On Mon, May 24, 1999 at 09:38:38AM +0100, Gordon Russell wrote:
>
> I am getting some reports that the new setserial 2.15-3 interferes
> with pcmcia cardservices. It would appear that setserial runs first,
> and initialises the pcmcia card for serial use. This is fine unless the
> card is a combo modem/ethernet card, in which case the ethernet part
> of the card is not initialised by cardservices.
>
> Is this a problem with setserial or with pcmcia??
I am the original submitter of this bug, and I've done some digging
since then.
As far as I can see, this is caused because setserial generates is
new-format config file _after_ pcmcia has been started (when you
install the new setserial package).
It therefore includes the ports on pc-cards that has already been
set up by pcmcia-cs, and somehow setserial is able to initialize
these ports even before pcmcia-cs has been started (perhaps the
pcmcia controller 'remembers' the state before boot).
A fix could be to do a "cardctl config" before generation of the
new-style config file, and avoid the ports, that pcmcia-cs is
responsible for, e.g.:
ford:~$ cardctl config
Socket 0:
not configured
Socket 1:
Vcc = 5.0, Vpp1 = 0.0, Vpp2 = 0.0
Interface type is memory and I/O
IRQ 3 is dynamic shared, level mode, enabled
Speaker output is enabled
Function 0:
Config register base = 0x03f8
Option = 0x68, status = 0x08
I/O window 1: 0x0300 to 0x031f, auto sized
I/O window 2: 0x03e8 to 0x03ef, 8 bit
In this example, 0x0300 (my netcard, not important) and 0x03e8 (my
modem) should be ignored when probing/saving the config.
Hope this solves the problem - I will work on a solution and send
the script to you when I have a moment to do this, but perhaps
you should do this yourself (since my time is sparse ATM), and then
send the package, and I will test it for you.
--
Andreas