Re: [PATCH v4 3/4] mm: Support batched unmap for lazyfree large folios during reclamation

On 26.06.25 03:17, Lance Yang wrote:

On 2025/6/26 05:03, Barry Song wrote:

quoted

On Thu, Jun 26, 2025 at 12:25 AM David Hildenbrand [off-list ref] wrote:

quoted

On 25.06.25 14:20, Lance Yang wrote:

[...]

quoted

quoted

Hmm... I have a question about the reference counting here ...

                if (vma->vm_flags & VM_LOCKED)
                        mlock_drain_local();
                folio_put(folio);
                /* We have already batched the entire folio */

Does anyone else still hold a reference to this folio after folio_put()?

The caller of the unmap operation should better hold a reference :)

Also, I am not sure why we don't perform a

folio_put_refs(folio, nr_pages);

Because we've already called folio_ref_sub(folio, nr_pages - 1);
Looking back, it’s kind of ugly, huh.

discard:
                  if (unlikely(folio_test_hugetlb(folio))) {
                          hugetlb_remove_rmap(folio);
                  } else {
                          folio_remove_rmap_ptes(folio, subpage, nr_pages, vma);
                          folio_ref_sub(folio, nr_pages - 1);
                  }

I assume Lance will send a patch? If so, remember to remove this
when switching to folio_put_refs(folio, nr_pages);

Ah, got it. Thanks for pointing that out!

Obviously I was hinting that the split refcount update can be merged 
into a single refcount update :)

-- 
Cheers,

David / dhildenb