Running linux on x86.
I am looking at the function free_pgd_range. It walks through the radix tree page table and "free" the page table at each level. For example, at pte level, free_pte_range calls pte_free_tlb which update the tlb stats and call page destructor. I am confused because I don't know where the linux kernel actually free a page allocated for a page table when destroy a process and exit from its associated mm_struct.
I also look into pgd_free and zap_page_range functions; they are critical while exiting from a mm_struct, but I don't think they free any pages used for page tables.

Any insight/pointer would be helpful! Thanks in advance!

This highly-technical question should be moved to Linux - Kernel.

This defers between 32 bit and 64 bit address(virtual).
if you go through this structure, you will find page table address. Out of this address, lower 12 bits used to refer address of particular page(Assuming that you have page size 4K), rest of the bits are used for multilevel-page table indexing.

Sorry I should have specified the arch well enough. I am running on x86-64. I know how those page tables are designed, but I guess my question in one sentence is that when and where do those page tables get freed?

Thanks for the suggestion. Let me migrate it there.

Duplicate of https://www.linuxquestions.org/quest...ed-4175711085/