1 files changed, 175 insertions, 94 deletions

diff --git a/fs/notify/mark.c b/fs/notify/mark.c
index e56e8768d67..d90deaa08e7 100644
--- a/fs/notify/mark.c
+++ b/fs/notify/mark.c
@@ -20,28 +20,29 @@
  * fsnotify inode mark locking/lifetime/and refcnting
  *
  * REFCNT:
- * The mark->refcnt tells how many "things" in the kernel currently are
- * referencing this object.  The object typically will live inside the kernel
- * with a refcnt of 2, one for each list it is on (i_list, g_list).  Any task
- * which can find this object holding the appropriete locks, can take a reference
- * and the object itself is guarenteed to survive until the reference is dropped.
+ * The group->recnt and mark->refcnt tell how many "things" in the kernel
+ * currently are referencing the objects. Both kind of objects typically will
+ * live inside the kernel with a refcnt of 2, one for its creation and one for
+ * the reference a group and a mark hold to each other.
+ * If you are holding the appropriate locks, you can take a reference and the
+ * object itself is guaranteed to survive until the reference is dropped.
  *
  * LOCKING:
- * There are 3 spinlocks involved with fsnotify inode marks and they MUST
- * be taken in order as follows:
+ * There are 3 locks involved with fsnotify inode marks and they MUST be taken
+ * in order as follows:
  *
+ * group->mark_mutex
  * mark->lock
- * group->mark_lock
  * inode->i_lock
  *
- * mark->lock protects 2 things, mark->group and mark->inode.  You must hold
- * that lock to dereference either of these things (they could be NULL even with
- * the lock)
- *
- * group->mark_lock protects the marks_list anchored inside a given group
- * and each mark is hooked via the g_list.  It also sorta protects the
- * free_g_list, which when used is anchored by a private list on the stack of the
- * task which held the group->mark_lock.
+ * group->mark_mutex protects the marks_list anchored inside a given group and
+ * each mark is hooked via the g_list.  It also protects the groups private
+ * data (i.e group limits).
+
+ * mark->lock protects the marks attributes like its masks and flags.
+ * Furthermore it protects the access to a reference of the group that the mark
+ * is assigned to as well as the access to a reference of the inode/vfsmount
+ * that is being watched by the mark.
  *
  * inode->i_lock protects the i_fsnotify_marks list anchored inside a
  * given inode and each mark is hooked via the i_list. (and sorta the
@@ -64,18 +65,11 @@
  * inode.  We take i_lock and walk the i_fsnotify_marks safely.  For each
  * mark on the list we take a reference (so the mark can't disappear under us).
  * We remove that mark form the inode's list of marks and we add this mark to a
- * private list anchored on the stack using i_free_list;  At this point we no
- * longer fear anything finding the mark using the inode's list of marks.
- *
- * We can safely and locklessly run the private list on the stack of everything
- * we just unattached from the original inode.  For each mark on the private list
- * we grab the mark-> and can thus dereference mark->group and mark->inode.  If
- * we see the group and inode are not NULL we take those locks.  Now holding all
- * 3 locks we can completely remove the mark from other tasks finding it in the
- * future.  Remember, 10 things might already be referencing this mark, but they
- * better be holding a ref.  We drop our reference we took before we unhooked it
- * from the inode.  When the ref hits 0 we can free the mark.
- *
+ * private list anchored on the stack using i_free_list; we walk i_free_list
+ * and before we destroy the mark we make sure that we dont race with a
+ * concurrent destroy_group by getting a ref to the marks group and taking the
+ * groups mutex.
+
  * Very similarly for freeing by group, except we use free_g_list.
  *
  * This has the very interesting property of being able to run concurrently with
@@ -85,17 +79,23 @@
 #include <linux/fs.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
+#include <linux/kthread.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
 #include <linux/slab.h>
 #include <linux/spinlock.h>
-#include <linux/writeback.h> /* for inode_lock */
+#include <linux/srcu.h>
 
-#include <asm/atomic.h>
+#include <linux/atomic.h>
 
 #include <linux/fsnotify_backend.h>
 #include "fsnotify.h"
 
+struct srcu_struct fsnotify_mark_srcu;
+static DEFINE_SPINLOCK(destroy_lock);
+static LIST_HEAD(destroy_list);
+static DECLARE_WAIT_QUEUE_HEAD(destroy_waitq);
+
 void fsnotify_get_mark(struct fsnotify_mark *mark)
 {
 	atomic_inc(&mark->refcnt);
@@ -103,8 +103,11 @@ void fsnotify_get_mark(struct fsnotify_mark *mark)
 
 void fsnotify_put_mark(struct fsnotify_mark *mark)
 {
-	if (atomic_dec_and_test(&mark->refcnt))
+	if (atomic_dec_and_test(&mark->refcnt)) {
+		if (mark->group)
+			fsnotify_put_group(mark->group);
 		mark->free_mark(mark);
+	}
 }
 
 /*
@@ -112,43 +115,50 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
  * The caller had better be holding a reference to this mark so we don't actually
  * do the final put under the mark->lock
  */
-void fsnotify_destroy_mark(struct fsnotify_mark *mark)
+void fsnotify_destroy_mark_locked(struct fsnotify_mark *mark,
+				  struct fsnotify_group *group)
 {
-	struct fsnotify_group *group;
-	struct inode *inode;
-
-	spin_lock(&mark->lock);
+	struct inode *inode = NULL;
 
-	group = mark->group;
-	inode = mark->i.inode;
+	BUG_ON(!mutex_is_locked(&group->mark_mutex));
 
-	BUG_ON(group && !inode);
-	BUG_ON(!group && inode);
+	spin_lock(&mark->lock);
 
-	/* if !group something else already marked this to die */
-	if (!group) {
+	/* something else already called this function on this mark */
+	if (!(mark->flags & FSNOTIFY_MARK_FLAG_ALIVE)) {
 		spin_unlock(&mark->lock);
 		return;
 	}
 
-	/* 1 from caller and 1 for being on i_list/g_list */
-	BUG_ON(atomic_read(&mark->refcnt) < 2);
+	mark->flags &= ~FSNOTIFY_MARK_FLAG_ALIVE;
 
-	spin_lock(&group->mark_lock);
-
-	if (mark->flags & FSNOTIFY_MARK_FLAG_INODE)
+	if (mark->flags & FSNOTIFY_MARK_FLAG_INODE) {
+		inode = mark->i.inode;
 		fsnotify_destroy_inode_mark(mark);
+	} else if (mark->flags & FSNOTIFY_MARK_FLAG_VFSMOUNT)
+		fsnotify_destroy_vfsmount_mark(mark);
 	else
 		BUG();
 
 	list_del_init(&mark->g_list);
-	mark->group = NULL;
-
-	fsnotify_put_mark(mark); /* for i_list and g_list */
 
-	spin_unlock(&group->mark_lock);
 	spin_unlock(&mark->lock);
 
+	if (inode && (mark->flags & FSNOTIFY_MARK_FLAG_OBJECT_PINNED))
+		iput(inode);
+	/* release lock temporarily */
+	mutex_unlock(&group->mark_mutex);
+
+	spin_lock(&destroy_lock);
+	list_add(&mark->destroy_list, &destroy_list);
+	spin_unlock(&destroy_lock);
+	wake_up(&destroy_waitq);
+	/*
+	 * We don't necessarily have a ref on mark from caller so the above destroy
+	 * may have actually freed it, unless this group provides a 'freeing_mark'
+	 * function which must be holding a reference.
+	 */
+
 	/*
 	 * Some groups like to know that marks are being freed.  This is a
 	 * callback to the group function to let it know that this mark
@@ -169,16 +179,34 @@ void fsnotify_destroy_mark(struct fsnotify_mark *mark)
 	 * is just a lazy update (and could be a perf win...)
 	 */
 
+	atomic_dec(&group->num_marks);
 
-	iput(inode);
+	mutex_lock_nested(&group->mark_mutex, SINGLE_DEPTH_NESTING);
+}
 
-	/*
-	 * it's possible that this group tried to destroy itself, but this
-	 * this mark was simultaneously being freed by inode.  If that's the
-	 * case, we finish freeing the group here.
-	 */
-	if (unlikely(atomic_dec_and_test(&group->num_marks)))
-		fsnotify_final_destroy_group(group);
+void fsnotify_destroy_mark(struct fsnotify_mark *mark,
+			   struct fsnotify_group *group)
+{
+	mutex_lock_nested(&group->mark_mutex, SINGLE_DEPTH_NESTING);
+	fsnotify_destroy_mark_locked(mark, group);
+	mutex_unlock(&group->mark_mutex);
+}
+
+void fsnotify_set_mark_mask_locked(struct fsnotify_mark *mark, __u32 mask)
+{
+	assert_spin_locked(&mark->lock);
+
+	mark->mask = mask;
+
+	if (mark->flags & FSNOTIFY_MARK_FLAG_INODE)
+		fsnotify_set_inode_mark_mask_locked(mark, mask);
+}
+
+void fsnotify_set_mark_ignored_mask_locked(struct fsnotify_mark *mark, __u32 mask)
+{
+	assert_spin_locked(&mark->lock);
+
+	mark->ignored_mask = mask;
 }
 
 /*
@@ -186,34 +214,26 @@ void fsnotify_destroy_mark(struct fsnotify_mark *mark)
  * These marks may be used for the fsnotify backend to determine which
  * event types should be delivered to which group.
  */
-int fsnotify_add_mark(struct fsnotify_mark *mark,
-		      struct fsnotify_group *group, struct inode *inode,
-		      struct vfsmount *mnt, int allow_dups)
+int fsnotify_add_mark_locked(struct fsnotify_mark *mark,
+			     struct fsnotify_group *group, struct inode *inode,
+			     struct vfsmount *mnt, int allow_dups)
 {
 	int ret = 0;
 
-	BUG_ON(mnt);
 	BUG_ON(inode && mnt);
 	BUG_ON(!inode && !mnt);
-
-	/*
-	 * if this group isn't being testing for inode type events we need
-	 * to start testing
-	 */
-	if (inode && unlikely(list_empty(&group->inode_group_list)))
-		fsnotify_add_inode_group(group);
-	else if (mnt && unlikely(list_empty(&group->vfsmount_group_list)))
-		fsnotify_add_vfsmount_group(group);
+	BUG_ON(!mutex_is_locked(&group->mark_mutex));
 
 	/*
 	 * LOCKING ORDER!!!!
+	 * group->mark_mutex
 	 * mark->lock
-	 * group->mark_lock
 	 * inode->i_lock
 	 */
 	spin_lock(&mark->lock);
-	spin_lock(&group->mark_lock);
+	mark->flags |= FSNOTIFY_MARK_FLAG_ALIVE;
 
+	fsnotify_get_group(group);
 	mark->group = group;
 	list_add(&mark->g_list, &group->marks_list);
 	atomic_inc(&group->num_marks);
@@ -223,11 +243,16 @@ int fsnotify_add_mark(struct fsnotify_mark *mark,
 		ret = fsnotify_add_inode_mark(mark, group, inode, allow_dups);
 		if (ret)
 			goto err;
+	} else if (mnt) {
+		ret = fsnotify_add_vfsmount_mark(mark, group, mnt, allow_dups);
+		if (ret)
+			goto err;
 	} else {
 		BUG();
 	}
 
-	spin_unlock(&group->mark_lock);
+	/* this will pin the object if appropriate */
+	fsnotify_set_mark_mask_locked(mark, mark->mask);
 	spin_unlock(&mark->lock);
 
 	if (inode)
@@ -235,37 +260,57 @@ int fsnotify_add_mark(struct fsnotify_mark *mark,
 
 	return ret;
 err:
-	mark->group = NULL;
+	mark->flags &= ~FSNOTIFY_MARK_FLAG_ALIVE;
 	list_del_init(&mark->g_list);
+	fsnotify_put_group(group);
+	mark->group = NULL;
 	atomic_dec(&group->num_marks);
-	fsnotify_put_mark(mark);
 
-	spin_unlock(&group->mark_lock);
 	spin_unlock(&mark->lock);
 
+	spin_lock(&destroy_lock);
+	list_add(&mark->destroy_list, &destroy_list);
+	spin_unlock(&destroy_lock);
+	wake_up(&destroy_waitq);
+
+	return ret;
+}
+
+int fsnotify_add_mark(struct fsnotify_mark *mark, struct fsnotify_group *group,
+		      struct inode *inode, struct vfsmount *mnt, int allow_dups)
+{
+	int ret;
+	mutex_lock(&group->mark_mutex);
+	ret = fsnotify_add_mark_locked(mark, group, inode, mnt, allow_dups);
+	mutex_unlock(&group->mark_mutex);
 	return ret;
 }
 
 /*
- * Given a group, destroy all of the marks associated with that group.
+ * clear any marks in a group in which mark->flags & flags is true
  */
-void fsnotify_clear_marks_by_group(struct fsnotify_group *group)
+void fsnotify_clear_marks_by_group_flags(struct fsnotify_group *group,
+					 unsigned int flags)
 {
 	struct fsnotify_mark *lmark, *mark;
-	LIST_HEAD(free_list);
 
-	spin_lock(&group->mark_lock);
+	mutex_lock_nested(&group->mark_mutex, SINGLE_DEPTH_NESTING);
 	list_for_each_entry_safe(mark, lmark, &group->marks_list, g_list) {
-		list_add(&mark->free_g_list, &free_list);
-		list_del_init(&mark->g_list);
-		fsnotify_get_mark(mark);
+		if (mark->flags & flags) {
+			fsnotify_get_mark(mark);
+			fsnotify_destroy_mark_locked(mark, group);
+			fsnotify_put_mark(mark);
+		}
 	}
-	spin_unlock(&group->mark_lock);
+	mutex_unlock(&group->mark_mutex);
+}
 
-	list_for_each_entry_safe(mark, lmark, &free_list, free_g_list) {
-		fsnotify_destroy_mark(mark);
-		fsnotify_put_mark(mark);
-	}
+/*
+ * Given a group, destroy all of the marks associated with that group.
+ */
+void fsnotify_clear_marks_by_group(struct fsnotify_group *group)
+{
+	fsnotify_clear_marks_by_group_flags(group, (unsigned int)-1);
 }
 
 void fsnotify_duplicate_mark(struct fsnotify_mark *new, struct fsnotify_mark *old)
@@ -273,6 +318,8 @@ void fsnotify_duplicate_mark(struct fsnotify_mark *new, struct fsnotify_mark *ol
 	assert_spin_locked(&old->lock);
 	new->i.inode = old->i.inode;
 	new->m.mnt = old->m.mnt;
+	if (old->group)
+		fsnotify_get_group(old->group);
 	new->group = old->group;
 	new->mask = old->mask;
 	new->free_mark = old->free_mark;
@@ -284,11 +331,45 @@ void fsnotify_duplicate_mark(struct fsnotify_mark *new, struct fsnotify_mark *ol
 void fsnotify_init_mark(struct fsnotify_mark *mark,
 			void (*free_mark)(struct fsnotify_mark *mark))
 {
+	memset(mark, 0, sizeof(*mark));
 	spin_lock_init(&mark->lock);
 	atomic_set(&mark->refcnt, 1);
-	INIT_HLIST_NODE(&mark->i.i_list);
-	mark->group = NULL;
-	mark->mask = 0;
-	mark->i.inode = NULL;
 	mark->free_mark = free_mark;
 }
+
+static int fsnotify_mark_destroy(void *ignored)
+{
+	struct fsnotify_mark *mark, *next;
+	struct list_head private_destroy_list;
+
+	for (;;) {
+		spin_lock(&destroy_lock);
+		/* exchange the list head */
+		list_replace_init(&destroy_list, &private_destroy_list);
+		spin_unlock(&destroy_lock);
+
+		synchronize_srcu(&fsnotify_mark_srcu);
+
+		list_for_each_entry_safe(mark, next, &private_destroy_list, destroy_list) {
+			list_del_init(&mark->destroy_list);
+			fsnotify_put_mark(mark);
+		}
+
+		wait_event_interruptible(destroy_waitq, !list_empty(&destroy_list));
+	}
+
+	return 0;
+}
+
+static int __init fsnotify_mark_init(void)
+{
+	struct task_struct *thread;
+
+	thread = kthread_run(fsnotify_mark_destroy, NULL,
+			     "fsnotify_mark");
+	if (IS_ERR(thread))
+		panic("unable to start fsnotify mark destruction thread.");
+
+	return 0;
+}
+device_initcall(fsnotify_mark_init);