if you’re going to nuke a disk you should zero out any previous partitions by nuking the master partition table.

THIS IS A POTENTIALLY DANGEROUS OPERATION, YOU PROCEED AT YOUR OWN RISK, I WILL NOT ASSUME LIABILITY FOR LOST DATA OR ANY DAMAGES.

Ok, so make 100% absolutely sure you know the drive letter you’re getting with something like

dmesg | tail

and you’ll see something like

sd 18:0:0:0: [sdf] 488397168 512-byte hardware sectors (250059 MB)
[3010413.606563] sd 18:0:0:0: [sdf] Test WP failed, assume Write Enabled
[3010413.606568] sd 18:0:0:0: [sdf] Assuming drive cache: write through
[3010413.607559] sd 18:0:0:0: [sdf] 488397168 512-byte hardware sectors (250059 MB)
[3010413.608807] sd 18:0:0:0: [sdf] Test WP failed, assume Write Enabled
[3010413.608811] sd 18:0:0:0: [sdf] Assuming drive cache: write through
[3010413.608815] sdf: sdf1 sdf2 < sdf5 >
[3010413.651837] sd 18:0:0:0: [sdf] Attached SCSI disk
[3010413.651880] sd 18:0:0:0: Attached scsi generic sg5 type 0
[3010415.111787] kjournald starting. Commit interval 5 seconds

Now run this (as root), substitude sdf for the device you found. BE WARNED that if Linux didn’t see your disk when you just plugged it in, you may be seeing the detection of an actual file system disk. The best way to tell is to run tail -f /var/log/kern.log (or /var/log/messages or something where the kernel logs, depending on your distribution of Linux)

fdisk -l /dev/sdf

This showed me

Disk /dev/sdf: 250.0 GB, 250059350016 bytes

255 heads, 63 sectors/track, 30401 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Disk identifier: 0x000969e0

Device Boot Start End Blocks Id System
/dev/sdf1 * 1 30024 241167748+ 83 Linux
/dev/sdf2 30025 30401 3028252+ 5 Extended
/dev/sdf5 30025 30401 3028221 82 Linux swap / Solaris

Now I check df -h and see that Ubunu already mounted /dev/sdf1 on /media/disk
So I (as root) unmount it with
umount /dev/sdf1

Now I zero the partition table, THIS CAN TOTALLY MESS UP YOUR DISK IN AN UNRECOVERABLE FASHION, SO BE REALLY REALLY REALLY SURE YOU HAVE THE CORRECT DISK DEVICE NODE. IF YOU ARE UNSURE DO NOT PROCEED.
(More info about the master boot record on the wikipedia article here. )

dd if=/dev/zero of=/dev/sdf bs=512 count=1

Now, proceed with creating the partition(s) and formatting
fdisk /dev/sdf
select n,p,enter,enter,w
mkfs.ext3 /dev/sdf1