[RFC] TI Calypso / HTC firmware support

[pHilipp Zabel]

On 3/28/07, Harald Welte <laforge at openmoko.org> wrote:
> On Tue, Mar 27, 2007 at 06:16:43PM +0200, Philipp Zabel wrote:
>
> > I wrote a line discipline for the simple multiplexing protocol that is
> > used by TI Calypso + HTC firmware phones like the HTC Blueangel,
> > Magician and Wallaby.  Thus it is possible to write AT commands
> > directly to the UART if the N_TIHTC line discipline is set, and it is
> > possible to run an unmodified gsmd on those devices.
>
> great!
>
> > Question: Is it in the scope of gsmd to set the line discipline in the
> > vendor plugin if needed, or should this be put in a separate machine
> > specific daemon, outside of gsmd? As I understand the line discipline on
> > a tty is reset to N_TTY as soon as the device file is closed.
>
> It is well within the scope of gsmd.
>
> I am still working on the multiplexing line discipline code for TS07.10.
> ACtually, what I have in mind is even more general.  the HTC and
> motorola  multiplex would just be a plugin to this generic GSM multiplex
> architecture.
>
> At that point, the interface for gsmd becomes completely transparent,
> i.e. it just uses the N_GSM line discipline and the rest is handled by
> the kernel, depending on which particular hardware is in use.

Good. My current implementation is at
http://www.handhelds.org/hypermail/kernel-discuss/current/3125.html.
As I have next to no experience in that area, I'd be very happy
to reimplement that inside some sane framework.

> > Further, the HTC firmware has a slightly different (smaller) featureset
> > than what is found on the Neo1973. Does that warrant another vendor
> > (plugin, as soon as that is possible), or should I look into extending
> > vendor_ti? The attached patch is what I use on the magician currently. I
> > just removed the unsupported commands and changed %CSQ to %HTCCSQ
> > parsing so far, but I think there is some more initial configuration to
> > do.
>
> I guess it boils down on how big those differences are.  As indicated
> before, I want to have chipset and machine plugins, plus the ability to
> specify init strings from the config files.

Right, I forgot. Except for the missing features (timezone related,
AT%CPRI not allowed) and the %CSQ/%HTCCSQ difference, it will probably
boil down to different initialization strings.

> One interesting question is auto-detection.  Is there any sensible
> response to manufacturer/model/serial number/... inquiry comands on your
> HTC machines?

When the device is powered on, I get the following string on the
debug/trace channel (0x12):
"GSM Version : 1.12.00 G23 Version : 1337.42"

On the AT command channel (0x16) I can issue the usual
vendor/model/revision queries:

> AT+CGMI
HTC

> AT+CGMM
MAGICIAN

> AT+CGMR
1.3.3.7

cheers
Philipp