NFSv42: Fix pagecache invalidation after COPY/CLONE
authorBenjamin Coddington <bcodding@redhat.com>
Tue, 16 Nov 2021 15:48:13 +0000 (10:48 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 8 Dec 2021 08:03:17 +0000 (09:03 +0100)
commit 3f015d89a47cd8855cd92f71fff770095bd885a1 upstream.

The mechanism in use to allow the client to see the results of COPY/CLONE
is to drop those pages from the pagecache.  This forces the client to read
those pages once more from the server.  However, truncate_pagecache_range()
zeros out partial pages instead of dropping them.  Let us instead use
invalidate_inode_pages2_range() with full-page offsets to ensure the client
properly sees the results of COPY/CLONE operations.

Cc: <stable@vger.kernel.org> # v4.7+
Fixes: 2e72448b07dc ("NFS: Add COPY nfs operation")
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/nfs/nfs42proc.c

index 4ebcd9dd153524d5cf93ef70ea4803d5ab92e02d..2587b1b8e2ef760ea482065c4d1ab091d8c65969 100644 (file)
@@ -362,8 +362,9 @@ static ssize_t _nfs42_proc_copy(struct file *src,
                        goto out;
        }
 
-       truncate_pagecache_range(dst_inode, pos_dst,
-                                pos_dst + res->write_res.count);
+       WARN_ON_ONCE(invalidate_inode_pages2_range(dst_inode->i_mapping,
+                                       pos_dst >> PAGE_SHIFT,
+                                       (pos_dst + res->write_res.count - 1) >> PAGE_SHIFT));
        spin_lock(&dst_inode->i_lock);
        NFS_I(dst_inode)->cache_validity |= (NFS_INO_REVAL_PAGECACHE |
                        NFS_INO_REVAL_FORCED | NFS_INO_INVALID_SIZE |