madvise:madvise_free_huge_pmd(): don't use mapcount() against large folio for sharing check

Commit fc986a3 ("mm: huge_memory: convert madvise_free_huge_pmd to
use a folio") replaced the page_mapcount() with folio_mapcount() to check
whether the folio is shared by other mapping.

It's not correct for large folios. folio_mapcount() returns the total
mapcount of large folio which is not suitable to detect whether the folio
is shared.

Use folio_estimated_sharers() which returns a estimated number of shares.
That means it's not 100% correct. It should be OK for madvise case here.

User-visible effects is that the THP is skipped when user call madvise.
But the correct behavior is THP should be split and processed then.

NOTE: this change is a temporary fix to reduce the user-visible effects
before the long term fix from David is ready.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: fc986a3 ("mm: huge_memory: convert madvise_free_huge_pmd to use a folio")
Signed-off-by: Yin Fengwei <[email protected]>
Reviewed-by: Yu Zhao <[email protected]>
Reviewed-by: Ryan Roberts <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: Kefeng Wang <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Minchan Kim <[email protected]>
Cc: Vishal Moola (Oracle) <[email protected]>
Cc: Yang Shi <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

Author: fyin1

mm/huge_memory.c

@@ -1612,7 +1612,7 @@ bool madvise_free_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma,
 	 * If other processes are mapping this folio, we couldn't discard
 	 * the folio unless they all do MADV_FREE so let's skip the folio.
 	 */
-	if (folio_mapcount(folio) != 1)
+	if (folio_estimated_sharers(folio) != 1)
 		goto out;
 
 	if (!folio_trylock(folio))