SMS Server Tools 3
 Menu
 Home 
Basic information:
 Slideshow, watch this first 
 License 
 How to compile/install 
 - Instructions for Windows 
 How to configure 
 - Configuring providers 
 Localizing 
 How to run/use 
 SMS file format 
 - User Data Header 
Additional information:
 Version history (3.x) 
 National language shift tables 
 Hardware compatibility 
 Status monitor, Statistics file 
 Event handler, Alarm handler 
 Blacklist and Whitelist 
Support:
 Frequently asked questions 
 SMSTools3 Community 
Get SMS Server Tools 3:
 Download 
Additional Options

 Sponsored links

 Search
Custom Search

 Visitor locations
 
 SMS Server Tools 3 Community
Welcome, Guest. The forum is currently read-only, but will open soon. Thu Nov 21, 2024 13:09
• SMSTools3 Community » Feature requests Bottom

better handling of disappering devices

Login and Post Reply

Page:  1

Author Post
lambda
Bottom
#1 Mon May 22, 2017 19:59, 91 months ago.
Member
Registered:
May 2017
Posts: 8
Location: Vienna, Austria
Hi,

this is somewhat related to this post:
http://smstools3.kekekasvi.com/topic.php?id=825

I'm on OpenWRT/LEDE (using custom built 3.1.21 package) and my modem has 3 devices:
/dev/ttyUSB[0-2] by default. I'm using ttyUSB2.

On the smsd side everything works fine, but the HW is used in an industrial environment with lots of EMI. This causes the usb bus to disconnect/reconnect the modem every now and then. (It can be simulated by unplugging/repluging the modem, I think.)

When this happens smsd recieves EIO, but insists on using the non-existent device. Ie I get a line of


every second without smsd ever recovering.

And what's even worse: It seems since ttyUSB2 is still "in use" after reconnecting the modem, the linux kernel assigns the modem to the devices ttyUSB0, ttyUSB1 and ttyUSB3.

I managed to work around this with an alarmhandler, that detects the EIO error and a startup script that links /dev/modem to the right device node, but it is clumsy and has some annoying side effects. I think smsd could handle this much better:

* do something sane on EIO - probably closing the device and following the usual startup procedure
* maybe also provide an option to use the highest number ttyUSB*, whatever it might be.
keke
Top
#2 Mon May 29, 2017 21:59, 91 months ago.
Administrator
Registered:
May 2009
Posts: 2070
Location: Jyväskylä, Finland
In the case of disappeared modem smsd should give up faster. I may change this in the next version.

Currently there are several settings which will help in your kind of environment:

devices = GSM

# Use small errorsleeptime to get rid of EIO faster:
errorsleeptime = 1

# Use an alarmhandler which will adjust the symlink to the device
alarmhandler = /usr/local/bin/smsd_alarmhandler.sh

[GSM]
# Use symlink:
device = /dev/modem

# Close the modem when it is not used:
keep_open = no

# When opening fails and alarmhandler was called, sleep less than 30 seconds:
device_open_errorsleeptime = 3

# Other relevant settings where the default value is ok:
# device_open_alarm_after = 0
# device_open_retries = 1

 
'smsdconf' Syntax Highlight powered by GeSHi


Here is the alarmhandler which I think is self explaining:

#!/bin/bash

echo "$2 $3,$4, $5: $6" >> /var/log/smstools/alarm.log

if [ "$5" == "GSM" ]; then
  if echo "$6" | grep "Couldn't open" >/dev/null; then

    DEVICE="/dev/modem"

    if [ ! -e "$DEVICE" ]; then
      echo "Device does not exist: $(readlink "$DEVICE")"

      while [ ! -e "$DEVICE" ]; do

        # Customize these as necessary:
        if [ -e /dev/ttyUSB3 ]; then
          ln -sf /dev/ttyUSB3 "$DEVICE"

        elif [ -e /dev/ttyUSB2 ]; then
          ln -sf /dev/ttyUSB2 "$DEVICE"

        else
          sleep 3
        fi

      done

      echo "Now using: $(readlink "$DEVICE")"
    fi
  fi
fi
 
 
'bash' Syntax Highlight powered by GeSHi


I just noticed that device_open_errorsleeptime had been forgotten in the documentation. It will be updated.

The solution is external because there are different systems and not all possibilities are reasonable or even possible inside the smsd.
keke
Top
#3 Mon May 29, 2017 22:11, 91 months ago.
Administrator
Registered:
May 2009
Posts: 2070
Location: Jyväskylä, Finland
As an addition, if you want that the device is always checked and set before opening, make the following change to the code: Edit src/modeminit.c, search for int openmodem and you are on the line 2347:

int openmodem()
{
  int retries = 0;

  // 3.1.7:
  //modem_handle = open(DEVICE.device, O_RDWR | O_NOCTTY | O_NONBLOCK);
 
'c' Syntax Highlight powered by GeSHi


Add highlighted lines as shown:

int openmodem()
{
  int retries = 0;

  char *DEVICE_device_open_preparator = "/usr/local/bin/smsd_preparator.sh"; // Change this as necessary.

  // 3.1.22: Run device_open_preparator if defined:
  if (DEVICE_device_open_preparator[0])
  {
    char cmdline[PATH_MAX + PATH_MAX + 32];
    int i;
    char *EXEC_DEVICE_OPEN_PREPARATOR = "device_open_preparator";

    snprintf(cmdline, sizeof(cmdline), "%s %s \"%s\" \"%s\"",
             DEVICE_device_open_preparator, DEVICE.name, DEVICE.device, DEVICE.description);

    if ((i = my_system(cmdline, EXEC_DEVICE_OPEN_PREPARATOR)) != 0)
      writelogfile(LOG_CRIT, 0, "Device_open_preparator script %s returned %i.", DEVICE_device_open_preparator, i);
  }

  // 3.1.7:
  //modem_handle = open(DEVICE.device, O_RDWR | O_NOCTTY | O_NONBLOCK);
 
'c' Syntax Highlight powered by GeSHi


With this change the preparator script could be like this:

#!/bin/bash

DEVICE="$2"

if [ ! -e "$DEVICE" ]; then
  echo "Device does not exist: $(readlink "$DEVICE")"

  while [ ! -e "$DEVICE" ]; do
    if [ -e /dev/ttyUSB3 ]; then
      ln -sf /dev/ttyUSB3 "$DEVICE"
    elif [ -e /dev/ttyUSB2 ]; then
      ln -sf /dev/ttyUSB2 "$DEVICE"
    else
      sleep 3
    fi
  done

  echo "Now using: $(readlink "$DEVICE")"
fi
 
'bash' Syntax Highlight powered by GeSHi


With this additional feature no alarmhandler is needed, but the setting keep_open = no is mandatory.
lambda
Top
#4 Wed Jun 14, 2017 15:24, 90 months ago.
Member
Registered:
May 2017
Posts: 8
Location: Vienna, Austria
Topic owner
Thanks for your input. The setup was already in production, when I read this, so I couldn't test this yet. I will come back to this when there is an opporunity (the setup needs maintainance or I get a new modem).

One unrelated request though: There is a patch in OpenWRT to compile smstools3:
https://github.com/haraldg/packages/blob/smstools3/utils/smstools3/patches/002-Makefile.patch

I don't think the current patch is the best way to handle this, but maybe you can improve the Makefile of smstools instead of me improving the patch in OpenWRT.

Thanks for your help,
Harald

Login and Post Reply

Page:  1

• SMSTools3 Community » Feature requests Top

 
Time in this board is UTC.  
Privacy Policy   SMS Server Tools 3 Copyright © Keijo Kasvi.