On Tue, Oct 12, 2021 at 5:03 AM Marek Szyprowski
<m.szyprowski@xxxxxxxxxxx> wrote:
>
> Decrease nr_thps counter in file's mapping to ensure that the page cache
> won't be dropped excessively on file write access if page has been
> already splitted.
>
> Signed-off-by: Marek Szyprowski <m.szyprowski@xxxxxxxxxxx>
> Fixes: 09d91cda0e82 ("mm,thp: avoid writes to file with THP in pagecache")
> Fixes: 06d3eff62d9d ("mm/thp: fix node page state in split_huge_page_to_list()")
Reviewed-by: Yang Shi <shy828301@xxxxxxxxx>
> ---
> I've analyzed the code a few times but either I missed something or the
> nr_thps counter is not decremented during the THP split on non-shmem file
> pages.
> ---
> mm/huge_memory.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index ec2bb93f7431..a6c2ba59abcd 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -2709,10 +2709,12 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
> }
> spin_unlock(&ds_queue->split_queue_lock);
> if (mapping) {
> - if (PageSwapBacked(head))
> + if (PageSwapBacked(head)) {
> __dec_node_page_state(head, NR_SHMEM_THPS);
> - else
> + } else {
> __dec_node_page_state(head, NR_FILE_THPS);
> + filemap_nr_thps_dec(mapping);
> + }
> }
>
> __split_huge_page(page, list, end, flags);
> --
> 2.17.1
>
>
