Subject: Re: How to fix a hard drive error
To: None <netbsd-users@netbsd.org>
From: Wolfgang S. Rupprecht <wolfgang+gnus20030130T220253@wsrcc.com>
List: netbsd-users
Date: 01/30/2003 22:18:37
phoffman@proper.com (Paul Hoffman) writes:
> I have a single IDE drive in a NetBSD 1.6 system. Starting recently,
> the console tells me:
> 
> wd0e: error reading fsbn 16596096 of 16596096-16596111 (wd0 bn
> 19775328; cn 19618 tn 6 sn 6), retrying
> wd0: (uncorrectable data error)
> 
> This repeats many times, always the exact same numbers.
> 
> How can I deal with this?

At one time errors like this were usually the sign of a failing disk.
This could also be cased by someone bumping the computer while it was
writing to disk.  You can cause a mistrack by taking a (fully backed
up!) running system and gently turning the tower case on its side
while doing a build (or anything else that writes to the disk.)  There
is no need to bump it, just changing the direction of gravity while
the disk is running is enough to cause a small number of bad writes *
After doing this a few times fsck will come up with quite a few weird
errors along with some semi-permanently screwed sectors that don't
read at all.  These aren't media errors, the head simply went
off-track during the write, perhaps scribbling over adjacent tracks
too.

To fix it I did a 'tar|tar' copy of the data and then cleared the old
disk with a "dd if=/dev/zero of=/dev/rwdXd bs=64k".  The original disk
was fine after that.

-wolfgang

* This was inadvertently discovered while toying with low-noise fans
  and turning the operating computer on its side to more easily
  measure various temperatures.  I started to get suspicious when some
  brand new replacement disks started showing the same "media" errors
  after only running for a day or two.
-- 
Wolfgang S. Rupprecht 		     http://www.wsrcc.com/wolfgang/
      Decoding genes for the sake of cloning is against the DMCA

(NOTE: The email address above is valid.  Edit it at your own peril.)