Some issues regarding IBL when using MAD utility, C6678

Dear Allen,

Thanks for the reply.

So to understand the sequence for MAD application, let's take NAND boot as an example.

1. Generate c6678-le.bin through MAP tool.

2. As requested, modify bootFormat, startAddress and branch Address in i2cConfig.gel

        - No need to modify ibl/src/util/iblConfig/src/device.c.  There are also the 3 items in device.c, what is the difference with those in i2cConfig.gel?

        - No need to re-generate i2crom_0x51_c6678_le.bin, use the default one in ibl\src\make\bin.

        - No need to re-generate i2cparam_0x51_c6678_le_0x500.out, use the default one in ibl\src\make\bin.

3. No need to re-program i2crom_0x51_c6678_le.bin into EEPROM since the file is not modified.

4. In CCS, program IBL Configuration into EEPROM: i2cparam_0x51_c6678_le_0x500.out + i2cConfig.gel

5. Write c6678-le.bin into NAND FLASH.

BTW, 

a. Usually user does NOT need to do the things stated in ibl_single_binary.txt.  But what ibl_single_binary.txt for ?

b. Anything to do with ibl\src\util\i2cConfig ?

Thanks.

Boll

Hi Boll,

In your step3, do you mean you have already programed the i2crom_0x51_c6678_le.bin into EEPROM? If so, there is no need to re-program, or you will need to program it at least one time.

ibl_single_binary.txt states how to generate the binary of ibl for the dedicated platform, so if you are working on some other EVM, you need to follow the steps in the txt to generate the customized ibl for that platform. There is originaly only binary for EVM6670/6678 within mcsdk. The source files within ibl\src\util\i2cConfig is related to the steps of ibl_single_binary.txt. So if you are using EVM6678, no need to edit these files.

Allen

Dear Allen,

Yes, in step 3 I have already programmed IBL bin into EEPROM when I perform single core booting example.

Actually, I am referring to http://processors.wiki.ti.com/index.php/MAD_Utils_User_Guide.  The only difference is that I am experimenting to boot from NOR FLASH.

I encounter problem in the last step: Application deployment on device, which make me scratch my head.

I can not figure out where I am wrong since I follow the instructions step by step.

Could you please help to review my procedure?

1. Build mad-utils\mad-loader\examples in MinGW

         - I build it in prelinker mode
                   build_examples_msys.sh C6678 little relocatable          => app_1.exe and app_2.exe generated.

         - Error will be encountered for prelinker by pass mode.  Maybe it is related with linker command file.

           "lnk_C6678.cmd", line 13: error: placement fails for object ".text", size 0x5e00 (page 0). 
            Available ranges:  PMEM size: 0x5000 unused 0x4ec0 max hole: 0x4ec0
            

2. No modification in deployment_template_C6678_windows.json since it is same as instructed.

3. Build ROMSFS:    maptool.py config-files\maptoolCfg_C6678_windows.json      c6678-le.bin and mad_load_symbols.gel generated in .\images folder

4. Set boot mode to NO BOOT.

5. Write c6678-le.bin to NOR FLASH
        - Copy c6678-le.bin to tools\writer\nor\evmc6678l\bin, and rename to app.bin
        - Load app.bin to memory 0x9E000000 of Core 0
        - Write to NOR FLASH.      => success displayed in CCS console

6.  Program IBL configuration to EEPROM
        - Modify \ibl\src\make\bin\i2cConfig.gel
                 ibl.bootModes[0].u.norBoot.bootFormat = ibl_BOOT_FORMAT_BBLOB;
                 ibl.bootModes[0].u.norBoot.blob[0][0].startAddress  = 0x9e000000;
                 ibl.bootModes[0].u.norBoot.blob[0][0].branchAddress = 0x9e001040;
        - Write: i2cparam_0x51_c6678_le_0x500.out + i2cConfig.gel          => success displayed in CCS console

7. Power off EVM; set boot mode to IBL NOR boot on image 0; power on.
    The following info is displayed in UART console.

                 IBL: PLL and DDR Initialization Complete
                 IBL Result code 00
                 IBL: Booting from NOR

8. Launch CCS, connect Core 0

        In CCS console:
                 C66xx_0: GEL Output: No initialization performed since bootmode = 0x00000005 
                 C66xx_0: GEL Output: You can manually initialize with GlobalDefaultSetup

9. Load mad_load_symbols.gel,  Run Scripts->refresh_symbols_app1:
                 C66xx_0: GEL Output: Symbols loaded for app1 C66xx_0: GEL Output: Symbols loaded for printf_dll

        However the variable "signature" is not displayed in "Variables" View, as is supposed in the MAD user guide, no matter I resume Core 0 or not.

        Q: do I need manually to perform "Run->Load->Load Symbols" ???  If needed, What to load since there is no .out file generated?

Thank you very much.

Boll

Hi Boll,

According to your figure, the signature is located in 0x00000004, which is an unavailable address in C6678EVM. It seems some sections are not allocated properly. I will try to repeat the situation on my EVM and let you know if I find the reason.

Allen 

Dear Allen,

1. I generate app_1.exe and app_2.exe in MinGW shell: build_examples_msys.sh C6678 little relocatable

2. I do not modify make file.

3. The only modification made in build_examples_msys.sh is to set CGTools to 7.3.1 since I am using the 7.3.1.

4. I do not touch deployment_template_C6678_windows.json, and I think the .far section is already included.

{
    "name" : "ddr-data",
    "vaddr" : "0xD0000000",
    "paddr" : [ "0x800000000", "0x801000000", "0x802000000", "0x803000000", "0x804000000", "0x805000000", "0x806000000", "0x807000000" ],
    "size" : "0x1000000",
    "secNamePat" : ["stack", "^.far$", "args", "neardata", "fardata", "rodata"],
    "cores" : [0,1,2,3,4,5,6,7],
    "permissions" : ["UR", "UW", "SR", "SW"],
    "cacheEnable" : true,
    "prefetch" : true,
    "priority" : 0,
    "shared" : false
}

BTW, I am evaluating C6678EVM, and using mcsdk_2_00_05_17, CGTool_7.3.1, MinGW 3.1.17, Python 2.7.2

Which versions are you using?

Boll

Dear Allen,

In app_1.c, signature is defined as a uninitialized array: char signature[128];

When I change it to initialized, everything will be OK: char signature[128] = " ";

However, in my memory both initialized and uninitialized array/structure/union are placed in .far segment.

Is the behavior changed in the new compiler of C66x, such as CGT 7.3.1?

In lkn_C6678.cmd:

     SECTIONS
     {
           .text > PMEM
           .const > PMEM
           .far > DMEM
           .fardata > DMEM
           .neardata > DMEM
           .stack > DMEM
     }

And in deployment_template_C6678_windows.json:

           "secNamePat" : ["stack", "^.far$", "args", "neardata", "fardata", "rodata"]

Boll

Dear Allen,

I have tried your suggestion, it does not help.

BTW, when I build app_1 with "static" option, a map file can be referred to.

• When signature is initialized, it will be placed in .neardata.
• When signature is uninitialized, it will be placed in .bss.
  Then I add "bss" into "secNamePat", still no help.

Hi guys, I've some questions regarding these configurations on the .gel file:

ibl.bootModes[1].u.nandBoot.blob[0][0].startAddress = 0x9E000000;
ibl.bootModes[1].u.nandBoot.blob[0][0].branchAddress = 0x9E001040;

I'm able to run other applications that i've made using these configurations of start adress em brach adress,not only the MAD example.

So, my question is: what exactly those 2 adresses are needed for?Why they work with another programs, not only no MAD example?

They're some kind of default adresses to boot multicore applications?

Thanks