DM365 EVM with Leopard Imaging LI-5M02 Camera -- Help

From the log it looks like the display portion is failing as opposed to the camera capture portion, and in particular it mentions it is being passed in invalid argument. If you are running an example from an older LSP version it is possible that the API has changed a bit and the call being made may need to be adjusted to fit the particular LSP you are using.

Thanks Bernie.

That makes sense and I will check it out.  Once I get it working I hope to post a detailed description on getting this camera to work on the EVM.

Jim

Interesting....

I changed the mem=76M in the bootargs to mem=116M and the program runs, but I still do not see any video out.

Jim

jmalone said:

I changed the mem=76M in the bootargs to mem=116M and the program runs, but I still do not see any video out.

This makes it sound like it may have had some memory allocation failure in there, though I would have hoped to see a more specific error regarding allocation, changing that 76M to 116M just gives more memory to the Linux kernel itself. Keep in mind that changing the amount of memory given to the kernel effects the amount of memory that can be used outside the kernel, namely by CMEM for the video accelerators, so you may run into issues trying to run standard DVSDK demos having done this.

As to the video not coming out you may want to check the LI application source to see what interface it is expecting to output to, as the DM365 has composite, s-video, and component. If the example runs like the DVSDK demo it will be using the component output. You may also want to check with LI to see what bootargs they suggest using, as the bootargs can configure the capture and display interfaces in addition to the memory change you have already tried.

Hi Bernie,

After contacting support, it turns out that the kernel patch and prebuilt uImage binary was corrupt from the CD that Leopard sends with the hardware.  They sent me a new RAR of the kernel patches which patches  support for the MT9P031 and adds support for MT9V034 sensors.  I am running it with success, with a cavet.  If anyone wants the details of this and/or the RAR, please let me know.

The cavet is that when trying to run the loadmodules.sh, the CMEM portion displays an error.

root@192.168.1.50:/opt/dvsdk/dm365# ./loadmodules.sh
CMEMK module: built on Sep  9 2009 at 11:27:08
  Reference Linux version 2.6.18
  File /home/jkmalone/dvsdk_2_10_01_18/linuxutils_2_24_02/packages/ti/sdo/linuxutils/cmem/src/module/cmemk.c
CMEM Range Overlaps Kernel Physical - allowing overlap
CMEM phys_start (0x85000000) overlaps kernel (0x80000000 -> 0x87400000)
CMEMK Error: Failed to request_mem_region(0x85000000, 50331648)
insmod: error inserting 'cmemk.ko': -1 Bad address
IRQK module: built on Sep  9 2009 at 11:27:14
  Reference Linux version 2.6.18
  File /home/jkmalone/dvsdk_2_10_01_18/linuxutils_2_24_02/packages/ti/sdo/linuxutils/irq/src/module/irqk.c
irqk initialized
EDMAK module: built on Sep  9 2009 at 11:27:12
  Reference Linux version 2.6.18
  File /home/jkmalone/dvsdk_2_10_01_18/linuxutils_2_24_02/packages/ti/sdo/linuxutils/edma/src/module/edmak.c
root@192.168.1.50:/opt/dvsdk/dm365#

I am not familiar enough with what these modules are doing yet to understand why CMEM is reporting a bad address, so any help or pointers to other threads would be great.

Thanks!

Jim

jmalone said:

After contacting support, it turns out that the kernel patch and prebuilt uImage binary was corrupt from the CD that Leopard sends with the hardware.  They sent me a new RAR of the kernel patches which patches  support for the MT9P031 and adds support for MT9V034 sensors.  Although the notes say this was built with the DM3655 2.10.0.14 LSP, I am running it with the 2.10.0.18 with success, with a cavet.  If anyone wants the details of this and/or the RAR, please let me know.

Just a quick note, DVSDK 2.10.0.18 includes LSP 2.10.0.14 (see name of PSP directory under DVSDK tree).

I just caught that and corrected my original post.  Thanks Juan.

I have the same outcome "CMEM phys_start (0x85000000) overlaps kernel (0x80000000 -> 0x87400000)", and can only found DVSDK 2.10.0.12 on TI's web site.

please share me the details of this and RAR.

thanks!

song10

Hi Jim,

I am trying to operate the Leopard board “LI-5M02 HD Camera Board” on the DM365 EVM board.

When executing the “capture_prev_rsz_onthe_fly_bayer”  it crash around the set the vpfe ccdc capture stage.

I am working with the original supplied linux by TI with the DM365EVM (linux-2.6.18_pro500)

And I have patched the drivers patch that supplied by Leopard CD.

As I understood, you have a much better patch that may help me..

So, can you send me the rar file with the new patch?

Additionally, I would like to ask: what should be the value of the boot parameter “davinci_capture.device_type” ?

My email is : Ilan.arad@odfopt.com

ilan

Hi, jmalone, would you please share the rar file leopard sent you? my email is liuhao@yahoo.com

thanks. i am having exactly same problem. changed to 76M but still composite does not output.

by the way, i dont think the leopard demo application (capture_prev_rsz_onfly_dm365) uses CMEM buftab, because it does not do any encoding/decoding.

jmalone,

i'm working with the dm365 evm and li-5m02. I tried Leopard Imaging support but have not gotten a response. will you please provide the .rar and any information you can offer? I appreciate it.

rreper@yahoo.com

thanks.

All,

I have not worked on this for a while and it is on another machine at work.  I am in the middle of an integration crunch for at least another week.  Once that let's up, I'll get something posted to help you out.

Jim

Could you please post the "better" patch?

Hi

I am planning to port the EUCDK architecture to my DVEVM for the DM365. Could you tell me the procedure,and the considerations to be kept in mind?

Ajay