Subject: Re: umass woe: sd0(umass0:0:0): readonly device & drive offline
To: Martin Husemann <martin@duskware.de>
From: Michael van Elst <mlelstv@serpens.de>
List: tech-kern
Date: 11/17/2005 23:17:41
On Thu, Nov 17, 2005 at 10:41:46PM +0100, Martin Husemann wrote:
> On Thu, Nov 17, 2005 at 09:56:46PM +0100, Hubert Feyrer wrote:
> > Following the discussion here, the fix in that PR seems to be the best
> > bet we have. If noone beats me to it, I'll commit it within the next few
> > days and try to get it into the netbsd-2 and netbsd-3 branches. (It's
> > already tested successfully on -current and netbsd-2).
>
> This does not sound like a good idea to me. This kind of quirks do not
> scale and are gone for a reason. The problem should be properly analyzed
> instead. Too bad noone picked this up so far (in nearly a year), but
> after all this time I see no need to rush in a brute force hack now.
I agree that there is no need to rush, but I don't consider this
a "hack".
We have devices that react erroneously when given a specific command.
The correct solution is to not send the command, in particular because
the command does not make any sense (there is no spoon^Wdoor).
The method of choice to handle such device anomalies is a quirk flag.
--
Michael van Elst
Internet: mlelstv@serpens.de
"A potential Snark may lurk in every tree."