fasync: re-organize fasync entry insertion to allow it under a spinlock

You currently cannot use "fasync_helper()" in an atomic environment to
insert a new fasync entry, because it will need to allocate the new
"struct fasync_struct".

Yet fcntl_setlease() wants to call this under lock_flocks(), which is in
the process of being converted from the BKL to a spinlock.

In order to fix this, this abstracts out the actual fasync list
insertion and the fasync allocations into functions of their own, and
teaches fs/locks.c to pre-allocate the fasync_struct entry.  That way
the actual list insertion can happen while holding the required
spinlock.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[bfields@redhat.com: rebase on top of my changes to Arnd's patch]
Tested-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>

2 files changed, 67 insertions, 17 deletions

diff --git a/fs/fcntl.c b/fs/fcntl.c
index f8cc34f542c..dcdbc6f5c33 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -640,7 +640,7 @@ static void fasync_free_rcu(struct rcu_head *head)
  * match the state "is the filp on a fasync list".
  *
  */
-static int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp)
+int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp)
 {
 	struct fasync_struct *fa, **fp;
 	int result = 0;
@@ -666,21 +666,28 @@ static int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp)
 	return result;
 }
 
+struct fasync_struct *fasync_alloc(void)
+{
+	return kmem_cache_alloc(fasync_cache, GFP_KERNEL);
+}
+
 /*
- * Add a fasync entry. Return negative on error, positive if
- * added, and zero if did nothing but change an existing one.
- *
- * NOTE! It is very important that the FASYNC flag always
- * match the state "is the filp on a fasync list".
+ * NOTE! This can be used only for unused fasync entries:
+ * entries that actually got inserted on the fasync list
+ * need to be released by rcu - see fasync_remove_entry.
  */
-static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fapp)
+void fasync_free(struct fasync_struct *new)
 {
-	struct fasync_struct *new, *fa, **fp;
-	int result = 0;
+	kmem_cache_free(fasync_cache, new);
+}
 
-	new = kmem_cache_alloc(fasync_cache, GFP_KERNEL);
-	if (!new)
-		return -ENOMEM;
+/*
+ * Insert a new entry into the fasync list.  Return the pointer to the
+ * old one if we didn't use the new one.
+ */
+struct fasync_struct *fasync_insert_entry(int fd, struct file *filp, struct fasync_struct **fapp, struct fasync_struct *new)
+{
+        struct fasync_struct *fa, **fp;
 
 	spin_lock(&filp->f_lock);
 	spin_lock(&fasync_lock);
@@ -691,8 +698,6 @@ static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fa
 		spin_lock_irq(&fa->fa_lock);
 		fa->fa_fd = fd;
 		spin_unlock_irq(&fa->fa_lock);
-
-		kmem_cache_free(fasync_cache, new);
 		goto out;
 	}
 
@@ -702,13 +707,42 @@ static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fa
 	new->fa_fd = fd;
 	new->fa_next = *fapp;
 	rcu_assign_pointer(*fapp, new);
-	result = 1;
 	filp->f_flags |= FASYNC;
 
 out:
 	spin_unlock(&fasync_lock);
 	spin_unlock(&filp->f_lock);
-	return result;
+	return fa;
+}
+
+/*
+ * Add a fasync entry. Return negative on error, positive if
+ * added, and zero if did nothing but change an existing one.
+ *
+ * NOTE! It is very important that the FASYNC flag always
+ * match the state "is the filp on a fasync list".
+ */
+static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fapp)
+{
+	struct fasync_struct *new;
+
+	new = fasync_alloc();
+	if (!new)
+		return -ENOMEM;
+
+	/*
+	 * fasync_insert_entry() returns the old (update) entry if
+	 * it existed.
+	 *
+	 * So free the (unused) new entry and return 0 to let the
+	 * caller know that we didn't add any new fasync entries.
+	 */
+	if (fasync_insert_entry(fd, filp, fapp, new)) {
+		fasync_free(new);
+		return 0;
+	}
+
+	return 1;
 }
 
 /*
diff --git a/fs/locks.c b/fs/locks.c
index 0391d2ff5a4..85fd9ce1aba 100644
--- a/fs/locks.c
+++ b/fs/locks.c
@@ -1505,6 +1505,7 @@ EXPORT_SYMBOL_GPL(vfs_setlease);
 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
 {
 	struct file_lock *fl;
+	struct fasync_struct *new;
 	struct inode *inode = filp->f_path.dentry->d_inode;
 	int error;
 
@@ -1512,12 +1513,25 @@ int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
 	if (IS_ERR(fl))
 		return PTR_ERR(fl);
 
+	new = fasync_alloc();
+	if (!new) {
+		locks_free_lock(fl);
+		return -ENOMEM;
+	}
 	lock_flocks();
 	error = __vfs_setlease(filp, arg, &fl);
 	if (error || arg == F_UNLCK)
 		goto out_unlock;
 
-	error = fasync_helper(fd, filp, 1, &fl->fl_fasync);
+	/*
+	 * fasync_insert_entry() returns the old entry if any.
+	 * If there was no old entry, then it used 'new' and
+	 * inserted it into the fasync list. Clear new so that
+	 * we don't release it here.
+	 */
+	if (!fasync_insert_entry(fd, filp, &fl->fl_fasync, new))
+		new = NULL;
+
 	if (error < 0) {
 		/* remove lease just inserted by setlease */
 		fl->fl_type = F_UNLCK | F_INPROGRESS;
@@ -1529,6 +1543,8 @@ int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
 	error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
 out_unlock:
 	unlock_flocks();
+	if (new)
+		fasync_free(new);
 	return error;
 }