vfs: make it possible to access the dentry hash/len as one 64-bit entry
[linux-3.10.git] / net / sunrpc / rpc_pipe.c
index 7df92d2..fd24239 100644 (file)
@@ -16,9 +16,9 @@
 #include <linux/namei.h>
 #include <linux/fsnotify.h>
 #include <linux/kernel.h>
+#include <linux/rcupdate.h>
 
 #include <asm/ioctls.h>
-#include <linux/fs.h>
 #include <linux/poll.h>
 #include <linux/wait.h>
 #include <linux/seq_file.h>
 #include <linux/workqueue.h>
 #include <linux/sunrpc/rpc_pipe_fs.h>
 #include <linux/sunrpc/cache.h>
+#include <linux/nsproxy.h>
+#include <linux/notifier.h>
 
-static struct vfsmount *rpc_mount __read_mostly;
-static int rpc_mount_count;
+#include "netns.h"
+#include "sunrpc.h"
+
+#define RPCDBG_FACILITY RPCDBG_DEBUG
+
+#define NET_NAME(net)  ((net == &init_net) ? " (init_net)" : "")
 
 static struct file_system_type rpc_pipe_fs_type;
 
@@ -38,7 +44,21 @@ static struct kmem_cache *rpc_inode_cachep __read_mostly;
 
 #define RPC_UPCALL_TIMEOUT (30*HZ)
 
-static void rpc_purge_list(struct rpc_inode *rpci, struct list_head *head,
+static BLOCKING_NOTIFIER_HEAD(rpc_pipefs_notifier_list);
+
+int rpc_pipefs_notifier_register(struct notifier_block *nb)
+{
+       return blocking_notifier_chain_cond_register(&rpc_pipefs_notifier_list, nb);
+}
+EXPORT_SYMBOL_GPL(rpc_pipefs_notifier_register);
+
+void rpc_pipefs_notifier_unregister(struct notifier_block *nb)
+{
+       blocking_notifier_chain_unregister(&rpc_pipefs_notifier_list, nb);
+}
+EXPORT_SYMBOL_GPL(rpc_pipefs_notifier_unregister);
+
+static void rpc_purge_list(wait_queue_head_t *waitq, struct list_head *head,
                void (*destroy_msg)(struct rpc_pipe_msg *), int err)
 {
        struct rpc_pipe_msg *msg;
@@ -51,32 +71,53 @@ static void rpc_purge_list(struct rpc_inode *rpci, struct list_head *head,
                msg->errno = err;
                destroy_msg(msg);
        } while (!list_empty(head));
-       wake_up(&rpci->waitq);
+       wake_up(waitq);
 }
 
 static void
 rpc_timeout_upcall_queue(struct work_struct *work)
 {
        LIST_HEAD(free_list);
-       struct rpc_inode *rpci =
-               container_of(work, struct rpc_inode, queue_timeout.work);
-       struct inode *inode = &rpci->vfs_inode;
+       struct rpc_pipe *pipe =
+               container_of(work, struct rpc_pipe, queue_timeout.work);
        void (*destroy_msg)(struct rpc_pipe_msg *);
+       struct dentry *dentry;
 
-       spin_lock(&inode->i_lock);
-       if (rpci->ops == NULL) {
-               spin_unlock(&inode->i_lock);
-               return;
+       spin_lock(&pipe->lock);
+       destroy_msg = pipe->ops->destroy_msg;
+       if (pipe->nreaders == 0) {
+               list_splice_init(&pipe->pipe, &free_list);
+               pipe->pipelen = 0;
        }
-       destroy_msg = rpci->ops->destroy_msg;
-       if (rpci->nreaders == 0) {
-               list_splice_init(&rpci->pipe, &free_list);
-               rpci->pipelen = 0;
+       dentry = dget(pipe->dentry);
+       spin_unlock(&pipe->lock);
+       if (dentry) {
+               rpc_purge_list(&RPC_I(dentry->d_inode)->waitq,
+                              &free_list, destroy_msg, -ETIMEDOUT);
+               dput(dentry);
        }
-       spin_unlock(&inode->i_lock);
-       rpc_purge_list(rpci, &free_list, destroy_msg, -ETIMEDOUT);
 }
 
+ssize_t rpc_pipe_generic_upcall(struct file *filp, struct rpc_pipe_msg *msg,
+                               char __user *dst, size_t buflen)
+{
+       char *data = (char *)msg->data + msg->copied;
+       size_t mlen = min(msg->len - msg->copied, buflen);
+       unsigned long left;
+
+       left = copy_to_user(dst, data, mlen);
+       if (left == mlen) {
+               msg->errno = -EFAULT;
+               return -EFAULT;
+       }
+
+       mlen -= left;
+       msg->copied += mlen;
+       msg->errno = 0;
+       return mlen;
+}
+EXPORT_SYMBOL_GPL(rpc_pipe_generic_upcall);
+
 /**
  * rpc_queue_upcall - queue an upcall message to userspace
  * @inode: inode of upcall pipe on which to queue given message
@@ -88,30 +129,31 @@ rpc_timeout_upcall_queue(struct work_struct *work)
  * initialize the fields of @msg (other than @msg->list) appropriately.
  */
 int
-rpc_queue_upcall(struct inode *inode, struct rpc_pipe_msg *msg)
+rpc_queue_upcall(struct rpc_pipe *pipe, struct rpc_pipe_msg *msg)
 {
-       struct rpc_inode *rpci = RPC_I(inode);
        int res = -EPIPE;
+       struct dentry *dentry;
 
-       spin_lock(&inode->i_lock);
-       if (rpci->ops == NULL)
-               goto out;
-       if (rpci->nreaders) {
-               list_add_tail(&msg->list, &rpci->pipe);
-               rpci->pipelen += msg->len;
+       spin_lock(&pipe->lock);
+       if (pipe->nreaders) {
+               list_add_tail(&msg->list, &pipe->pipe);
+               pipe->pipelen += msg->len;
                res = 0;
-       } else if (rpci->flags & RPC_PIPE_WAIT_FOR_OPEN) {
-               if (list_empty(&rpci->pipe))
+       } else if (pipe->flags & RPC_PIPE_WAIT_FOR_OPEN) {
+               if (list_empty(&pipe->pipe))
                        queue_delayed_work(rpciod_workqueue,
-                                       &rpci->queue_timeout,
+                                       &pipe->queue_timeout,
                                        RPC_UPCALL_TIMEOUT);
-               list_add_tail(&msg->list, &rpci->pipe);
-               rpci->pipelen += msg->len;
+               list_add_tail(&msg->list, &pipe->pipe);
+               pipe->pipelen += msg->len;
                res = 0;
        }
-out:
-       spin_unlock(&inode->i_lock);
-       wake_up(&rpci->waitq);
+       dentry = dget(pipe->dentry);
+       spin_unlock(&pipe->lock);
+       if (dentry) {
+               wake_up(&RPC_I(dentry->d_inode)->waitq);
+               dput(dentry);
+       }
        return res;
 }
 EXPORT_SYMBOL_GPL(rpc_queue_upcall);
@@ -125,29 +167,26 @@ rpc_inode_setowner(struct inode *inode, void *private)
 static void
 rpc_close_pipes(struct inode *inode)
 {
-       struct rpc_inode *rpci = RPC_I(inode);
-       const struct rpc_pipe_ops *ops;
+       struct rpc_pipe *pipe = RPC_I(inode)->pipe;
        int need_release;
+       LIST_HEAD(free_list);
 
        mutex_lock(&inode->i_mutex);
-       ops = rpci->ops;
-       if (ops != NULL) {
-               LIST_HEAD(free_list);
-               spin_lock(&inode->i_lock);
-               need_release = rpci->nreaders != 0 || rpci->nwriters != 0;
-               rpci->nreaders = 0;
-               list_splice_init(&rpci->in_upcall, &free_list);
-               list_splice_init(&rpci->pipe, &free_list);
-               rpci->pipelen = 0;
-               rpci->ops = NULL;
-               spin_unlock(&inode->i_lock);
-               rpc_purge_list(rpci, &free_list, ops->destroy_msg, -EPIPE);
-               rpci->nwriters = 0;
-               if (need_release && ops->release_pipe)
-                       ops->release_pipe(inode);
-               cancel_delayed_work_sync(&rpci->queue_timeout);
-       }
+       spin_lock(&pipe->lock);
+       need_release = pipe->nreaders != 0 || pipe->nwriters != 0;
+       pipe->nreaders = 0;
+       list_splice_init(&pipe->in_upcall, &free_list);
+       list_splice_init(&pipe->pipe, &free_list);
+       pipe->pipelen = 0;
+       pipe->dentry = NULL;
+       spin_unlock(&pipe->lock);
+       rpc_purge_list(&RPC_I(inode)->waitq, &free_list, pipe->ops->destroy_msg, -EPIPE);
+       pipe->nwriters = 0;
+       if (need_release && pipe->ops->release_pipe)
+               pipe->ops->release_pipe(inode);
+       cancel_delayed_work_sync(&pipe->queue_timeout);
        rpc_inode_setowner(inode, NULL);
+       RPC_I(inode)->pipe = NULL;
        mutex_unlock(&inode->i_mutex);
 }
 
@@ -162,31 +201,39 @@ rpc_alloc_inode(struct super_block *sb)
 }
 
 static void
-rpc_destroy_inode(struct inode *inode)
+rpc_i_callback(struct rcu_head *head)
 {
+       struct inode *inode = container_of(head, struct inode, i_rcu);
        kmem_cache_free(rpc_inode_cachep, RPC_I(inode));
 }
 
+static void
+rpc_destroy_inode(struct inode *inode)
+{
+       call_rcu(&inode->i_rcu, rpc_i_callback);
+}
+
 static int
 rpc_pipe_open(struct inode *inode, struct file *filp)
 {
-       struct rpc_inode *rpci = RPC_I(inode);
+       struct rpc_pipe *pipe;
        int first_open;
        int res = -ENXIO;
 
        mutex_lock(&inode->i_mutex);
-       if (rpci->ops == NULL)
+       pipe = RPC_I(inode)->pipe;
+       if (pipe == NULL)
                goto out;
-       first_open = rpci->nreaders == 0 && rpci->nwriters == 0;
-       if (first_open && rpci->ops->open_pipe) {
-               res = rpci->ops->open_pipe(inode);
+       first_open = pipe->nreaders == 0 && pipe->nwriters == 0;
+       if (first_open && pipe->ops->open_pipe) {
+               res = pipe->ops->open_pipe(inode);
                if (res)
                        goto out;
        }
        if (filp->f_mode & FMODE_READ)
-               rpci->nreaders++;
+               pipe->nreaders++;
        if (filp->f_mode & FMODE_WRITE)
-               rpci->nwriters++;
+               pipe->nwriters++;
        res = 0;
 out:
        mutex_unlock(&inode->i_mutex);
@@ -196,38 +243,39 @@ out:
 static int
 rpc_pipe_release(struct inode *inode, struct file *filp)
 {
-       struct rpc_inode *rpci = RPC_I(inode);
+       struct rpc_pipe *pipe;
        struct rpc_pipe_msg *msg;
        int last_close;
 
        mutex_lock(&inode->i_mutex);
-       if (rpci->ops == NULL)
+       pipe = RPC_I(inode)->pipe;
+       if (pipe == NULL)
                goto out;
        msg = filp->private_data;
        if (msg != NULL) {
-               spin_lock(&inode->i_lock);
+               spin_lock(&pipe->lock);
                msg->errno = -EAGAIN;
                list_del_init(&msg->list);
-               spin_unlock(&inode->i_lock);
-               rpci->ops->destroy_msg(msg);
+               spin_unlock(&pipe->lock);
+               pipe->ops->destroy_msg(msg);
        }
        if (filp->f_mode & FMODE_WRITE)
-               rpci->nwriters --;
+               pipe->nwriters --;
        if (filp->f_mode & FMODE_READ) {
-               rpci->nreaders --;
-               if (rpci->nreaders == 0) {
+               pipe->nreaders --;
+               if (pipe->nreaders == 0) {
                        LIST_HEAD(free_list);
-                       spin_lock(&inode->i_lock);
-                       list_splice_init(&rpci->pipe, &free_list);
-                       rpci->pipelen = 0;
-                       spin_unlock(&inode->i_lock);
-                       rpc_purge_list(rpci, &free_list,
-                                       rpci->ops->destroy_msg, -EAGAIN);
+                       spin_lock(&pipe->lock);
+                       list_splice_init(&pipe->pipe, &free_list);
+                       pipe->pipelen = 0;
+                       spin_unlock(&pipe->lock);
+                       rpc_purge_list(&RPC_I(inode)->waitq, &free_list,
+                                       pipe->ops->destroy_msg, -EAGAIN);
                }
        }
-       last_close = rpci->nwriters == 0 && rpci->nreaders == 0;
-       if (last_close && rpci->ops->release_pipe)
-               rpci->ops->release_pipe(inode);
+       last_close = pipe->nwriters == 0 && pipe->nreaders == 0;
+       if (last_close && pipe->ops->release_pipe)
+               pipe->ops->release_pipe(inode);
 out:
        mutex_unlock(&inode->i_mutex);
        return 0;
@@ -237,39 +285,40 @@ static ssize_t
 rpc_pipe_read(struct file *filp, char __user *buf, size_t len, loff_t *offset)
 {
        struct inode *inode = filp->f_path.dentry->d_inode;
-       struct rpc_inode *rpci = RPC_I(inode);
+       struct rpc_pipe *pipe;
        struct rpc_pipe_msg *msg;
        int res = 0;
 
        mutex_lock(&inode->i_mutex);
-       if (rpci->ops == NULL) {
+       pipe = RPC_I(inode)->pipe;
+       if (pipe == NULL) {
                res = -EPIPE;
                goto out_unlock;
        }
        msg = filp->private_data;
        if (msg == NULL) {
-               spin_lock(&inode->i_lock);
-               if (!list_empty(&rpci->pipe)) {
-                       msg = list_entry(rpci->pipe.next,
+               spin_lock(&pipe->lock);
+               if (!list_empty(&pipe->pipe)) {
+                       msg = list_entry(pipe->pipe.next,
                                        struct rpc_pipe_msg,
                                        list);
-                       list_move(&msg->list, &rpci->in_upcall);
-                       rpci->pipelen -= msg->len;
+                       list_move(&msg->list, &pipe->in_upcall);
+                       pipe->pipelen -= msg->len;
                        filp->private_data = msg;
                        msg->copied = 0;
                }
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&pipe->lock);
                if (msg == NULL)
                        goto out_unlock;
        }
        /* NOTE: it is up to the callback to update msg->copied */
-       res = rpci->ops->upcall(filp, msg, buf, len);
+       res = pipe->ops->upcall(filp, msg, buf, len);
        if (res < 0 || msg->len == msg->copied) {
                filp->private_data = NULL;
-               spin_lock(&inode->i_lock);
+               spin_lock(&pipe->lock);
                list_del_init(&msg->list);
-               spin_unlock(&inode->i_lock);
-               rpci->ops->destroy_msg(msg);
+               spin_unlock(&pipe->lock);
+               pipe->ops->destroy_msg(msg);
        }
 out_unlock:
        mutex_unlock(&inode->i_mutex);
@@ -280,13 +329,12 @@ static ssize_t
 rpc_pipe_write(struct file *filp, const char __user *buf, size_t len, loff_t *offset)
 {
        struct inode *inode = filp->f_path.dentry->d_inode;
-       struct rpc_inode *rpci = RPC_I(inode);
        int res;
 
        mutex_lock(&inode->i_mutex);
        res = -EPIPE;
-       if (rpci->ops != NULL)
-               res = rpci->ops->downcall(filp, buf, len);
+       if (RPC_I(inode)->pipe != NULL)
+               res = RPC_I(inode)->pipe->ops->downcall(filp, buf, len);
        mutex_unlock(&inode->i_mutex);
        return res;
 }
@@ -294,17 +342,18 @@ rpc_pipe_write(struct file *filp, const char __user *buf, size_t len, loff_t *of
 static unsigned int
 rpc_pipe_poll(struct file *filp, struct poll_table_struct *wait)
 {
-       struct rpc_inode *rpci;
-       unsigned int mask = 0;
+       struct inode *inode = filp->f_path.dentry->d_inode;
+       struct rpc_inode *rpci = RPC_I(inode);
+       unsigned int mask = POLLOUT | POLLWRNORM;
 
-       rpci = RPC_I(filp->f_path.dentry->d_inode);
        poll_wait(filp, &rpci->waitq, wait);
 
-       mask = POLLOUT | POLLWRNORM;
-       if (rpci->ops == NULL)
+       mutex_lock(&inode->i_mutex);
+       if (rpci->pipe == NULL)
                mask |= POLLERR | POLLHUP;
-       if (filp->private_data || !list_empty(&rpci->pipe))
+       else if (filp->private_data || !list_empty(&rpci->pipe->pipe))
                mask |= POLLIN | POLLRDNORM;
+       mutex_unlock(&inode->i_mutex);
        return mask;
 }
 
@@ -312,23 +361,26 @@ static long
 rpc_pipe_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 {
        struct inode *inode = filp->f_path.dentry->d_inode;
-       struct rpc_inode *rpci = RPC_I(inode);
+       struct rpc_pipe *pipe;
        int len;
 
        switch (cmd) {
        case FIONREAD:
-               spin_lock(&inode->i_lock);
-               if (rpci->ops == NULL) {
-                       spin_unlock(&inode->i_lock);
+               mutex_lock(&inode->i_mutex);
+               pipe = RPC_I(inode)->pipe;
+               if (pipe == NULL) {
+                       mutex_unlock(&inode->i_mutex);
                        return -EPIPE;
                }
-               len = rpci->pipelen;
+               spin_lock(&pipe->lock);
+               len = pipe->pipelen;
                if (filp->private_data) {
                        struct rpc_pipe_msg *msg;
                        msg = filp->private_data;
                        len += msg->len - msg->copied;
                }
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&pipe->lock);
+               mutex_unlock(&inode->i_mutex);
                return put_user(len, (int __user *)arg);
        default:
                return -EINVAL;
@@ -351,12 +403,15 @@ rpc_show_info(struct seq_file *m, void *v)
 {
        struct rpc_clnt *clnt = m->private;
 
-       seq_printf(m, "RPC server: %s\n", clnt->cl_server);
+       rcu_read_lock();
+       seq_printf(m, "RPC server: %s\n",
+                       rcu_dereference(clnt->cl_xprt)->servername);
        seq_printf(m, "service: %s (%d) version %d\n", clnt->cl_protname,
                        clnt->cl_prog, clnt->cl_vers);
        seq_printf(m, "address: %s\n", rpc_peeraddr2str(clnt, RPC_DISPLAY_ADDR));
        seq_printf(m, "protocol: %s\n", rpc_peeraddr2str(clnt, RPC_DISPLAY_PROTO));
        seq_printf(m, "port: %s\n", rpc_peeraddr2str(clnt, RPC_DISPLAY_PORT));
+       rcu_read_unlock();
        return 0;
 }
 
@@ -413,24 +468,7 @@ struct rpc_filelist {
        umode_t mode;
 };
 
-struct vfsmount *rpc_get_mount(void)
-{
-       int err;
-
-       err = simple_pin_fs(&rpc_pipe_fs_type, &rpc_mount, &rpc_mount_count);
-       if (err != 0)
-               return ERR_PTR(err);
-       return rpc_mount;
-}
-EXPORT_SYMBOL_GPL(rpc_get_mount);
-
-void rpc_put_mount(void)
-{
-       simple_release_fs(&rpc_mount, &rpc_mount_count);
-}
-EXPORT_SYMBOL_GPL(rpc_put_mount);
-
-static int rpc_delete_dentry(struct dentry *dentry)
+static int rpc_delete_dentry(const struct dentry *dentry)
 {
        return 1;
 }
@@ -448,13 +486,13 @@ rpc_get_inode(struct super_block *sb, umode_t mode)
        inode->i_ino = get_next_ino();
        inode->i_mode = mode;
        inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
-       switch(mode & S_IFMT) {
-               case S_IFDIR:
-                       inode->i_fop = &simple_dir_operations;
-                       inode->i_op = &simple_dir_inode_operations;
-                       inc_nlink(inode);
-               default:
-                       break;
+       switch (mode & S_IFMT) {
+       case S_IFDIR:
+               inode->i_fop = &simple_dir_operations;
+               inode->i_op = &simple_dir_inode_operations;
+               inc_nlink(inode);
+       default:
+               break;
        }
        return inode;
 }
@@ -466,7 +504,7 @@ static int __rpc_create_common(struct inode *dir, struct dentry *dentry,
 {
        struct inode *inode;
 
-       BUG_ON(!d_unhashed(dentry));
+       d_drop(dentry);
        inode = rpc_get_inode(dir->i_sb, mode);
        if (!inode)
                goto out_err;
@@ -513,12 +551,47 @@ static int __rpc_mkdir(struct inode *dir, struct dentry *dentry,
        return 0;
 }
 
-static int __rpc_mkpipe(struct inode *dir, struct dentry *dentry,
-                       umode_t mode,
-                       const struct file_operations *i_fop,
-                       void *private,
-                       const struct rpc_pipe_ops *ops,
-                       int flags)
+static void
+init_pipe(struct rpc_pipe *pipe)
+{
+       pipe->nreaders = 0;
+       pipe->nwriters = 0;
+       INIT_LIST_HEAD(&pipe->in_upcall);
+       INIT_LIST_HEAD(&pipe->in_downcall);
+       INIT_LIST_HEAD(&pipe->pipe);
+       pipe->pipelen = 0;
+       INIT_DELAYED_WORK(&pipe->queue_timeout,
+                           rpc_timeout_upcall_queue);
+       pipe->ops = NULL;
+       spin_lock_init(&pipe->lock);
+       pipe->dentry = NULL;
+}
+
+void rpc_destroy_pipe_data(struct rpc_pipe *pipe)
+{
+       kfree(pipe);
+}
+EXPORT_SYMBOL_GPL(rpc_destroy_pipe_data);
+
+struct rpc_pipe *rpc_mkpipe_data(const struct rpc_pipe_ops *ops, int flags)
+{
+       struct rpc_pipe *pipe;
+
+       pipe = kzalloc(sizeof(struct rpc_pipe), GFP_KERNEL);
+       if (!pipe)
+               return ERR_PTR(-ENOMEM);
+       init_pipe(pipe);
+       pipe->ops = ops;
+       pipe->flags = flags;
+       return pipe;
+}
+EXPORT_SYMBOL_GPL(rpc_mkpipe_data);
+
+static int __rpc_mkpipe_dentry(struct inode *dir, struct dentry *dentry,
+                              umode_t mode,
+                              const struct file_operations *i_fop,
+                              void *private,
+                              struct rpc_pipe *pipe)
 {
        struct rpc_inode *rpci;
        int err;
@@ -527,10 +600,8 @@ static int __rpc_mkpipe(struct inode *dir, struct dentry *dentry,
        if (err)
                return err;
        rpci = RPC_I(dentry->d_inode);
-       rpci->nkern_readwriters = 1;
        rpci->private = private;
-       rpci->flags = flags;
-       rpci->ops = ops;
+       rpci->pipe = pipe;
        fsnotify_create(dir, dentry);
        return 0;
 }
@@ -546,6 +617,22 @@ 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;
@@ -560,16 +647,12 @@ static int __rpc_unlink(struct inode *dir, struct dentry *dentry)
 static int __rpc_rmpipe(struct inode *dir, struct dentry *dentry)
 {
        struct inode *inode = dentry->d_inode;
-       struct rpc_inode *rpci = RPC_I(inode);
 
-       rpci->nkern_readwriters--;
-       if (rpci->nkern_readwriters != 0)
-               return 0;
        rpc_close_pipes(inode);
        return __rpc_unlink(dir, dentry);
 }
 
-static struct dentry *__rpc_lookup_create(struct dentry *parent,
+static struct dentry *__rpc_lookup_create_exclusive(struct dentry *parent,
                                          struct qstr *name)
 {
        struct dentry *dentry;
@@ -577,27 +660,13 @@ static struct dentry *__rpc_lookup_create(struct dentry *parent,
        dentry = d_lookup(parent, name);
        if (!dentry) {
                dentry = d_alloc(parent, name);
-               if (!dentry) {
-                       dentry = ERR_PTR(-ENOMEM);
-                       goto out_err;
-               }
+               if (!dentry)
+                       return ERR_PTR(-ENOMEM);
        }
-       if (!dentry->d_inode)
-               dentry->d_op = &rpc_dentry_operations;
-out_err:
-       return dentry;
-}
-
-static struct dentry *__rpc_lookup_create_exclusive(struct dentry *parent,
-                                         struct qstr *name)
-{
-       struct dentry *dentry;
-
-       dentry = __rpc_lookup_create(parent, name);
-       if (IS_ERR(dentry))
-               return dentry;
-       if (dentry->d_inode == NULL)
+       if (dentry->d_inode == NULL) {
+               d_set_d_op(dentry, &rpc_dentry_operations);
                return dentry;
+       }
        dput(dentry);
        return ERR_PTR(-EEXIST);
 }
@@ -752,7 +821,7 @@ static int rpc_rmdir_depopulate(struct dentry *dentry,
  * @private: private data to associate with the pipe, for the caller's use
  * @ops: operations defining the behavior of the pipe: upcall, downcall,
  *     release_pipe, open_pipe, and destroy_msg.
- * @flags: rpc_inode flags
+ * @flags: rpc_pipe flags
  *
  * Data is made available for userspace to read by calls to
  * rpc_queue_upcall().  The actual reads will result in calls to
@@ -765,9 +834,8 @@ static int rpc_rmdir_depopulate(struct dentry *dentry,
  * The @private argument passed here will be available to all these methods
  * from the file pointer, via RPC_I(file->f_dentry->d_inode)->private.
  */
-struct dentry *rpc_mkpipe(struct dentry *parent, const char *name,
-                         void *private, const struct rpc_pipe_ops *ops,
-                         int flags)
+struct dentry *rpc_mkpipe_dentry(struct dentry *parent, const char *name,
+                                void *private, struct rpc_pipe *pipe)
 {
        struct dentry *dentry;
        struct inode *dir = parent->d_inode;
@@ -775,9 +843,9 @@ struct dentry *rpc_mkpipe(struct dentry *parent, const char *name,
        struct qstr q;
        int err;
 
-       if (ops->upcall == NULL)
+       if (pipe->ops->upcall == NULL)
                umode &= ~S_IRUGO;
-       if (ops->downcall == NULL)
+       if (pipe->ops->downcall == NULL)
                umode &= ~S_IWUGO;
 
        q.name = name;
@@ -785,24 +853,11 @@ struct dentry *rpc_mkpipe(struct dentry *parent, const char *name,
        q.hash = full_name_hash(q.name, q.len),
 
        mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
-       dentry = __rpc_lookup_create(parent, &q);
+       dentry = __rpc_lookup_create_exclusive(parent, &q);
        if (IS_ERR(dentry))
                goto out;
-       if (dentry->d_inode) {
-               struct rpc_inode *rpci = RPC_I(dentry->d_inode);
-               if (rpci->private != private ||
-                               rpci->ops != ops ||
-                               rpci->flags != flags) {
-                       dput (dentry);
-                       err = -EBUSY;
-                       goto out_err;
-               }
-               rpci->nkern_readwriters++;
-               goto out;
-       }
-
-       err = __rpc_mkpipe(dir, dentry, umode, &rpc_pipe_fops,
-                          private, ops, flags);
+       err = __rpc_mkpipe_dentry(dir, dentry, umode, &rpc_pipe_fops,
+                                 private, pipe);
        if (err)
                goto out_err;
 out:
@@ -815,7 +870,7 @@ out_err:
                        err);
        goto out;
 }
-EXPORT_SYMBOL_GPL(rpc_mkpipe);
+EXPORT_SYMBOL_GPL(rpc_mkpipe_dentry);
 
 /**
  * rpc_unlink - remove a pipe
@@ -888,7 +943,7 @@ struct dentry *rpc_create_client_dir(struct dentry *dentry,
 
 /**
  * rpc_remove_client_dir - Remove a directory created with rpc_create_client_dir()
- * @dentry: directory to remove
+ * @clnt: rpc client
  */
 int rpc_remove_client_dir(struct dentry *dentry)
 {
@@ -926,7 +981,7 @@ static void rpc_cachedir_depopulate(struct dentry *dentry)
 }
 
 struct dentry *rpc_create_cache_dir(struct dentry *parent, struct qstr *name,
-                                   mode_t umode, struct cache_detail *cd)
+                                   umode_t umode, struct cache_detail *cd)
 {
        return rpc_mkdir_populate(parent, name, umode, NULL,
                        rpc_cachedir_populate, cd);
@@ -959,6 +1014,7 @@ enum {
        RPCAUTH_statd,
        RPCAUTH_nfsd4_cb,
        RPCAUTH_cache,
+       RPCAUTH_nfsd,
        RPCAUTH_RootEOF
 };
 
@@ -991,13 +1047,67 @@ static const struct rpc_filelist files[] = {
                .name = "cache",
                .mode = S_IFDIR | S_IRUGO | S_IXUGO,
        },
+       [RPCAUTH_nfsd] = {
+               .name = "nfsd",
+               .mode = S_IFDIR | S_IRUGO | S_IXUGO,
+       },
 };
 
+/*
+ * This call can be used only in RPC pipefs mount notification hooks.
+ */
+struct dentry *rpc_d_lookup_sb(const struct super_block *sb,
+                              const unsigned char *dir_name)
+{
+       struct qstr dir = QSTR_INIT(dir_name, strlen(dir_name));
+
+       dir.hash = full_name_hash(dir.name, dir.len);
+       return d_lookup(sb->s_root, &dir);
+}
+EXPORT_SYMBOL_GPL(rpc_d_lookup_sb);
+
+void rpc_pipefs_init_net(struct net *net)
+{
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+
+       mutex_init(&sn->pipefs_sb_lock);
+}
+
+/*
+ * This call will be used for per network namespace operations calls.
+ * Note: Function will be returned with pipefs_sb_lock taken if superblock was
+ * found. This lock have to be released by rpc_put_sb_net() when all operations
+ * will be completed.
+ */
+struct super_block *rpc_get_sb_net(const struct net *net)
+{
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+
+       mutex_lock(&sn->pipefs_sb_lock);
+       if (sn->pipefs_sb)
+               return sn->pipefs_sb;
+       mutex_unlock(&sn->pipefs_sb_lock);
+       return NULL;
+}
+EXPORT_SYMBOL_GPL(rpc_get_sb_net);
+
+void rpc_put_sb_net(const struct net *net)
+{
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+
+       BUG_ON(sn->pipefs_sb == NULL);
+       mutex_unlock(&sn->pipefs_sb_lock);
+}
+EXPORT_SYMBOL_GPL(rpc_put_sb_net);
+
 static int
 rpc_fill_super(struct super_block *sb, void *data, int silent)
 {
        struct inode *inode;
        struct dentry *root;
+       struct net *net = data;
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+       int err;
 
        sb->s_blocksize = PAGE_CACHE_SIZE;
        sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
@@ -1006,30 +1116,60 @@ rpc_fill_super(struct super_block *sb, void *data, int silent)
        sb->s_time_gran = 1;
 
        inode = rpc_get_inode(sb, S_IFDIR | 0755);
-       if (!inode)
+       sb->s_root = root = d_make_root(inode);
+       if (!root)
                return -ENOMEM;
-       sb->s_root = root = d_alloc_root(inode);
-       if (!root) {
-               iput(inode);
-               return -ENOMEM;
-       }
        if (rpc_populate(root, files, RPCAUTH_lockd, RPCAUTH_RootEOF, NULL))
                return -ENOMEM;
+       dprintk("RPC:   sending pipefs MOUNT notification for net %p%s\n", net,
+                                                               NET_NAME(net));
+       sn->pipefs_sb = sb;
+       err = blocking_notifier_call_chain(&rpc_pipefs_notifier_list,
+                                          RPC_PIPEFS_MOUNT,
+                                          sb);
+       if (err)
+               goto err_depopulate;
+       sb->s_fs_info = get_net(net);
        return 0;
+
+err_depopulate:
+       blocking_notifier_call_chain(&rpc_pipefs_notifier_list,
+                                          RPC_PIPEFS_UMOUNT,
+                                          sb);
+       sn->pipefs_sb = NULL;
+       __rpc_depopulate(root, files, RPCAUTH_lockd, RPCAUTH_RootEOF);
+       return err;
 }
 
-static int
-rpc_get_sb(struct file_system_type *fs_type,
-               int flags, const char *dev_name, void *data, struct vfsmount *mnt)
+static struct dentry *
+rpc_mount(struct file_system_type *fs_type,
+               int flags, const char *dev_name, void *data)
 {
-       return get_sb_single(fs_type, flags, data, rpc_fill_super, mnt);
+       return mount_ns(fs_type, flags, current->nsproxy->net_ns, rpc_fill_super);
+}
+
+static void rpc_kill_sb(struct super_block *sb)
+{
+       struct net *net = sb->s_fs_info;
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+
+       mutex_lock(&sn->pipefs_sb_lock);
+       sn->pipefs_sb = NULL;
+       mutex_unlock(&sn->pipefs_sb_lock);
+       put_net(net);
+       dprintk("RPC:   sending pipefs UMOUNT notification for net %p%s\n", net,
+                                                               NET_NAME(net));
+       blocking_notifier_call_chain(&rpc_pipefs_notifier_list,
+                                          RPC_PIPEFS_UMOUNT,
+                                          sb);
+       kill_litter_super(sb);
 }
 
 static struct file_system_type rpc_pipe_fs_type = {
        .owner          = THIS_MODULE,
        .name           = "rpc_pipefs",
-       .get_sb         = rpc_get_sb,
-       .kill_sb        = kill_litter_super,
+       .mount          = rpc_mount,
+       .kill_sb        = rpc_kill_sb,
 };
 
 static void
@@ -1039,16 +1179,8 @@ init_once(void *foo)
 
        inode_init_once(&rpci->vfs_inode);
        rpci->private = NULL;
-       rpci->nreaders = 0;
-       rpci->nwriters = 0;
-       INIT_LIST_HEAD(&rpci->in_upcall);
-       INIT_LIST_HEAD(&rpci->in_downcall);
-       INIT_LIST_HEAD(&rpci->pipe);
-       rpci->pipelen = 0;
+       rpci->pipe = NULL;
        init_waitqueue_head(&rpci->waitq);
-       INIT_DELAYED_WORK(&rpci->queue_timeout,
-                           rpc_timeout_upcall_queue);
-       rpci->ops = NULL;
 }
 
 int register_rpc_pipefs(void)
@@ -1062,17 +1194,27 @@ int register_rpc_pipefs(void)
                                init_once);
        if (!rpc_inode_cachep)
                return -ENOMEM;
+       err = rpc_clients_notifier_register();
+       if (err)
+               goto err_notifier;
        err = register_filesystem(&rpc_pipe_fs_type);
-       if (err) {
-               kmem_cache_destroy(rpc_inode_cachep);
-               return err;
-       }
-
+       if (err)
+               goto err_register;
        return 0;
+
+err_register:
+       rpc_clients_notifier_unregister();
+err_notifier:
+       kmem_cache_destroy(rpc_inode_cachep);
+       return err;
 }
 
 void unregister_rpc_pipefs(void)
 {
+       rpc_clients_notifier_unregister();
        kmem_cache_destroy(rpc_inode_cachep);
        unregister_filesystem(&rpc_pipe_fs_type);
 }
+
+/* Make 'mount -t rpc_pipefs ...' autoload this module. */
+MODULE_ALIAS("rpc_pipefs");