delayed mntput

On final mntput() we want fs shutdown to happen before return to
userland; however, the only case where we want it happen right
there (i.e. where task_work_add won't do) is MNT_INTERNAL victim.
Those have to be fully synchronous - failure halfway through module
init might count on having vfsmount killed right there.  Fortunately,
final mntput on MNT_INTERNAL vfsmounts happens on shallow stack.
So we handle those synchronously and do an analog of delayed fput
logics for everything else.

As the result, we are guaranteed that fs shutdown will always happen
on shallow stack.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

diff --git a/fs/mount.h b/fs/mount.h
index 6740a621552950e7a795b249f635b5eeea7065d1..8f2a14ae38a20ea8132316a9c7573ecb3ed156b1 100644 (file)
--- a/fs/mount.h
+++ b/fs/mount.h
@@ -29,7 +29,10 @@ struct mount {
        struct mount *mnt_parent;
        struct dentry *mnt_mountpoint;
        struct vfsmount mnt;
-       struct rcu_head mnt_rcu;
+       union {
+               struct rcu_head mnt_rcu;
+               struct llist_node mnt_llist;
+       };
 #ifdef CONFIG_SMP
        struct mnt_pcp __percpu *mnt_pcp;
 #else

diff --git a/fs/namespace.c b/fs/namespace.c
index ef42d9bee2121f8e6a68937a5ecaa7670721ded9..044134315f9339a1824ee03a75ac45a399afa0dd 100644 (file)
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -23,6 +23,7 @@
 #include <linux/proc_ns.h>
 #include <linux/magic.h>
 #include <linux/bootmem.h>
+#include <linux/task_work.h>
 #include "pnode.h"
 #include "internal.h"
 
@@ -957,6 +958,46 @@ static struct mount *clone_mnt(struct mount *old, struct dentry *root,
        return ERR_PTR(err);
 }
 
+static void cleanup_mnt(struct mount *mnt)
+{
+       /*
+        * This probably indicates that somebody messed
+        * up a mnt_want/drop_write() pair.  If this
+        * happens, the filesystem was probably unable
+        * to make r/w->r/o transitions.
+        */
+       /*
+        * The locking used to deal with mnt_count decrement provides barriers,
+        * so mnt_get_writers() below is safe.
+        */
+       WARN_ON(mnt_get_writers(mnt));
+       if (unlikely(mnt->mnt_pins.first))
+               mnt_pin_kill(mnt);
+       fsnotify_vfsmount_delete(&mnt->mnt);
+       dput(mnt->mnt.mnt_root);
+       deactivate_super(mnt->mnt.mnt_sb);
+       mnt_free_id(mnt);
+       call_rcu(&mnt->mnt_rcu, delayed_free_vfsmnt);
+}
+
+static void __cleanup_mnt(struct rcu_head *head)
+{
+       cleanup_mnt(container_of(head, struct mount, mnt_rcu));
+}
+
+static LLIST_HEAD(delayed_mntput_list);
+static void delayed_mntput(struct work_struct *unused)
+{
+       struct llist_node *node = llist_del_all(&delayed_mntput_list);
+       struct llist_node *next;
+
+       for (; node; node = next) {
+               next = llist_next(node);
+               cleanup_mnt(llist_entry(node, struct mount, mnt_llist));
+       }
+}
+static DECLARE_DELAYED_WORK(delayed_mntput_work, delayed_mntput);
+
 static void mntput_no_expire(struct mount *mnt)
 {
        rcu_read_lock();
@@ -982,24 +1023,18 @@ static void mntput_no_expire(struct mount *mnt)
        list_del(&mnt->mnt_instance);
        unlock_mount_hash();
 
-       /*
-        * This probably indicates that somebody messed
-        * up a mnt_want/drop_write() pair.  If this
-        * happens, the filesystem was probably unable
-        * to make r/w->r/o transitions.
-        */
-       /*
-        * The locking used to deal with mnt_count decrement provides barriers,
-        * so mnt_get_writers() below is safe.
-        */
-       WARN_ON(mnt_get_writers(mnt));
-       if (unlikely(mnt->mnt_pins.first))
-               mnt_pin_kill(mnt);
-       fsnotify_vfsmount_delete(&mnt->mnt);
-       dput(mnt->mnt.mnt_root);
-       deactivate_super(mnt->mnt.mnt_sb);
-       mnt_free_id(mnt);
-       call_rcu(&mnt->mnt_rcu, delayed_free_vfsmnt);
+       if (likely(!(mnt->mnt.mnt_flags & MNT_INTERNAL))) {
+               struct task_struct *task = current;
+               if (likely(!(task->flags & PF_KTHREAD))) {
+                       init_task_work(&mnt->mnt_rcu, __cleanup_mnt);
+                       if (!task_work_add(task, &mnt->mnt_rcu, true))
+                               return;
+               }
+               if (llist_add(&mnt->mnt_llist, &delayed_mntput_list))
+                       schedule_delayed_work(&delayed_mntput_work, 1);
+               return;
+       }
+       cleanup_mnt(mnt);
 }
 
 void mntput(struct vfsmount *mnt)