Operating system name and version: Debian squeeze/sid
Version of smsd: 3.1.6
Smsd installed from: package repository
Name and model of a modem / phone: SE K750i
Interface: USB

Sometimes (not often) my interface device eg. /dev/ttyACM0 decides to suddenly change and become /dev/ttyACM1. I only have one mobil connected so smsd wont send any more sms's if this happens. I'm getting "Cannot open serial port /dev/ttyACM0, error: No such file or directory" in the logs.
I'm not really asking for help on solving my unstable telephone, but will a workaround be to add several "modems" to my smsd.conf, one for ACM0, ACM1 and ACM2..? Two of them will always be unavailable, but one of them will most likely always work.. And how can I configure smsd.conf to minimize delays if one ACM dies?

Is there anything bad in this workaround?

I'm only using this for private usage, so a rock stable solution isnt really necessary..


Thanks!

Regards
Lars

When smsd is starting and a modem port cannot be opened, that process will stop permanently. Therefore you cannot have ACM1 and ACM2 as "ready for use" when your modem is ACM0. Also, even if they could be as a spare, there will be very much error messages in the log.

You could monitor alarm events, and change the port setting and restart smsd by the external process. It could work in this way:

First, use keep_open = no for a modem. This will ensure that in case of an error you will get a message which is easy to detect.

Define an alarmhandler. When opening fails, smsd will call that script and $6 is the alarm text. In your case it contains "Cannot open serial port". When your script sees this, it can create a flag-file to somewhere. You could also log all alarm events to your own log file, it may be usefull if you later get some other messages.

Next, create a watchdog script and run it by the cron, for example once per minute. It should monitor if a flag-file exists. When the file exists, it can call init.d/smstools stop.

Then you need to check what is the port used currently. It can be seen using dmesg. Take a look at this post. It contains a sample code to detect the port and change the smsd.conf setting.

After the port setting is changed, your watchdog can start the smsd again, and it will run using the new port. A flag-file should be removed and case should be taken that the watchdog script will not run if it's already running.

How this kind of a functionality sounds for you? And are you good in scripting?

Thank you very much for your answer, it made sense!
I have experience in scripting, and this doesn't sound to hard

Regards

Here is an another way to act:

Modify the init.d/smstools (or sms3) script. In the case of start, check what is the current port using dmesg. Then create (or update) a symbolic link to that port, for example (if running as a root):
ln -f -s /dev/ttyACM2 /dev/smsd_port

or use some other directory if running as an unpriviledged user.

In the smsd.conf, just define device = /dev/smsd_port (or link in some other directory). There is no need to change it every time.

The alarmhandler can call /etc/init.d/smstools restart and then just exit, when it has seen that the serial port cannot be opened. No other scripts are needed (or at least I think so )

That looks like a much more elegant solution! Will use that one.. Thanks a lot!