Issue reading program data from the flash

[David Beltran Lacasa]

Good afternoon,

I would like to share an issue I’m facing with a stm8s105c4.

A feature of our main application is to calculate a CRC of the entire program in the flash. Unfortunately, when I read flash data from addresses close to the instruction in execution, the data read it isn’t always correct.

I’ve been able to reproduce the issue in this small program:

 main()
{
               unsigned char* p_address;
               unsigned char dataRead;
               volatile dbg;
 
               /* Loop and read the registers*/
               for (p_address = (unsigned char*)0x80ce; ((unsigned int)p_address) < 0x80FF; p_address++)
               {                             
                              dataRead = *p_address;
                              dbg++;
               }
               dataRead++;    
}

In this case, when p_address points to 0x80dc, this is what we can observe in the dissasembly and the core registers:

1)      This set the address we want to read in X

0x80d7 <main+9>            0x1E04         LDW  X,(0x04,SP) 

 

2)     This reads the value in the address point by X :

0x80d9 <main+11>           0xF6           LD   A,(X)

 

But the data stored in the accumulator (0x8b) is wrong according to the memory view:

 

Or the Srecord:

 

On the main application, all the data that was read incorrectly was where a LD or LDW instruction was placed. Also in this example, as 0x80dc has this disassembly:

0x80dc <main+14>           0x1E01         LDW  X,(0x01,SP)

Cosmic and ST Visual Develop was used for this example. I’ve also reproduced this in a stm8s105C6 using a Discovery board. A zip with the project is provided.

Any help will be appreciated. Kind regards,

David

[S.Ma · Accepted solution]

Have you tried to make the flash reading instruction reside in RAM and called from the flash program to see if there is a difference?

[David Beltran Lacasa]

I haven't tried, I had a bit of pressure to find a solution so I decided to forget about the algorithm we were using and switch to the _checksum16 routines provided by Cosmic.

Then I found out the same issue was happening:

https://community.st.com/s/question/0D50X00009XkhppSAB/cosmic-checksum16-problems

The answer to my problem is there: a software breakpoint. I wasn't familiarized with software breakpoints implementations so I didn't figured out before.

Thanks for the help.

David