Why does dd give me an error at the end of zero-writing a disk?

Question

I'm running Cygwin on Windows 7.

I was using the following dd command to securely erase a 1TB mechanical disk:

dd if=/dev/zero of=/dev/sde bs=4M status=progress

About 2 and a half hours into the wipe, around the time I was expecting it to complete and after successfully writing about 930GB (or GiB, more likely) to the disk, I got the error:

dd: error writing '/dev/sde': Input/output error

I ran the command again with seek to try and reproduce the error by zero-writing the last few gigabytes of the file:

dd if=/dev/zero of=/dev/sde bs=4M status=progress seek=231100

...and sure enough I got the same error:

dd: error writing '/dev/sde': Input/output error
7368+0 records in
7367+0 records out
35196174335 bytes (35 GB, 33 GiB) copied, 405.703 s, 86.8 MB/s

It seems that the wipe ran successfully, but if that was the case, then why is the error being generated?

Is it normal? If not, how can I avoid it?

Answer 1

My guess would be a bad sector. AFAIK that's one of the cases that could lead to I/O error.

Anyway, it is unlikely that the error is given is because dd was trying to write past the end of the drive. Even with my answer to your another question, you should still see No space left on device instead:

For the record, because of the Cygwin bug, 35196174335 is unlikely the actual number of bytes that have been written. Instead, it should be 35196174335 - 4294967295 = 30901207040. (You can try to "mod" these numbers against 512.)

In that case, the number of bytes / sectors that have been successfully written should be 231100 * 4 * 1024 ^ 2 + 30901207040 = 1000204861440 / 1953525120. Base on other information that you have given in your questions, this is unlikely the size of the entire drive anyway. (It might be worth mentioning that 30901207040 bytes is not "4M" (4 * 1024 ^ 2) divisible either.)

Answer 2

On further investigation this appears to be a Cygwin bug that prevents the last X sectors of a drive from wiped, although it's possible that this bug is entirely unrelated to the error message.

I tested by first writing a single pass of ones to the entire disk with:

badblocks -t "1" -vws /dev/sda

I verified that this was successful on the disk by running:

od /dev/sda

I then did:

dd if=/dev/zero of=/dev/sda bs=4M status=progress

...on the entire disk. This failed with the typical I/O error. Inspecting the disk using a hex editor shows that the entire drive was successfully zeroed apart from the last 48 sectors:

I worked out that 1953525168 sectors - 48 = 1953525120 sectors, and used this figure in combination with seek to wipe the last 48 sectors of the disk:

$ dd if=/dev/zero of=/dev/sda seek=1953525120
dd: writing to '/dev/sda': Input/output error
49+0 records in
48+1 records out
4294991871 bytes (4.3 GB, 4.0 GiB) copied, 0.230269 s, 18.7 GB/s

This successfuly zeroes the disk from sector 48 all the way until the end, which I again confirmed using a hex editor and od /dev/sda. Note the I/O error still exists when it reaches the end, indicating that the error does not necessarily indicate a bad wipe.

Finally, I booted Linux Kali from a bootable USB and wiped the entire disk with the same command:

dd if=/dev/zero of=/dev/sda bs=4M status=progress

Checking the wiped disk with a hex editor again confirms that the command wipes the entirety of the disk with no non-zero characters remaining.

In conclusion:

• Running dd on the entire disk in Cygwin prevents the last 48 sectors from being wiped

• Using seek to selectively wipe the disk wipes the full disk, but still causes the I/O error

• Wiping the entire disk from a bootable Linux system wipes the whole disk without the I/O error

For confirmation of this bug, I also ran the same dd command on the only other spare drive I have available, a 128GB SSD, and here it also wipes all but the last 48 sectors: