Re: [PATCH] [RFC] [Modem Manager] Fix for CDMA modems not able to connect if SID=0

[Dan Williams <dcbw redhat com>]

On Wed, 2010-03-24 at 20:54 -0500, Jerone Young wrote:
> Apparently in some countries, such as India, some providers have an SID
> ID of 0. 

This isn't actually the right way to solve the problem.  The problem is
that we're using SID as a proxy of registration state on devices which
don't have extended AT commands for registration state.  Unfortunately,
SID is not a perfect proxy.  The SID should not be 0, and in fact that's
not really legal.  Every network must have a SID, otherwise roaming and
other stuff simply doesn't work.  The SID is like the MNC of a GSM
device.

SID is also only valid for 1x registration state, since EVDO does not
use the SID (instead it uses "subnet codes").

So the problems we've observed are:

1) if the device can't find a 1x signal, but has an EVDO signal, then
the SID isn't necessarily known.  This happens for some providers in
Russia and India that are EVDO-only, and the modems for those providers
(AnyDATA and ZTE specifically) often don't have extended AT commands for
determining registration state.  In the US there really aren't any
EVDO-only providers, and so the modems always have a 1X registration and
thus the SID is always known.

2) On some devices where there *are* extended AT commands, the SID isn't
necessarily reported; this was seen on Sierra CDMA modems in some cases.

The real solution is to use qcdm to determine the actual registration
state of the device, instead of depending on AT commands that may or may
not represent registration state.  I've implemented that recently in
ModemManager git (e7d1e4adb9bf984736ae2bfadbdd616ebc6ade80 and
39326f249105b7d71c63125f29e3bee2143a82d2) and am waiting for some
feedback on that from testers.

The problem that occurs if you only look for SID 99999 is that if the
modem *isn't* registered, the connection will immediately fail instead
of waiting for the device to register.  This was a problem last summer,
and the +CSS check was introduced during registration as a fix for that
problem.  It was most visible with connections that were supposed to
autoconnect the 3G.

That patch is essentially equivalent to ignoring the GSM +CREG state and
blindly trying to connect the modem not matter whether it's registered
or not.

Dan

> According to this spec this is the case (page 11):
> http://www.tiaonline.org/standards/technology/cdma2000/documents/TIA-EIA-IS-707-A-2.pdf
> 
> The simple patch to fix this can be found here:
> https://bugs.launchpad.net/ubuntu/+source/modemmanager/+bug/461096/comments/14
> 
> You can see from the bug this patch fixes issue for CDMA users who's
> providers are following this:
> https://bugs.launchpad.net/ubuntu/+source/modemmanager/+bug/461096
> 
> 
> 			Thanks,
> 				Jerone
> 
> 
> 
> 
> _______________________________________________
> NetworkManager-list mailing list
> NetworkManager-list gnome org
> http://mail.gnome.org/mailman/listinfo/networkmanager-list