gstreamer application crashing after running for a few times on dm368

 A gstreamer C application is written to be run on Dm368 for  DVI+rtp streaming.
Also recording activity takes place on board based on  start and stop recording commands as well as image capture commands.Currently being received over the socket.
There are 2 threads created in main 
supervisor() handles socket programming for recording/capture commands:
doSomeThing(void *arg) is the therad handling gstreamer pipeline:
In this function:
The DVI +stream pipeline is created using  
	create_pipeline();
all the elements are linked using
	link_elements();

Design:
v4l2src-->queue-->filtercaps--
                              \
                               -->tee-->queue-->TIDmaivideosink(DVI output)
                                      |
                                      |-->queue-->dmaienc_h264-->rtph264pay-->udpsink(streaming)



Approach for Recording(Start/Stop) and Image Snaphot


1)First Approach 
On start recording command(Send 1 over socket)
 A recording bin is created
using create_recBin()
|----------------------------------------------------------|
|recording bin                                             |
|                                                          |
|   queue-->dmaienc_h264-->qtmux-->filesink                |
|                                                          |
|                                                          |
|----------------------------------------------------------|
 
Then start_recording() is called which performs:
1)adding recording bin to the pipeline 
2)set to playing state 
3)and it is attached to the tee by requesting a pad from it.
On the stop command(send 5 over socket)
stop_recording(&pipedata) function is called which in turn calls handle_pad_block() 
The following are the steps in handle_pad_block():
1)the recording bin ghost pad is unlinked from the tee.
2)send EOS to the recording bin  .
4)the recording bin is removed from the pipeline.
6)the recording bin is set to null 
7)the recording bin is unref.
8)the  requested tee pad is released using gst_element_release_request_pad ()

On similar lines on start capture (send 2 over socket) a image a capture bin is created using 
create_imgBin()

|----------------------------------------------------------|
|capture bin                                               |
|                                                          |
|   queue-->jpegenc-->filesink                             |
|                                                          |
|                                                          |
|----------------------------------------------------------|

and attached to the tee by requesting a pad from it when the start capture command is received.
Here  a sleep function is set then image_done() function is called
1)the requested tee pad and the ghost pad of the capture bin is then unlinked.
2)a EOS is sent to the capture bin.
A event probe is used to detect whether the eos is received at the filesink
3)the capture bin is removed  from the pipeline.
4)the capture bin is set to null 
5)the capture bin is and unref.  
6)the  requested tee pad is released using gst_element_release_request_pad ()
                                                                  

The application runs for  quite a few initial no. of times of recording and capture before getting crashed (crash dump attached).Can any one sugest a suitable technique to solve the problem.
8875.kernelPagingREquestErrorOnlyRec+capture.txt0410.video_capture_application.c