I get the below error message when I start the ULINK debugger with my KEIL MCBSTM32 eval board:
Cannot access Memory *** error 57: illegal address (0x8000000)
I've searched KEIL's website for error 57, but couldn't find anything. I've searched uVsion help, but couldn't find anything. Can someone point me to where these types of errors are documented?
The debugger is actually stopped at address 0x80001A4, and these are the instructions before and at that line:
0x080001A0 F002BBF6 B.W 0x080002990 0x080001A4 4804 LDR r0,[PC,#32] ; @0x080001C8
and this is what's at 0x080001C8: 0x080001C8 00ED LSLS r5,r5,#3
and this is what's at 0x08002990: 0x08002990 4770 BX lr
Address 0x8000000 is the start of flash ROM for the uC on this eval board (an STM32F103RBT6). I'm using the default KEIL assembly startup code (STM32F103x.s) and the default project options for where ROM and RAM is (IROM1 at 0x8000000 with size = 0x20000, and IRAM1 at 0x20000000 with size = 0x5000).
If I start the debugger then stop it, I get this error message:
ULINK - Cortex-M Error Could not stop Cortex-M device! Please check the JTAG cable
I found some knowledgebase articles that suggested slowing down the JTAG clock, using internal or external 3.3V inside the ULINK, and/or making sure that a hardware reset was used with the ULINK; so I tried those suggestions. Unfortunately, they didn't change my symptoms.
I've been trying to find somwhere where I could insert a breakpoint to try to step through the startup code, but uVision won't let me put a breakpoint anywhere in the startup assembly code (and yes, I do have the "Debug Information" checkbox checked in the project options). I've unchecked the debugger "Reset and Run" and "Run to main()" options, but when I start the debugger and try to single step, I can only step a few times and then the code appears to start and never stop. If I hit the "halt" button, I get the same error message about not being able to stop the uC.
The debugger always starts at address 0x080001A4 with the error message about not being able to access memory.
Can anyone suggest something else to do that might fix these problems?
Thanks, Dave.
Hi David,
You are right. I thought that I had read this thread before, so I didn't scroll up, and then lost a lot of information in the earlier posts.
I have a question, who creates/builds the problematic project? Is that some example you downloaded from Internet?
If it is a project not created/built by yourself, then I will suggest that: 1.Extract all the source code out. 2.Create a totally new project, include all the source code in. 3.Configure the new project as your need.
John,
The project is of my own creation. I did start it from scratch; something I've done for many, many years using KEIL tools on 805x, C16x, and ARM projects.
I did restart it, from scratch again, using a different STM32F eval board from a different vendor. The fact that I couldn't reliably (if at all) start the debugger until I followed a suggestion from someone else to load the code into the flash ROM manually (and disable starting the debugger from doing that) is what makes me think it's not my code/project that's the problem. My code runs fine on both eval boards, and I can use the debugger with ease on both eval boards, as long as I prevent starting the debugger from trying to load the code into the board. Well, at least that's how it's working now.
So... the immediate problem has been solved (I can use the debugger now). However, the underlying cause is still there (starting the debugger and letting it load code into my eval boards prevents the debugger from being able to set a breakpoint or execute a single line of code).
What's worse is I still haven't been able to find this error documented anywhere; or have anyone tell me what it really means and what causes it.
I don't know much about this, but I am really curious about such a problem.
I have an assumption. (Sorry for my limited technical ability and limited English ability.)
[ I did try the Blinky project, and it worked. ]
Maybe, when your specific application is running, the MCU enters a specific status, so that the MCU can not be halt/reset properly.
If this is the case, then, after you fully erase the program on the MCU, the problem should disappear for once.
My suggestions are: - use Blinky as base and rename it, add your files and try it - compare all the settings in Bliny with settings in your project - make a while (1) loop at the beginning of main function to check if your code is producing problems
You should be able to find out what the problem was by using last 2 suggestions, and by using first one you should be able to see if a setting in a project is a problem.
Since I can't even execute the 1st line of code when I have this error, I don't see how it could be my app's fault that the debugger won't run.
But, I went ahead and tried your suggestion anyway after re-enabling "Download to Flash" under "Download Options" in the "Cortex-M Target Driver Setup". As soon as I did that, my error 57 problem re-appeared. I then manually erased the flash ROM and tried to start the debugger again, but error 57 came right back.
There are over 70 source files in my project. I'll try putting them into the blinky project and see if that does anything.
Dave.
I copied the Blinky OPT & UV2 files to a temp directory, stripped out all the source files, put all mine in, did a build all and tried to start the debugger... same problem. I copied the startup ASM code from the Blinky project to my project, rebuilt it, etc., but it made no difference.
So, I tried the MCBSTM32's Blinky project, and now it has the same problem too. It worked the last time I tried it.
Argh! Quite frustrating.
OOPS... I just got my MCBSTM32 board back from a coworker and didn't notice that he'd moved the BOOT0 jumper (he's working on a bootstrap loader). I had been using the STM3210E eval board till this morning. I put the BOOT0 jumper back to where it was and now at least both my project and the Blinky project run once more.
However, they both exhibit the error 57 problem when I let the debugger download code to flash, and the debugger works with both of them when I disable the debugger from downloading code to flash.