[szary_a AT mail.mrrau.hyperreal.info: Re: failure notice]

[David Hinds]

On Sun, Jan 30, 2005 at 10:36:55PM +0100, Dominik Brodowski wrote:
> ----- Forwarded message from "Kot. Czarny." <szary_a at mail.mrrau.hyperreal.info> -----
> 
> the problem is.. i have sandisk ata-flash ide card, bios detects it and
> connects aa /dev/hdc, in old (standalone) i82365 driver it just says:
>     host opts [1]: already alive!
> and everything works fine. with 2.4 in-kernel modules, or with 2.6 
> yenta_socket it reinitializes it and /dev/hdc is lost, and it couldn't mount
> rootfs anymore. when booted from /dev/hda, card is redetected as /dev/hde, 
> but booted alone it doesn't even get detected as /dev/hde by the time vfs
> rootfs mount..
> so.. can old functionality be restored? (that it just leaves it as it is)
> i would gladly apply and test patches :)
> 
> ----- End forwarded message -----
> 
> Any ideas?

Well... what the i82365 driver did (and continues to do, as far as I
know), was to try to determine if a socket was already "running" at
initialization, and if so, to leave it alone.  It did this by seeing
if the socket was powered up, in "memory and IO mode", with IO windows
configured, and with those IO resources already reserved by the
kernel, meaning some Linux driver was already talking to the card.

The "is_alive()" function implements this.

The status of an "alive" socket is always reported as empty, to
discourage cardmgr etc from interfering.

This was particularly useful as a simple alternative to initrd etc,
when the BIOS directly supported PCMCIA boot devices.

-- Dave