gsmd and modem power on

Hi all,

I just though about something regarding the init sequence of
gsmd. 

Currently, on cold start, my openmoko gsmd is not up and
running fine
without killing xsession, restarting gsmd, and starting
again xsession.
Or by adding a gsmd restart at the beginning of matchbox
session file. 

As I really suspected that the power-on of the modem takes a
bit of time
till it's ready to issue at commands, I moved the actual
power on (echo
1 > ..../power_on) into a separate init file that I call
really early in
the init stages. 

This trick allows the gsmd to be up and running each time I
rebooted my
moko without any gsmd restart trick required.

what do you guy think about this approach ? 

Thanks,

Alex.

----- Original Message ----- 
From: "Alexandre d'Alton" <alexalexdalton.org>
To: <gsmd-devellists.openmoko.org>
Sent: Wednesday, October 03, 2007 2:23 PM
Subject: gsmd and modem power on


> Hi all,
>
> I just though about something regarding the init
sequence of gsmd.
>
> Currently, on cold start, my openmoko gsmd is not up
and running fine
> without killing xsession, restarting gsmd, and starting
again xsession.
> Or by adding a gsmd restart at the beginning of
matchbox session file.

How recent is the image that you are running?  I believe
that the issue
which you describe was fixed earlier this week; recent
images should not
exibit this behavior.  You can tell by seeing if the line
"stty -F
"$GSM_DEV" -echo" is present in your
/etc/init.d/gsmd file.  If it is not,
try upgrading to a more recent image (or just upgrade the
gsmd ipkg - that
should get you the updated startup script as well).

> As I really suspected that the power-on of the modem
takes a bit of time
> till it's ready to issue at commands, I moved the
actual power on (echo
> 1 > ..../power_on) into a separate init file that I
call really early in
> the init stages.

This would be contrary to what I've found, in fact.  I have
complete success
(with the change mentioned above) -- in fact, it turns out
that the 1 second
delay after the power_on is completely unnecessary -- the
only thing it does
for me is waste one second of time at boot.  Just to
clarify, from your
/tmp/gsm.log file, is the problem that you are solving the
inability of gsmd
to syncronize with the GSM modem and establish
communications, or is that
the GSM modem is unable to register?

> This trick allows the gsmd to be up and running each
time I rebooted my
> moko without any gsmd restart trick required.
>
> what do you guy think about this approach ?

I agree completely with starting gsmd earlier in the boot,
and in general
with the concept of re-ordering and optimizing the boot
sequence to reduce
the (lengthy) startup time.  However, in my opinion, we
should be able to
start gsmd at any time, and it should be able to establish
communications
with the GSM modem.  If we rely on ordering, and bootup
delays to get the
communications established, then we're just deferring the
problem -- our end
goal should be a system that boots, and is able to place a
call, in a matter
of seconds and when we get there, I'd hate to find that
we've been depending
on long startup delays for other tasks in order to get gsmd
to work.  JMO.


> Thanks,
>
> Alex.


Mike (mwester)

> How recent is the image that you are running?  
Actually, just after I sent this mail, I realized that it's
been at
least one week since a ipkg update & upgrade were not
installing new
packets. By updating my build environment, I also realized
that gsmd in
my phone were not up-to-date with the built one.
By forcing installation of this newly built gsmd, the
problem
effectively disapeared without my hack.

> I agree completely with starting gsmd earlier in the
boot, and in general
> with the concept of re-ordering and optimizing the boot
sequence to reduce
> the (lengthy) startup time.  However, in my opinion, we
should be able to
> start gsmd at any time, and it should be able to
establish communications
> with the GSM modem.  
I agree.

> Mike (mwester)
> 
Thanks for you comments and suggestion,

Alex.