demand_paging: eviction: add kconfig CONFIG_EVICTION_TRACKING

[dcpleung]

This adds a new kconfig for eviction algorithm which needs page tracking. When enabled, k_mem_paging_eviction_add()/_remove() and k_mem_paging_eviction_accessed() must be implemented. If an algorithm does not do page tracking, there is no need to implement these functions, and no need for the kernel MMU code to call into empty functions. This should save a few function calls and some CPU cycles.

Note that arm64 unconditionally calls those functions so forces CONFIG_EVICTION_TRACKING to be enabled there.

[npitre]

OK for the Kconfig symbol, however it might be nicer to define
empty static inline stubs when disabled rather than adding more #ifdef
in kernel/mmu.c.

[dcpleung]

Being static inline does not necessarily eliminate the related code when no optimization is used. These inlined functions are converted into non-inlined functions and will be called. Even changing the static inline to static ALWAYS_INLINE will simply replace the assembly call into nop. So they are not totally eliminated. One could argue that this is expected from using no optimizations, but it would be great if the code is actually removed when kconfig is disabled, such that there are fewer surprised when debugging.

[npitre]

Being `static inline` does not necessarily eliminate the related code when no optimization is used.

Then why would you even care? Code readability is more important than optimizing the "no optimization" case.

[dcpleung]

Because lots of time I do debug with no optimizations.

[npitre]

Because lots of time I do debug with no optimizations.

Then you must be used to skip over all those empty inline stubs already.

[peter-mitsis]

I'm OK with the #ifdef stuff, but if it is blocking issue, would defining the affected routines as empty macros be a satisfactory compromise? It is an approach with precedence and should hopefully work well with the no optimization scenario.

[dcpleung]

Because lots of time I do debug with no optimizations.
Then you must be used to skip over all those empty inline stubs already.

Well... doesn't mean that I want to make it worse.

Besides, there is now code being run which should not have been there, for example the for loop in z_mem_manage_init() to add page frames in queue. Instead of skipping over the whole loop, there would be actual looping to call into a function that does nothing (but there is still a call in assembly and all the stack manipulation). Semantically speaking, it would not be the same code. Therefore if possible, I would like those part of code to not appear in compiled code, regardless of optimization level.

[dcpleung]

I guess a comprise can be done via IS_ENABLED().