Background:
On some microcontrollers memory is split into PMEM (programm memory) and DMEM (data memory) with only PMEM being executable (e.g. instructions can’t even be fetched from DMEM). PMEM can be changed at runtime for e.g. updates, but that is relatively seldom and can be easily monitored. This is useful to e.g. prevent time-of-check-time-of-use-attacks for attestation.

Question:
My question is how/if it is possible to adapt this to unified memory so that i get notified when there are changes to the executed code (e.g. by malicous actor)?

So while there is one unified memory some part/address-range should store the programm and only that memory-range should be used for instructions.
Basically if the programm-memory-part is being written to/changed i want to get notified and if some part other than the programm-memory-part is somehow getting executed, i also want to get notified.
Is that possible on e.g. common arm-platforms or similar (especially without operating system support)?

(Like one could write code that checks the processors signals for whether its currently writing and whether its writing to a predefined area, right?
With that technology could someone limit executing to the predefined area?
Could that maybe be done using a Trusted Execution Environment (e.g. keeping part of the code there or checking from there the other parts)?
And would that likely have a very high overhead?)

Further background:
E.g. this paper (On the TOCTOU Problem in Remote Attestation) seems to describe monitoring signals ???????????? (= whether MCU is writing to memory), ???????????????????? (=address for an MCU memory-access), ???????????????????????????? (memory address being accessed by DMA, if any) , and ???????????????????? (= signals that DMA is currently accessing memory) to detect changes to the attested range (AR) of the memory.