reply: [PATCH AUTOSEL 6.11 47/76] fs: nfs: fix missing refcnt by replacing folio_set_private by folio_attach_private
From: 黄朝阳 (Zhaoyang Huang)
Date: Sat Oct 05 2024 - 00:00:21 EST
>From: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
>
>[ Upstream commit 03e02b94171b1985dd0aa184296fe94425b855a3 ]
>
>This patch is inspired by a code review of fs codes which aims at folio's extra
>refcnt that could introduce unwanted behavious when judging refcnt, such
>as[1].That is, the folio passed to mapping_evict_folio carries the refcnts from
>find_lock_entries, page_cache, corresponding to PTEs and folio's private if has.
>However, current code doesn't take the refcnt for folio's private which could
>have mapping_evict_folio miss the one to only PTE and lead to call
>filemap_release_folio wrongly.
>
>[1]
>long mapping_evict_folio(struct address_space *mapping, struct folio *folio)
>{ ...
>//current code will misjudge here if there is one pte on the folio which is be
>deemed as the one as folio's private
> if (folio_ref_count(folio) >
> folio_nr_pages(folio) + folio_has_private(folio) +
>1)
> return 0;
> if (!filemap_release_folio(folio, 0))
> return 0;
>
> return remove_mapping(mapping, folio); }
>
>Signed-off-by: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
>Signed-off-by: Anna Schumaker <anna.schumaker@xxxxxxxxxx>
>Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
Sorry, this patch should be dropped since Trond has explained that there is no need to grab refcnt here as nfs_page has hold the refcount on this folio.
>---
> fs/nfs/write.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
>diff --git a/fs/nfs/write.c b/fs/nfs/write.c index
>d074d0ceb4f01..80c6ded5f74c6 100644
>--- a/fs/nfs/write.c
>+++ b/fs/nfs/write.c
>@@ -772,8 +772,7 @@ static void nfs_inode_add_request(struct nfs_page
>*req)
> nfs_lock_request(req);
> spin_lock(&mapping->i_private_lock);
> set_bit(PG_MAPPED, &req->wb_flags);
>- folio_set_private(folio);
>- folio->private = req;
>+ folio_attach_private(folio, req);
> spin_unlock(&mapping->i_private_lock);
> atomic_long_inc(&nfsi->nrequests);
> /* this a head request for a page group - mark it as having an @@
>-797,8 +796,7 @@ static void nfs_inode_remove_request(struct nfs_page
>*req)
>
> spin_lock(&mapping->i_private_lock);
> if (likely(folio)) {
>- folio->private = NULL;
>- folio_clear_private(folio);
>+ folio_detach_private(folio);
> clear_bit(PG_MAPPED,
>&req->wb_head->wb_flags);
> }
> spin_unlock(&mapping->i_private_lock);
>--
>2.43.0