Memory acquisition problem with yavta and media control.
A message that is not in the archive (sent off-list or not yet received): 05019EE783D8408881382E5648ED48BE@store
Hi Nilesh,
Quoted 56 lines · On Wednesday 04 September 2013 11:30:09 [email protected] wrote:
Quoted 48 lines · Nilesh Mane
Quoted 44 lines · On Sunday 01 September 2013 12:35:35 [email protected] wrote:
On Sunday, September 01, 2013 1:59 AM Laurent Pinchart wrote:
Quoted 12 lines · On Saturday 31 August 2013 10:07:14 [email protected] wrote:
Quoted 11 lines
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.
Quoted 4 lines
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 releasedExample 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 releasedQuoted 2 lines
As per your guidance can I conclude that memory acquired during triggering and image grab process is an bug in kernel or driver ?
Quoted 6 lines
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 ?
Quoted 6 lines
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.
Quoted 14 lines
Quoted 9 lines
Quoted 8 lines
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 : MT9V032Quoted 4 lines
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.
Quoted 4 lines
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.
Signature · 3 lines
Regards, Laurent Pinchart