In order to comply with the application binary interface (ABI) of the CPU, it is required to assign space in memory for the execution stack. This can be done in different ways, but usually it is preferable to mark the end of the stack space in the linker script, and associate the stack space to a specific area in RAM not in use by any section.

The address obtained through the END_STACK symbol, exported by the linker script, points to the end of an unused area in RAM. As mentioned earlier, its value must be stored at the beginning of the vector table, at address 0 in our case, just before the IV. The address of the end of the stack has to be constant, and cannot be calculated at runtime, because the IV content is stored in the flash memory and thus cannot be modified later on.

Properly sizing the execution stack in memory is a delicate task that includes the assessment of the whole code base, keeping in mind stack usage from local variables and the depth of the call trace at any time during the execution. The analysis of all the factors related to stack usage and troubleshooting are part of a wider topic that is covered in the next chapter. Our simple startup code provided here has a stack size that is big enough to contain the local variables and the functions call stack, as it is mapped by the linker script as far as possible from the .bss and .data sections. More aspects about the placement of the stack are considered through Chapter 5, Memory Management.