Subject: Re: tape driver behavior
To: None <>
From: Charles Shannon Hendrix <>
List: current-users
Date: 09/09/2001 18:23:22
On Thu, Sep 06, 2001 at 03:33:01PM -0700, Matthew Jacob wrote:

> See patches below for a first pass at addressing this.


> The architecture for NetBSD's tape driver is that unless you enable early
> warning, a hard I/O error is your indication that you've hit EOT. Without
> early warning set, you'll get literally hundreds of EOM check conditions,
> you'll then get a complete flood of 'errors' as you approach physical EOT.

Yes, as much as several thousand of them.

> Actual early EOM  handling *should* work, but I wouldn't be surprised if it
> didn't.

Yeah, I'm surprised that, even with the problems, dump doesn't pick up
the error and ask for the next volume. Unless something is broken in the
api somewhere, dump should be getting the error value, seeing that it is
related to EOM, and do the right thing.

> The tape tester is part of my tool kit package which is clonable via BitKeeper
> at bk://

BitKeeper?  Have to look that one up.

I'll try to test this in the coming week and get back to you.

"Star Wars Moral Number 17: Teddy bears are dangerous in herds."