Motion functions properly for a few minutes and then stops logging motion detection

Question

I have a Logitech 820 Clicksmart webcam, I am using the gvcaps driver.

"motion" logs images and movies to the local directory for a few minutes, and then all logging stops. No more images are logged, the movie size does not increase, when for example I move my hand in front of the camera in attempt to trigger motion detection.

For my configuration file, I made a copy of the config file auto-generated, or provided on install, located in /etc/motion.conf, naming the copy "myconfig".

At the top it had the comment

-------------------------------------------------------------
# /etc/motion.conf
#
# This config file was generated by motion 3.2.8

I made the following changes:

• framerate chagned from 2 to 15
• pre_capture changed from 0 to 5
• post_capture changed from 0 to 5
• gap changed from 60 to 0
• target_dir changed from "/usr/local/apache2/htdocs/cam1" to commented out, so that it defaults to the current directory
• webcam_port changed from 8081 to 0, to disable
• control_port changed from 8080 to 0, to disable
• sql_log_image changed from on to off
• sql_log_snapshot changed from on to off

I ran motion using the command motion -d 9 -c myconfig

The output started as:

-----------------------------
[0] Processing thread 0 - config file motionconf
[0] Thread 1 is from motionconf
[1] Thread started
[1] Not a V4L2 device?
[1] Using VIDEO_PALETTE_YUV420P palette
[1] Using V4L1
[1] File of type 8 saved to: ./01-20071123153123.avi
[1] File of type 1 saved to: ./01-20071123153122-11.jpg
[1] File of type 1 saved to: ./01-20071123153122-12.jpg
[1] File of type 1 saved to: ./01-20071123153122-13.jpg
...

In this case, it stopped logging images and video after about 10 seconds.

Sometimes it seems that lack of motion for a long time causes it to stop logging, and other times it seems that too much causes it, for example, if I pick the camera up and move it.

Environment

Motion version:	3.2.8
ffmpeg version:	0.4.9pre1
Libraries:	unkown
Server OS:	2.6.22.9 Mandriva desktop system

-- MichaelOverlin - 23 Nov 2007

Answer

Your config is a bit odd.

Running with gap 0 is normally not adviced. It means that many features related to events do not work at all.

Read up on the ConfigOptionGap

But besides this - does Motion write anything in the syslog? Normally /var/log/messages.

The command dmesg can give some hints if it is a camera problem.

A 3.2.9 was just released fixing some seg faults. Perhaps you should try that.

-- KennethLavrsen - 24 Nov 2007

Reply (MichaelOverlin)

I changed "gap" to the default value of 60, this made no difference.

I looked in the /var/log/messages, relevant messages seem to be:

ov 23 20:18:13 localhost motion: [0] Processing thread 0 - config file motionconf
Nov 23 20:18:13 localhost motion: [0] Thread 1 is from motionconf
Nov 23 20:18:13 localhost kernel: /var/lib/dkms/gspca/1.00.18-3mdv2008.0/build/Sunplus-jpeg/sp5xxfw2.h: [spca50x_GetFirmware:662] FirmWare : 2 0 0 5 4
Nov 23 20:18:13 localhost kernel: /var/lib/dkms/gspca/1.00.18-3mdv2008.0/build/gspca_core.c: [spca5xx_set_light_freq:1889] Sensor currently not support light frequency banding filters.
Nov 23 20:18:13 localhost kernel: /var/lib/dkms/gspca/1.00.18-3mdv2008.0/build/gspca_core.c: [gspca_set_isoc_ep:903] ISO EndPoint found 0x81 AlternateSet 7
Nov 23 20:18:14 localhost kernel: /var/lib/dkms/gspca/1.00.18-3mdv2008.0/build/Sunplus-jpeg/sp5xxfw2.h: [spca50x_GetFirmware:662] FirmWare : 2 0 0 5 4
Nov 23 20:18:16 localhost motion: [1] Not a V4L2 device?
Nov 23 20:20:02 localhost kernel: /var/lib/dkms/gspca/1.00.18-3mdv2008.0/build/gspca_core.c: [spca50x_isoc_irq:1110] Non-zero status (-84) in isoc completion handler.
Nov 23 20:24:13 localhost kernel: /var/lib/dkms/gspca/1.00.18-3mdv2008.0/build/gspca_core.c: [spca50x_isoc_irq:1110] Non-zero status (-84) in isoc completion handler.
Nov 23 20:25:27 localhost kernel: /var/lib/dkms/gspca/1.00.18-3mdv2008.0/build/gspca_core.c: [spca50x_isoc_irq:1110] Non-zero status (-84) in isoc completion handler.
Nov 23 20:25:28 localhost kernel: /var/lib/dkms/gspca/1.00.18-3mdv2008.0/build/gspca_core.c: [spca50x_isoc_irq:1110] Non-zero status (-18) in isoc completion handler.

The -84 and -18 errors might indicate a problem with the driver. Looking around on the internet a bit, I found that some suggest changing a costant in the gspca driver source file "gspca.h", FRAMES_PER_DESC , from 16 to 128. If I try this I'll let you know if it fixed the problem.

Thanks.

-- MichaelOverlin

Reply (MichaelOverlin)

Quite a few people seem to have this problem with the gspca driver. I found a solution that works for me. I made a few tiny changes to the driver that make it just ignore these -84, -18 etc. errors at these locations in the code, and I also increased the constant FRAMES_PER_DESC from 16 to 64.

With these changes my Logitech Clicksmart 820 works reliably for about two hours, and then it fails with a message (in /var/log/messages)

localhost kernel: usb 2-1: reset full speed USB device
using uhci_hcd and address 2

I work around this by writing a little perl-script that runs "motion" one hour at a time.

Here is my perl script

$motionCommand = "motion -c motionconf";
# RUN ONE HOUR AT A TIME
$motionTime = 3600; 

my $k = 0;
for(;;) {
    $k ++;
    print "DoMotion count $k \n";
    my $pid = DoMotion();
    sleep $motionTime;
    kill 9, $pid;
}

sub DoMotion {
    my $pid;
    if (!defined($pid = fork)) {
        print "fork failed\n";
        exit;
    } elsif ($pid) {
        print  "begat $pid \n";
        return $pid;
    }
    exec $motionCommand;
}

The hack I made to the gspca driver:

VERSION gspcav1-20070508

***** IN FILE "gspca_core.c" *****
** IN FUNCTION  "spca50x_isoc_irq" ABOUT LINE 1110
CHANGE SWITCH BLOCK FROM 
   switch (urb->status) {
   case 0:
      break;
   default:
      PDEBUG(0, "Non-zero status (%d) in isoc "
             "completion handler.", urb->status);
   case -ENOENT:      /* usb_kill_urb() called. */
   case -ECONNRESET:   /* usb_unlink_urb() called. */
   case -ESHUTDOWN:   /* The endpoint is being disabled. */
      return;
   }
TO
   switch (urb->status) {
   case 0:
      break;
   default:
      PDEBUG(0, "Ignoring non-zero status (%d) in isoc "
             "completion handler.", urb->status);
      break;
   case -ENOENT:      /* usb_kill_urb() called. */
   case -ECONNRESET:   /* usb_unlink_urb() called. */
   case -ESHUTDOWN:   /* The endpoint is being disabled. */
      PDEBUG(0, "Non-zero status (%d) in isoc "
             "completion handler.", urb->status);
      return;
   }
** IN FUNCTION "spca50x_move_data" ABOUT LINE 1565
WITHIN THE FIRST "for" LOOP THAT READ  "for (i = 0; i < urb->number_of_packets; i++) {"
CHANGE AN ERROR-CHECKING IF STATEMENT BLOCK FROM
      if (st) {
         PDEBUG(0, "ISOC data error: [%d] len=%d, status=%d \n",
                i, datalength, st);
         continue;
      }
TO
      if (st) {
         st = urb->iso_frame_desc[i].status = 0;
      }


***** IN FILE "gspca.h" *****
** CONSTANT "FRAMES_PER_DESC" ABOUT LINE 28
CHANGE FROM 
#define FRAMES_PER_DESC      16 /* Default value, should be reasonable */
TO 
#define FRAMES_PER_DESC      64 /* Default value, should be reasonable */

-- MichaelOverlin

---

Lets see if driver's authour can help you, that is out of motion scope.

-- AngelCarpintero - 06 Jan 2008