NFS: Avoid flushing data while holding directory locks in nfs_rename()

[ Upstream commit dcd21b609d4abc7303f8683bce4f35d78d7d6830 ]

The Linux client assumes that all filehandles are non-volatile for
renames within the same directory (otherwise sillyrename cannot work).
However, the existence of the Linux 'subtree_check' export option has
meant that nfs_rename() has always assumed it needs to flush writes
before attempting to rename.

Since NFSv4 does allow the client to query whether or not the server
exhibits this behaviour, and since knfsd does actually set the
appropriate flag when 'subtree_check' is enabled on an export, it
should be OK to optimise away the write flushing behaviour in the cases
where it is clearly not needed.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index a8930e6c417fc0550d734068f574ee3ed28aa39e..de4ad41b14e2a9950112b57ec28d4ef40a45648e 100644 (file)
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -1052,6 +1052,8 @@ struct nfs_server *nfs_create_server(struct fs_context *fc)
                if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
                        server->namelen = NFS2_MAXNAMLEN;
        }
+       /* Linux 'subtree_check' borkenness mandates this setting */
+       server->fh_expire_type = NFS_FH_VOL_RENAME;
 
        if (!(fattr->valid & NFS_ATTR_FATTR)) {
                error = ctx->nfs_mod->rpc_ops->getattr(server, ctx->mntfh,

diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 70660ff248b798d398e9fc164b70bb5afd6c59ed..1876978107ca10a34d0fcd55d4a4c9dbbc0abc50 100644 (file)
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -2632,6 +2632,18 @@ nfs_unblock_rename(struct rpc_task *task, struct nfs_renamedata *data)
        unblock_revalidate(new_dentry);
 }
 
+static bool nfs_rename_is_unsafe_cross_dir(struct dentry *old_dentry,
+                                          struct dentry *new_dentry)
+{
+       struct nfs_server *server = NFS_SB(old_dentry->d_sb);
+
+       if (old_dentry->d_parent != new_dentry->d_parent)
+               return false;
+       if (server->fh_expire_type & NFS_FH_RENAME_UNSAFE)
+               return !(server->fh_expire_type & NFS_FH_NOEXPIRE_WITH_OPEN);
+       return true;
+}
+
 /*
  * RENAME
  * FIXME: Some nfsds, like the Linux user space nfsd, may generate a
@@ -2719,7 +2731,8 @@ int nfs_rename(struct user_namespace *mnt_userns, struct inode *old_dir,
 
        }
 
-       if (S_ISREG(old_inode->i_mode))
+       if (S_ISREG(old_inode->i_mode) &&
+           nfs_rename_is_unsafe_cross_dir(old_dentry, new_dentry))
                nfs_sync_inode(old_inode);
        task = nfs_async_rename(old_dir, new_dir, old_dentry, new_dentry,
                                must_unblock ? nfs_unblock_rename : NULL);

diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index 9ea9f9087a7124c920c7ca152fe5df0fb3ef81fc..a9671f9300848429f7054d41e84637e1e3cf38eb 100644 (file)
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -196,6 +196,15 @@ struct nfs_server {
        char                    *fscache_uniq;  /* Uniquifier (or NULL) */
 #endif
 
+       /* The following #defines numerically match the NFSv4 equivalents */
+#define NFS_FH_NOEXPIRE_WITH_OPEN (0x1)
+#define NFS_FH_VOLATILE_ANY (0x2)
+#define NFS_FH_VOL_MIGRATION (0x4)
+#define NFS_FH_VOL_RENAME (0x8)
+#define NFS_FH_RENAME_UNSAFE (NFS_FH_VOLATILE_ANY | NFS_FH_VOL_RENAME)
+       u32                     fh_expire_type; /* V4 bitmask representing file
+                                                  handle volatility type for
+                                                  this filesystem */
        u32                     pnfs_blksize;   /* layout_blksize attr */
 #if IS_ENABLED(CONFIG_NFS_V4)
        u32                     attr_bitmask[3];/* V4 bitmask representing the set
@@ -219,9 +228,6 @@ struct nfs_server {
        u32                     acl_bitmask;    /* V4 bitmask representing the ACEs
                                                   that are supported on this
                                                   filesystem */
-       u32                     fh_expire_type; /* V4 bitmask representing file
-                                                  handle volatility type for
-                                                  this filesystem */
        struct pnfs_layoutdriver_type  *pnfs_curr_ld; /* Active layout driver */
        struct rpc_wait_queue   roc_rpcwaitq;
        void                    *pnfs_ld_data;  /* per mount point data */