From:   Benjamin Coddington <bcodding@redhat.com>
To:     trond.myklebust@hammerspace.com, anna.schumaker@netapp.com
Cc:     linux-nfs@vger.kernel.org
Subject: [PATCH 2/3] NFSv42: Don't drop NFS_INO_INVALID_CHANGE if we hold a delegation
Date:   Tue, 16 Nov 2021 08:49:23 -0500
Message-Id: <c91e224b847e697e42b25cdc36cd164a61ad1ade.1637069577.git.bcodding@redhat.com>
In-Reply-To: <cover.1637069577.git.bcodding@redhat.com>
References: <cover.1637069577.git.bcodding@redhat.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

The nfs_set_cache_invalid() helper drops NFS_INO_INVALID_CHANGE if we hold
a delegation, but after a copy or clone the change attribute can be updated
on the server.  After commit b6f80a2ebb97 "NFS: Fix open coded versions of
nfs_set_cache_invalid() in NFSv4", the client stopped updating the change
attribute after copy or clone while holding a read delegation.

We can use NFS_INO_REVAL_PAGECACHE to help nfs_set_cache_invalid() know
when we really want to keep NFS_INO_INVALID_CHANGE, even if the client
holds a delegation.

Fixes: b6f80a2ebb97 ("NFS: Fix open coded versions of nfs_set_cache_invalid() in NFSv4")
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
---
 fs/nfs/inode.c     | 4 +++-
 fs/nfs/nfs42proc.c | 3 ++-
 2 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 853213b3a209..296ed8ea3273 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -202,7 +202,9 @@ void nfs_set_cache_invalid(struct inode *inode, unsigned long flags)
 			flags &= ~(NFS_INO_INVALID_MODE |
 				   NFS_INO_INVALID_OTHER |
 				   NFS_INO_INVALID_XATTR);
-		flags &= ~(NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE);
+		if (!(flags & NFS_INO_REVAL_PAGECACHE))
+			flags &= ~NFS_INO_INVALID_CHANGE;
+		flags &= ~NFS_INO_INVALID_SIZE;
 	} else if (flags & NFS_INO_REVAL_PAGECACHE)
 		flags |= NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE;
 
diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c
index bbcd4c80c5a6..fc3c36e1f656 100644
--- a/fs/nfs/nfs42proc.c
+++ b/fs/nfs/nfs42proc.c
@@ -292,7 +292,8 @@ static void nfs42_copy_dest_done(struct inode *inode, loff_t pos, loff_t len)
 	spin_lock(&inode->i_lock);
 	if (newsize > i_size_read(inode))
 		i_size_write(inode, newsize);
-	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
+	nfs_set_cache_invalid(inode, NFS_INO_REVAL_PAGECACHE |
+					     NFS_INO_INVALID_CHANGE |
 					     NFS_INO_INVALID_CTIME |
 					     NFS_INO_INVALID_MTIME |
 					     NFS_INO_INVALID_BLOCKS);
-- 
2.31.1