Move FASYNC bit handling to f_op->fasync()

Removing the BKL from FASYNC handling ran into the challenge of keeping the
setting of the FASYNC bit in filp->f_flags atomic with regard to calls to
the underlying fasync() function.  Andi Kleen suggested moving the handling
of that bit into fasync(); this patch does exactly that.  As a result, we
have a couple of internal API changes: fasync() must now manage the FASYNC
bit, and it will be called without the BKL held.

As it happens, every fasync() implementation in the kernel with one
exception calls fasync_helper().  So, if we make fasync_helper() set the
FASYNC bit, we can avoid making any changes to the other fasync()
functions - as long as those functions, themselves, have proper locking.
Most fasync() implementations do nothing but call fasync_helper() - which
has its own lock - so they are easily verified as correct.  The BKL had
already been pushed down into the rest.

The networking code has its own version of fasync_helper(), so that code
has been augmented with explicit FASYNC bit handling.

Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: David Miller <davem@davemloft.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>

4 files changed, 29 insertions, 27 deletions

diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index ec6a939..4e78ce6 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -437,8 +437,11 @@ grab BKL for cases when we close a file that had been opened r/w, but that
 can and should be done using the internal locking with smaller critical areas).
 Current worst offender is ext2_get_block()...
 
-->fasync() is a mess. This area needs a big cleanup and that will probably
-affect locking.
+->fasync() is called without BKL protection, and is responsible for
+maintaining the FASYNC bit in filp->f_flags.  Most instances call
+fasync_helper(), which does that maintenance, so it's not normally
+something one needs to worry about.  Return values > 0 will be mapped to
+zero in the VFS layer.
 
 ->readdir() and ->ioctl() on directories must be changed. Ideally we would
 move ->readdir() to inode_operations and use a separate method for directory
diff --git a/fs/fcntl.c b/fs/fcntl.c
index 04df857..431bb64 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -141,7 +141,7 @@ SYSCALL_DEFINE1(dup, unsigned int, fildes)
 	return ret;
 }
 
-#define SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | FASYNC | O_DIRECT | O_NOATIME)
+#define SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT | O_NOATIME)
 
 static int setfl(int fd, struct file * filp, unsigned long arg)
 {
@@ -177,23 +177,19 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
 		return error;
 
 	/*
-	 * We still need a lock here for now to keep multiple FASYNC calls
-	 * from racing with each other.
+	 * ->fasync() is responsible for setting the FASYNC bit.
 	 */
-	lock_kernel();
-	if ((arg ^ filp->f_flags) & FASYNC) {
-		if (filp->f_op && filp->f_op->fasync) {
-			error = filp->f_op->fasync(fd, filp, (arg & FASYNC) != 0);
-			if (error < 0)
-				goto out;
-		}
+	if (((arg ^ filp->f_flags) & FASYNC) && filp->f_op &&
+			filp->f_op->fasync) {
+		error = filp->f_op->fasync(fd, filp, (arg & FASYNC) != 0);
+		if (error < 0)
+			goto out;
 	}
-
 	spin_lock(&filp->f_lock);
 	filp->f_flags = (arg & SETFL_MASK) | (filp->f_flags & ~SETFL_MASK);
 	spin_unlock(&filp->f_lock);
+
  out:
-	unlock_kernel();
 	return error;
 }
 
@@ -518,7 +514,7 @@ static DEFINE_RWLOCK(fasync_lock);
 static struct kmem_cache *fasync_cache __read_mostly;
 
 /*
- * fasync_helper() is used by some character device drivers (mainly mice)
+ * fasync_helper() is used by almost all character device drivers
  * to set up the fasync queue. It returns negative on error, 0 if it did
  * no changes and positive if it added/deleted the entry.
  */
@@ -557,6 +553,13 @@ int fasync_helper(int fd, struct file * filp, int on, struct fasync_struct **fap
 		result = 1;
 	}
 out:
+	/* Fix up FASYNC bit while still holding fasync_lock */
+	spin_lock(&filp->f_lock);
+	if (on)
+		filp->f_flags |= FASYNC;
+	else
+		filp->f_flags &= ~FASYNC;
+	spin_unlock(&filp->f_lock);
 	write_unlock_irq(&fasync_lock);
 	return result;
 }
diff --git a/fs/ioctl.c b/fs/ioctl.c
index 421aab4..e8e89ed 100644
--- a/fs/ioctl.c
+++ b/fs/ioctl.c
@@ -427,19 +427,11 @@ static int ioctl_fioasync(unsigned int fd, struct file *filp,
 	/* Did FASYNC state change ? */
 	if ((flag ^ filp->f_flags) & FASYNC) {
 		if (filp->f_op && filp->f_op->fasync)
+			/* fasync() adjusts filp->f_flags */
 			error = filp->f_op->fasync(fd, filp, on);
 		else
 			error = -ENOTTY;
 	}
-	if (error)
-		return error;
-
-	spin_lock(&filp->f_lock);
-	if (on)
-		filp->f_flags |= FASYNC;
-	else
-		filp->f_flags &= ~FASYNC;
-	spin_unlock(&filp->f_lock);
 	return error;
 }
 
@@ -507,10 +499,7 @@ int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
 		break;
 
 	case FIOASYNC:
-		/* BKL needed to avoid races tweaking f_flags */
-		lock_kernel();
 		error = ioctl_fioasync(fd, filp, argp);
-		unlock_kernel();
 		break;
 
 	case FIOQSIZE:
diff --git a/net/socket.c b/net/socket.c
index 35dd737..0f75746 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -1030,6 +1030,13 @@ static int sock_fasync(int fd, struct file *filp, int on)
 
 	lock_sock(sk);
 
+	spin_lock(&filp->f_lock);
+	if (on)
+		filp->f_flags |= FASYNC;
+	else
+		filp->f_flags &= ~FASYNC;
+	spin_unlock(&filp->f_lock);
+
 	prev = &(sock->fasync_list);
 
 	for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)