Re: kernel memory leak: freeing pagetables in vmfree_area_pages in

[Hugh Dickins]

On Thu, 28 Jun 2001 [e-mail] wrote:
> I was talking about this leak 2 days back but my mail ot lost..
> ----
> we have in vfree -->
> vmfree_area_pages (calling) free_area_pmd (calling) free_area_pte (calling)
> free_page.
> The final free_page frees all the pages that are allocated to a memory
> region in vmalloc.
> Now where are we freeing pages that are allocated to page table themselves.
> For simplicity we can assume 2 level page tables (pgd == pmd)

They're not freed, but I don't see that as a memory leak.  The page
tables stay there ready for the next time a vmalloc() needs to expand
into them.  Seems sensible to me: do you have an actual case where it
wastes significant memory?

(I haven't thought it through, but I would not be surprised if there
turned out to be awkward races if those page tables were freed when
emptied: the code's probably simpler and surer the way it is.)

Hugh