Re: [PATCH 2/5] mm,hugetlb: Document the reason to lock the folio in the faulting path

[David Hildenbrand]

> I see, this makes a lot of sense, thanks for walking me through David!

Well, I hope the same logic applies to hugetlb :D

> Alright, then, with all this clear now we should:
>
> - Not take any locks on hugetlb_fault()->hugetlb_wp(), hugetlb_wp() will take it
>    if it's an anonymous folio (re-use check)
> - Drop the lock in hugetlb_no_page() after we have mapped the page in
>    the pagetables
> - hugetlb_wp() will take the lock IFF the folio is anonymous
>
> This will lead to something like the following:
>
>   diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>   index dfa09fc3b2c6..4d48cda8a56d 100644
>   --- a/mm/hugetlb.c
>   +++ b/mm/hugetlb.c
>   @@ -6198,6 +6198,8 @@ static vm_fault_t hugetlb_wp(struct vm_fault *vmf)
>    	 * in scenarios that used to work. As a side effect, there can still
>    	 * be leaks between processes, for example, with FOLL_GET users.
>    	 */
>   +	if (folio_test_anon(old_folio))
>   +		folio_lock(old_folio);

If holding the PTL, this would not work. You'd have to unlock PTL, lock 
folio, retake PTL, check pte_same.

>    	if (folio_mapcount(old_folio) == 1 && folio_test_anon(old_folio)) {
>    		if (!PageAnonExclusive(&old_folio->page)) {
>    			folio_move_anon_rmap(old_folio, vma);
>   @@ -6212,6 +6214,8 @@ static vm_fault_t hugetlb_wp(struct vm_fault *vmf)
>    	}
>    	VM_BUG_ON_PAGE(folio_test_anon(old_folio) &&
>    		       PageAnonExclusive(&old_folio->page), &old_folio->page);
>   +	if (folio_test_anon(old_folio))
>   +		folio_unlock(old_folio);

[...]

>    
> This should be patch#2 with something like "Sorting out locking" per
> title, and maybe explaining a bit more why the lock in hugelb_wp for
> anonymous folios.

Jup.


--
Cheers,

David / dhildenb