why this MPU_ALIGN(__ramfunc_size) can work in linker file before __ramfunc_size is defined

[pasuzw]

I happened to see the following code from SDK_FOLDER/zephyr/arch/common/ramfunc.ld:

_SECTION_DATA_PROLOGUE(.ramfunc,,)
{
	MPU_ALIGN(__ramfunc_size);
	__ramfunc_start = .;
	*(.ramfunc)
	*(".ramfunc.*")
	MPU_ALIGN(__ramfunc_size);
	__ramfunc_end = .;
} GROUP_DATA_LINK_IN(RAMABLE_REGION, ROMABLE_REGION)
__ramfunc_size = __ramfunc_end - __ramfunc_start;
__ramfunc_load_start = LOADADDR(.ramfunc);_

My question is that why __ramfunc_size could be used together with MPU_ALIGN() even before __ramfunc_size is defined/calculated.

Background info:

	#if defined(CONFIG_MPU_REQUIRES_POWER_OF_TWO_ALIGNMENT)
		#define MPU_ALIGN(region_size) \
			. = ALIGN(MPU_MIN_SIZE); \
			. = ALIGN( 1 << LOG2CEIL(region_size))

[pasuzw]

anybody could help here ?

[yashi]

Becase MPU_ALIGN() is just a macro which exapands with __ramfunc_size as is?

[nbhume]

__ramfunc_size can be used before definition because:
• GNU ld allows forward symbol references
• Linker scripts are declarative
• Expressions are resolved after full script parsing
• Symbol values are finalized during layout phase
There will not be sequential execution like C.