Seagate disk problems (NCQ bug???)
Wolfgang S. Rupprecht
wolfgang.rupprecht+gnus200905 at gmail.com
Mon May 11 22:17:11 UTC 2009
"D. Hugh Redelmeier" <hugh at mimosa.com> writes:
> Could you dd an equivallent volume from /dev/zero to see if it also
> causes a problem? That eliminates one more variable from the problem.
I'll try. I have only seen it happen with a dvd as source. Disk to
disk copies such as "cp -ax / /mnt/backup" didn't seem to cause
problems.
> Why do you call it "streaming conditions"? Do you think that the fact
> that the DVD delivers bytes more slowly than the hard drive would accept
> them is important?
I had assumed that the bug required a high sustained data rate that
would dump a large number of consecutive blocks into the disk's on-disk
cache, but was still slow enough to allow the on-disk cache to drain.
But then again, it was just a stab at explaining why it only seemed to
happen when reading from dvd.
> Or is there some other aspect that you think is relevant?
Hard to say. I'd only really had 3 or 4 lockups and each time it was
the morning after I'd added a new video.
> NCQ has been around long enough that I'd expect/hope that the nits
> have been picked. Of course I could be wrong.
Well, I've seen the bug described as a on-disk cache flushing bug and/or
NCQ bug.
In any case, Seagate support got me some SD3B firmware to load onto the
disk. Strangely my disk now identifies itself as ST31500341AS instead
of ST31500343AS (with a "1" instead of a "3"). That also explains why I
couldn't find other folks with bug reports -- only a few drives went out
with the trailing "3" model number. Lets see how the drive holds up.
I'm beating on it right now.
-wolfgang
--
Wolfgang S. Rupprecht http://www.full-steam.org/ (ipv6-only)
You may need to config 6to4 to see the above pages.
More information about the fedora-list
mailing list