Starting an application

[Teunderbolt]

Hello,

I am using your project as an example to write my own bootloader, so I am able to update the pi pico using a network connection. There is only one thing I don't understand yet, and that is the assembler code to start the application.

asm volatile 
(
        "mov r0, %[start]\n"
        "ldr r1, =%[vtable]\n"
        "str r0, [r1]\n"
        "ldmia r0, {r0, r1}\n"
        "msr msp, r0\n"
        "bx r1\n"
        :
        : [start] "r" (sStart + 0x100), [vtable] "X" (PPB_BASE + M0PLUS_VTOR_OFFSET)
        :
);

I am able to flash a hex file to the pico's flash, but the application won't run at the moment.
I hope that when you clarify the piece of assembler code I will find the thing that goes wrong in my code.

Kind regards,
Teun

[rhulme]

Hi,

That code is effectively the same as used in the vector_into_flash function in the SDK - https://github.com/raspberrypi/pico-sdk/blob/master/src/rp2_common/boot_stage2/asminclude/boot2_helpers/exit_from_boot2.S

Here's a commented version:

asm volatile (
"mov r0, %[start]\n"    // Get address of application's vector table (start of image + 0x100)
"ldr r1, =%[vtable]\n"  // Get address of RP2040 VTOR register
"str r0, [r1]\n"        // Point RP2040 at application's vector table
"ldmia r0, {r0, r1}\n"  // Put values starting at address r0 into r0 and r1
"msr msp, r0\n"         // Set the stack pointer
"bx r1\n"               // Jump to start of application
:
: [start] "r" (sStart + 0x100), [vtable] "X" (PPB_BASE + M0PLUS_VTOR_OFFSET)
:
);

See https://developer.arm.com/documentation/dui0497/a/the-cortex-m0-processor/exception-model/vector-table for the layout of the vector table. As you can see there, the first two entries are the stack pointer and the entry point.

So it's basically just telling the processor where the application's vector table is and setting up the stack pointer before jumping to the application's entry point.

[Teunderbolt]

Hi Rhulme,

During debugging I found out, that the data I readed back from flash always was the same value: 4294967295 in decimal or 0xFFFFFFFF. When I read the data direct from flash using a debugger I got the correct values on the correct addresses, so the rp2040 is writing to flash, but for some reason I can't read from flash. Is there something that needs to be set to enable reading from flash?

Kind regards,
Teun

[rhulme]

Hi Teun,

No, you don't need to do anything to enable reading. Just read from the appropriate address.

You are reading from the correct absolute address, aren't you? Not from the offset within flash (i.e. the address you pass to flash_range_erase and flash_range_program)?

Are you seeing this problem with the example code (app.c) or with your own code? If with your own, have you confirmed that the example works correctly?

Richard

[Teunderbolt]

Hi Richard,

As it turns out the debugger is breaking the flash reading function for some reason. The debugger I used is a Segger Jlink EDU V10.1. This is something I didn't expect and this makes debugging a lot harder. I am still researching my problem and if I got something new I will let you know.

Kind regards,
Teun

[Teunderbolt]

Hi Richard,

I finaly found my problems, as it turns out I saved and disabled the interrupts in the bootloader this prevented the application to perform a sleep_ms(). You could bypass this by writing the interrupt status to a watchdog scratch register, and recall the scratch register in the application.

Also the use of the stdio.h library prevented the application to run, but this has something todo with my pc (computer, not program counter). When it is initializing the stdio, it is setting up the usb as an serial port, when this is happening my pc is power cycling the usb-bus which sometimes causes a reset.

I hope, that my problom will help other people in the future who are experiencing the same problem.

Kind regards,
Teun

[rhulme]

Ah, OK. You'd changed the bootloader. Yes, disabling interrupts for any amount of time will often cause problems.