[Paranoia] No such device or address errors

[Jonathan Morace]

> On Sun, 23 Jan 2005 00:54:15 -0500, Monty <xiphmont at xiph.org> wrote:
>
> Starting in late 1998-ish, Plextor drives went over the top trying to
> put paranoia-like capability into the drive firmware such that if they
> hit a media error, they would retry endlessly and as far as the
> computer was concerned, the drive had disappeared.  There was no way
> to interrupt the drive or abort it; it would not reappear after a bus
> reset because it was still occupied trying to get that damned sector.
> They would sometimes do this even on brand new commercial disks that
> had manufacturing defects that wouldn't make other drives even hiccup.
>
> The behavior appeared in a specific firmware update (1.03 maybe?); in
> addition, it was a firmware update that couldn't be backed out by
> reverting to an earlier revision, and at that point I abandoned
> buying/recommending Plextor.  I don't know if they ever fixed the
> problem.

Well, I've been tinkering a bit more with these drives, trying to
figure out which ones displayed this problem.  I tested each drive
with two heavily damaged cd's and every single one failed on at least
one of them.  Does anyone else see this problem?  I'm surprised I'm
able to cause it so easily but not find any references to it on the
web.

I also tested the drives with cdda2wav and of course saw the same
behavior.  Interestingly, I can't reproduce it with windows
applications like EAC.  Just based on watching how EAC stalls, I'm
betting it patiently waits for the drive to respond instead of trying
to abort the command and causing the drive to go missing.  My guess is
that the drive would eventually respond if it were given the time, but
it's aborted instead.  If I wanted to test that theory out, could
someone suggest where I might find the code performing the abort so I
could just test what happens when I remove it?

Here are the TLA numbers (the first two digits are hardware version
and the last two are firmware version) and dates of the drives I
tested:

PX-12CSI
0000: October 1996
0000: November 1996
0301: April 1997
0301: May 1997
0304: September 1997
0404: April 1998

PX-32CSI
0000: December 1997
0101: April 1998
0101: May 1998
0202: April 1999

Thanks again,

Jonathan