Re: Memory acquisition problem with yavta and media control.

Re: Memory acquisition problem with yavta and media control.

[Laurent Pinchart]

Hi Nilesh,

On Wednesday 04 September 2013 11:30:09 purchase@utopiacontrol.com wrote:
> On Wednesday, September 04, 2013 11:19 AM Nilesh Mane wrote:
> > On Sunday 01 September 2013 12:35:35 purchase@utopiacontrol.com wrote:
> >> On Sunday, September 01, 2013 1:59 AM Laurent Pinchart wrote:
> > > On Saturday 31 August 2013 10:07:14 purchase@utopiacontrol.com wrote:
> >> >> You have done a great work for snapshot mode image sensor driver for
> >> >> linux. I am using your media control tool with yavta test application
> >> >> for interfacing the mt9v032 image sensor with Gumstix Overo Water Com
> >> >> board. I have successfully tested the snapshot mode with this
> >> >> combination. But the problem is that, when I attempt to grab lots of
> >> >> images (thousands) of images by this test application yavta. I found
> >> >> that the free memory goes increasing by some amount which will not get
> >> >> free. Afterwards I have calculate the amount of ram acquires on every
> >> >> snap is about 0.618 KB (after averaging 100000 frames). Will you
> >> >> please Give me any reason why this is happening with this test
> >> >> application? And how can I get overcome on this problem.
> >> > 
> >> > That's definitely not expected and should be debugged. First of all, is
> >> > the memory released when you stop yavta ? If it isn't then we have a
> >> > kernel bug, if it is the bug could be either on the kernel side or the
> >> > application side.
> >> 
> >> Thank you for quick reply,
> >> 
> >> I have checked the memory status before starting yavta and while running
> >> .
> >> then after killing yavta. the results are as below.
> >> 
> >> 
> >> Example condition 1
> >> 1) Before Starting yavta Free memory is 370528KB
> >> 2) After running yavta and before first trigger free memory is 370108KB
> >> //acquired   420KB for initial buffer allocations
> >> 3) After 1008 triggers free memory is 369488KB
> >> //unexpected 620KB for 1008 images ( I.e. 0.61KB / image ) acquired
> >> 4) After killing yavta free memory is 369908KB
> >> //620KB not released only    420KB released
> >> 
> >> Example condition 2
> >> 1) Before Starting yavta Free memory is 366064KB
> >> 2) After running yavta and before first trigger free memory is 365520KB
> >> //acquired   544KB for initial buffer allocations
> >> 3) After 10000 triggers free memory is 359328KB
> >> //unexpected 6192KB for 10000 images ( I.e. 0.6192KB / image ) acquired
> >> 4) After killing yavta free memory is 359864KB
> >> //6200KB not released only    536KB released
> >> 
> >> As per your guidance can I conclude that memory acquired during
> >> triggering and image grab process is an bug in kernel or driver ?
> >
> > Not yet, you should be careful about how you compute the free memory.
> > Merely running free won't give you an accurate information about potential
> > memory leaks.
> >
> > I also don't know what you mean by trigger above, could you please
> > elaborate ?
>
> First of all, I mean to say triggering is the image acquisition by
> application on external trigger applied to camera which is in snapshot mode.
> ok I can understand the size I have calculate is not exact but I just want
> to tell you that the memory  acquired by the application is partially
> released ,  memory Acquired during the images grabbed on external triggers
> is not released after killing the application.

The mainline mt9v032 driver doesn't support snapshot mode as far as I know. 
This leads me to believe that the kernel you are using contains modified code 
compared to the mainline version. A memory leak could have been introduced 
there. This is one more reason to test a mainline kernel first.

> >> If it is how can I found it to get resolve my error?
> >> 
> >> From where can I get the correct source for MT9V032 driver working in
> >> snapshot mode and  kernel source code(3.2.0+ or any). for my platform?
> >> Platform      :    Gumstix Overo Water Com
> >> OS            :    Angstrom
> >> Kernel        :    3.2.0+
> >> Image sensor  :    MT9V032
> >
> > 3.2.0 is old, you should upgrade to the latest mainline kernel. The
> > mt9v032 driver is present in mainline, and sample board code is available
> > at http://git.linuxtv.org/pinchartl/media.git/shortlog/refs/heads/board/
> > overo/mt9v032.
>
> Can  I get the detailed documentation by referring it I can Download
> mainline kernel, build it with the MT9V032 camera driver in snapshot mode 
> for specific board (Gumstix over water com) and test it with media control
> and yavta test application.

I'm afraid my spare time is limited, and I can't give you a detailed step-by-
step explanation. I'm pretty sure you can find plenty of resources online to 
help you building a custom kernel for the Gumstix Overo.

-- 
Regards,

Laurent Pinchart