[v2] nfsd: Clone should commit src file metadata too
diff mbox series

Message ID 20191218195723.395277-1-trond.myklebust@hammerspace.com
State New
Headers show
Series
  • [v2] nfsd: Clone should commit src file metadata too
Related show

Commit Message

Trond Myklebust Dec. 18, 2019, 7:57 p.m. UTC
vfs_clone_file_range() can modify the metadata on the source file too,
so we need to commit that to stable storage as well.

Reported-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---

 v2: Reverse order of vfs_fsync_range()/commit_inode_metadata() for
     efficiency as suggested by Dave Chinner.

 fs/nfsd/vfs.c | 19 ++++++++++++++-----
 1 file changed, 14 insertions(+), 5 deletions(-)

Comments

Dave Chinner Dec. 18, 2019, 9:12 p.m. UTC | #1
On Wed, Dec 18, 2019 at 02:57:23PM -0500, Trond Myklebust wrote:
> vfs_clone_file_range() can modify the metadata on the source file too,
> so we need to commit that to stable storage as well.
> 
> Reported-by: Dave Chinner <david@fromorbit.com>
> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

Looks ok to me now.

Acked-by: Dave Chinner <dchinner@redhat.com>
J. Bruce Fields Dec. 19, 2019, 10:44 p.m. UTC | #2
On Thu, Dec 19, 2019 at 08:12:51AM +1100, Dave Chinner wrote:
> On Wed, Dec 18, 2019 at 02:57:23PM -0500, Trond Myklebust wrote:
> > vfs_clone_file_range() can modify the metadata on the source file too,
> > so we need to commit that to stable storage as well.
> > 
> > Reported-by: Dave Chinner <david@fromorbit.com>
> > Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
> 
> Looks ok to me now.
> 
> Acked-by: Dave Chinner <dchinner@redhat.com>

Thanks!  Applying for 5.6, with Dave's ACK.--b.

Patch
diff mbox series

diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index f0bca0e87d0c..82cf80dde5c7 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -280,19 +280,25 @@  nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name,
  * Commit metadata changes to stable storage.
  */
 static int
-commit_metadata(struct svc_fh *fhp)
+commit_inode_metadata(struct inode *inode)
 {
-	struct inode *inode = d_inode(fhp->fh_dentry);
 	const struct export_operations *export_ops = inode->i_sb->s_export_op;
 
-	if (!EX_ISSYNC(fhp->fh_export))
-		return 0;
-
 	if (export_ops->commit_metadata)
 		return export_ops->commit_metadata(inode);
 	return sync_inode_metadata(inode, 1);
 }
 
+static int
+commit_metadata(struct svc_fh *fhp)
+{
+	struct inode *inode = d_inode(fhp->fh_dentry);
+
+	if (!EX_ISSYNC(fhp->fh_export))
+		return 0;
+	return commit_inode_metadata(inode);
+}
+
 /*
  * Go over the attributes and take care of the small differences between
  * NFS semantics and what Linux expects.
@@ -537,6 +543,9 @@  __be32 nfsd4_clone_file_range(struct file *src, u64 src_pos, struct file *dst,
 	if (sync) {
 		loff_t dst_end = count ? dst_pos + count - 1 : LLONG_MAX;
 		int status = vfs_fsync_range(dst, dst_pos, dst_end, 0);
+
+		if (!status)
+			status = commit_inode_metadata(file_inode(src));
 		if (status < 0)
 			return nfserrno(status);
 	}