UM1020 support

[Roman Kagan]

On Sat, Feb 19, 2005 at 08:47:24PM +0100, Marco d'Itri wrote:
> The following patch adds support for the Aethra Starmodem UM1020

Applied, thanks.  Just for bookkeeping: do you have any idea which
Conexant product ID it's equivalent to (likely cb00 or cb06)?  It may be
possible to guess from *.inf files in the Windows package.

> [...] when I try to
> load the new cxacru with the cxfirm4.bin firmware it fails ("wrong data
> length 0x19 in response").

Well that's something I'd like people to comment.  The point is that
older firmware revisions use slightly different format[*] when
communicating with the host computer.  OTOH newer firmware seems to be
available on the internet.  Besides, the only place where it matters is
when the driver polls the modem for its status; the ATM layer currently
ignores it so I was able to use the old firmware to connect to the
internet, except that the driver kept complaining every 5 seconds.

So there are 3 ways to handle this situation:

 1) leave it as it is, and expect people to use recent firmware;

 2) make the driver stop polling the modem after several, say 10,
    consecutive failures;

 3) add a workaround to the driver to make it parse the old firmware
    responses too.

Naturally I'd prefer 1) as I wouldn't have to do anything but I'd like
to know other opinions too.

[*] The modem exchanges command/control messages (URBs) with the
computer on endpoint 0x01 arranged into one or more 64-byte packets,
containing 4-byte header and 60-byte data payload.  When the data
payload is to carry elements of integer arrays (configuration, status,
etc.), it consists of (all le32) the number of elements in the current
packet (up to 7) followed by index-value pairs.  The older firmware
revisions, instead of the number of elements in the current packet, seem
to put the number of elements in the whole array into each packet in the
message.

Cheers,
  Roman.