Subject: Re: Diaspora, politics, and MIf
To: Jukka Marin <>
From: Charles M. Hannum <>
List: current-users
Date: 09/19/1996 17:21:20
Jukka Marin <> writes:

> Instead of telling me WHAT is wrong
> and should be changed, they're just flaming me because the code is "bad".
> Well, that "bad" code happens to WORK - something that the cy.c in the
> source tree does NOT.

You're probably referring here to something I wrote in private mail.
It's true that I haven't had time to give you a full analysis of the
problems with the driver.  That's a shame, but other things do have

Anyway, there are some architectural problems in this code:

1) The separation of the ISA and PCI code is rather poor.
2) It should be split into separate `board' and `chip' layers.
3) It uses fixed-size global arrays for some information.  (Worse, it
keeps state in them that there's really no reason to keep.)
4) If CY_HW_RTS is not enabled, there appears to be several
handshaking bugs.

Also, there are some other minor niglets:

A) It doesn't use tty_attach().
B) It uses ttioctl() where it should use l_ioctl.
C) It doesn't seem to handle MDMBUF.
D) It needs to use spltty() rather than disable_intr().
E) The overrun messages should be output asynchronously, like in the
com driver, to avoid making overload conditions worse.

In addition, it would be nice if some of the performance-related
changes that Onno is working on were incorporated into this driver.

That said, I have offered to work on the code, but I can't reasonably
do this without some way to test it.  Access to the console of a
machine with such a card would probably work, even remotely.