Re: ModemManager: issues with Nokia handset in git master

[Dan Williams <dcbw redhat com>]

On Sat, 2011-08-20 at 17:41 +0200, Aleksander Morgado wrote:
> Hey Dan & everyone,
> 
> The recent commit 0f6d1b2 [1] to fix some N900 behaviour in git master
> seems to be breaking my Nokia C7 connection. Without the PORT_SEND_DELAY
> setting, I seem to get the same strange behaviour as described in commit
> 46d757f [2]: we get some but not all echo characters back. I tried to
> re-enable the PORT_SEND_DELAY property value to 5000 again, and when
> that set, the modem now sends all echo-ed characters back (still doesn't
> disable the echo anyway).
> 
> But, while debugging the issue, I also found that I could properly fully
> disable the echo if I first open the port with minicom and run ATE0 in
> there. After some more debugging, it seems that if we use ATE1 (which is
> part of the minicom initialization commands) and then ATE0 just when
> opening the serial connection, the modem ends up properly disabling the
> echo completely.
> 
> Attached a patch that follows this logic both in the modem enabling
> process and in the setup before probing. I tested it with my Nokia C7
> and seems to work quite well; no need of PORT_SEND_DELAY and no need of
> having the INIT command twice... Can someone test it with a N900 or some
> other Nokia handset? If it ends up working properly, we can even skip
> using the custom v1_e1 response parser in the Nokia plugin. Or maybe I'm
> just too optimistic... :-)

Unfortunately the trick doesn't work on the N900, so I've applied
everything but the mm-generic-gsm.c hunk.  Oddly though, now the N900
echos back everything upon disconnect, which is the same problem we
recently had with the Ericsson 5521gw, where "flashing" (setting baud to
0 for a short period) resets the ephemeral port attributes like echo and
such.  Seems to work OK though.

Dan