NFS: idmap PipeFS notifier introduced

v2:
1) Added "nfs_idmap_init" and "nfs_idmap_quit" definitions for kernels built
without CONFIG_NFS_V4 option set.

This patch subscribes NFS clients to RPC pipefs notifications. Idmap notifier
is registering on NFS module load. This notifier callback is responsible for
creation/destruction of PipeFS idmap pipe dentry for NFS4 clients.

Since ipdmap pipe is created in rpc client pipefs directory, we have make sure,
that this directory has been created already. IOW RPC client notifier callback
has been called already. To achive this, PipeFS notifier priorities has been
introduced (RPC clients notifier priority is greater than NFS idmap one).
But this approach gives another problem: unlink for RPC client directory will
be called before NFS idmap pipe unlink on UMOUNT event and will fail, because
directory is not empty.
The solution, introduced in this patch, is to try to remove client directory
once again after idmap pipe was unlinked. This looks like ugly hack, so
probably it should be replaced in some more elegant way.

Note that no locking required in notifier callback because PipeFS superblock
pointer is passed as an argument from it's creation or destruction routine and
thus we can be sure about it's validity.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

7 files changed, 116 insertions, 12 deletions

diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index 64815b725409..ca9a4aa38dff 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -52,8 +52,8 @@
 
 #define NFSDBG_FACILITY         NFSDBG_CLIENT
 
-static DEFINE_SPINLOCK(nfs_client_lock);
-static LIST_HEAD(nfs_client_list);
+DEFINE_SPINLOCK(nfs_client_lock);
+LIST_HEAD(nfs_client_list);
 static LIST_HEAD(nfs_volume_list);
 static DECLARE_WAIT_QUEUE_HEAD(nfs_client_active_wq);
 #ifdef CONFIG_NFS_V4
diff --git a/fs/nfs/idmap.c b/fs/nfs/idmap.c
index 769274ed51c4..ff084d258c41 100644
--- a/fs/nfs/idmap.c
+++ b/fs/nfs/idmap.c
@@ -377,6 +377,7 @@ int nfs_map_gid_to_group(const struct nfs_server *server, __u32 gid, char *buf,
 #include <linux/nfs_fs.h>
 
 #include "nfs4_fs.h"
+#include "internal.h"
 
 #define IDMAP_HASH_SZ          128
 
@@ -530,6 +531,80 @@ nfs_idmap_delete(struct nfs_client *clp)
         kfree(idmap);
 }
 
+static int __rpc_pipefs_event(struct nfs_client *clp, unsigned long event,
+                              struct super_block *sb)
+{
+        int err = 0;
+
+        switch (event) {
+        case RPC_PIPEFS_MOUNT:
+                BUG_ON(clp->cl_rpcclient->cl_dentry == NULL);
+                err = __nfs_idmap_register(clp->cl_rpcclient->cl_dentry,
+                                                clp->cl_idmap,
+                                                clp->cl_idmap->idmap_pipe);
+                break;
+        case RPC_PIPEFS_UMOUNT:
+                if (clp->cl_idmap->idmap_pipe) {
+                        struct dentry *parent;
+
+                        parent = clp->cl_idmap->idmap_pipe->dentry->d_parent;
+                        __nfs_idmap_unregister(clp->cl_idmap->idmap_pipe);
+                        /*
+                         * Note: This is a dirty hack. SUNRPC hook has been
+                         * called already but simple_rmdir() call for the
+                         * directory returned with error because of idmap pipe
+                         * inside. Thus now we have to remove this directory
+                         * here.
+                         */
+                        if (rpc_rmdir(parent))
+                                printk(KERN_ERR "%s: failed to remove clnt dir!\n", __func__);
+                }
+                break;
+        default:
+                printk(KERN_ERR "%s: unknown event: %ld\n", __func__, event);
+                return -ENOTSUPP;
+        }
+        return err;
+}
+
+static int rpc_pipefs_event(struct notifier_block *nb, unsigned long event,
+                            void *ptr)
+{
+        struct super_block *sb = ptr;
+        struct nfs_client *clp;
+        int error = 0;
+
+        spin_lock(&nfs_client_lock);
+        list_for_each_entry(clp, &nfs_client_list, cl_share_link) {
+                if (clp->net != sb->s_fs_info)
+                        continue;
+                if (clp->rpc_ops != &nfs_v4_clientops)
+                        continue;
+                error = __rpc_pipefs_event(clp, event, sb);
+                if (error)
+                        break;
+        }
+        spin_unlock(&nfs_client_lock);
+        return error;
+}
+
+#define PIPEFS_NFS_PRIO         1
+
+static struct notifier_block nfs_idmap_block = {
+        .notifier_call  = rpc_pipefs_event,
+        .priority       = SUNRPC_PIPEFS_NFS_PRIO,
+};
+
+int nfs_idmap_init(void)
+{
+        return rpc_pipefs_notifier_register(&nfs_idmap_block);
+}
+
+void nfs_idmap_quit(void)
+{
+        rpc_pipefs_notifier_unregister(&nfs_idmap_block);
+}
+
 /*
  * Helper routines for manipulating the hashtable
  */
diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index d602188f889f..2b9836fe4434 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -182,6 +182,10 @@ static inline void nfs_fs_proc_exit(void)
 {
 }
 #endif
+#ifdef CONFIG_NFS_V4
+extern spinlock_t nfs_client_lock;
+extern struct list_head nfs_client_list;
+#endif
 
 /* nfs4namespace.c */
 #ifdef CONFIG_NFS_V4
diff --git a/include/linux/nfs_idmap.h b/include/linux/nfs_idmap.h
index 308c18877018..3c9eeb7da646 100644
--- a/include/linux/nfs_idmap.h
+++ b/include/linux/nfs_idmap.h
@@ -69,31 +69,32 @@ struct nfs_server;
 struct nfs_fattr;
 struct nfs4_string;
 
-#ifdef CONFIG_NFS_USE_NEW_IDMAPPER
-
+#ifdef CONFIG_NFS_V4
 int nfs_idmap_init(void);
 void nfs_idmap_quit(void);
-
-static inline int nfs_idmap_new(struct nfs_client *clp)
+#else
+static inline int nfs_idmap_init(void)
 {
         return 0;
 }
 
-static inline void nfs_idmap_delete(struct nfs_client *clp)
-{
-}
+static inline void nfs_idmap_quit(void)
+{}
+#endif
 
-#else /* CONFIG_NFS_USE_NEW_IDMAPPER not set */
+#ifdef CONFIG_NFS_USE_NEW_IDMAPPER
 
-static inline int nfs_idmap_init(void)
+static inline int nfs_idmap_new(struct nfs_client *clp)
 {
         return 0;
 }
 
-static inline void nfs_idmap_quit(void)
+static inline void nfs_idmap_delete(struct nfs_client *clp)
 {
 }
 
+#else /* CONFIG_NFS_USE_NEW_IDMAPPER not set */
+
 int nfs_idmap_new(struct nfs_client *);
 void nfs_idmap_delete(struct nfs_client *);
 
diff --git a/include/linux/sunrpc/rpc_pipe_fs.h b/include/linux/sunrpc/rpc_pipe_fs.h
index 0d1f748f76da..ca32ebd14c18 100644
--- a/include/linux/sunrpc/rpc_pipe_fs.h
+++ b/include/linux/sunrpc/rpc_pipe_fs.h
@@ -49,6 +49,11 @@ RPC_I(struct inode *inode)
         return container_of(inode, struct rpc_inode, vfs_inode);
 }
 
+enum {
+        SUNRPC_PIPEFS_NFS_PRIO,
+        SUNRPC_PIPEFS_RPC_PRIO,
+};
+
 extern int rpc_pipefs_notifier_register(struct notifier_block *);
 extern void rpc_pipefs_notifier_unregister(struct notifier_block *);
 
@@ -78,6 +83,8 @@ extern struct dentry *rpc_create_cache_dir(struct dentry *,
                                            struct cache_detail *);
 extern void rpc_remove_cache_dir(struct dentry *);
 
+extern int rpc_rmdir(struct dentry *dentry);
+
 struct rpc_pipe *rpc_mkpipe_data(const struct rpc_pipe_ops *ops, int flags);
 void rpc_destroy_pipe_data(struct rpc_pipe *pipe);
 extern struct dentry *rpc_mkpipe_dentry(struct dentry *, const char *, void *,
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index ed7c22de9319..4c6848017168 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -222,6 +222,7 @@ static int rpc_pipefs_event(struct notifier_block *nb, unsigned long event,
 
 static struct notifier_block rpc_clients_block = {
         .notifier_call  = rpc_pipefs_event,
+        .priority       = SUNRPC_PIPEFS_RPC_PRIO,
 };
 
 int rpc_clients_notifier_register(void)
diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c
index 6b417fcabdbf..bae4e71d8663 100644
--- a/net/sunrpc/rpc_pipe.c
+++ b/net/sunrpc/rpc_pipe.c
@@ -616,6 +616,22 @@ static int __rpc_rmdir(struct inode *dir, struct dentry *dentry)
         return ret;
 }
 
+int rpc_rmdir(struct dentry *dentry)
+{
+        struct dentry *parent;
+        struct inode *dir;
+        int error;
+
+        parent = dget_parent(dentry);
+        dir = parent->d_inode;
+        mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
+        error = __rpc_rmdir(dir, dentry);
+        mutex_unlock(&dir->i_mutex);
+        dput(parent);
+        return error;
+}
+EXPORT_SYMBOL_GPL(rpc_rmdir);
+
 static int __rpc_unlink(struct inode *dir, struct dentry *dentry)
 {
         int ret;