[PATCH] Fix file lookup without ref

There are places in the kernel where we look up files in fd tables and
access the file structure without holding refereces to the file.  So, we
need special care to avoid the race between looking up files in the fd
table and tearing down of the file in another CPU.  Otherwise, one might
see a NULL f_dentry or such torn down version of the file.  This patch
fixes those special places where such a race may happen.

Signed-off-by: Dipankar Sarma <dipankar@in.ibm.com>
Acked-by: "Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

3 files changed, 28 insertions, 10 deletions

diff --git a/drivers/char/tty_io.c b/drivers/char/tty_io.c
index 53d3d066554..edaee70b2e6 100644
--- a/drivers/char/tty_io.c
+++ b/drivers/char/tty_io.c
@@ -2706,7 +2706,11 @@ static void __do_SAK(void *arg)
 		}
 		task_lock(p);
 		if (p->files) {
-			rcu_read_lock();
+			/*
+			 * We don't take a ref to the file, so we must
+			 * hold ->file_lock instead.
+			 */
+			spin_lock(&p->files->file_lock);
 			fdt = files_fdtable(p->files);
 			for (i=0; i < fdt->max_fds; i++) {
 				filp = fcheck_files(p->files, i);
@@ -2721,7 +2725,7 @@ static void __do_SAK(void *arg)
 					break;
 				}
 			}
-			rcu_read_unlock();
+			spin_unlock(&p->files->file_lock);
 		}
 		task_unlock(p);
 	} while_each_task_pid(session, PIDTYPE_SID, p);
diff --git a/fs/locks.c b/fs/locks.c
index 909eab8fb1d..e75ac392a31 100644
--- a/fs/locks.c
+++ b/fs/locks.c
@@ -2212,7 +2212,12 @@ void steal_locks(fl_owner_t from)
 
 	lock_kernel();
 	j = 0;
-	rcu_read_lock();
+
+	/*
+	 * We are not taking a ref to the file structures, so
+	 * we need to acquire ->file_lock.
+	 */
+	spin_lock(&files->file_lock);
 	fdt = files_fdtable(files);
 	for (;;) {
 		unsigned long set;
@@ -2230,7 +2235,7 @@ void steal_locks(fl_owner_t from)
 			set >>= 1;
 		}
 	}
-	rcu_read_unlock();
+	spin_unlock(&files->file_lock);
 	unlock_kernel();
 }
 EXPORT_SYMBOL(steal_locks);
diff --git a/fs/proc/base.c b/fs/proc/base.c
index 20feb7568de..c192cb2af2c 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -294,16 +294,20 @@ static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsm
 
 	files = get_files_struct(task);
 	if (files) {
-		rcu_read_lock();
+		/*
+		 * We are not taking a ref to the file structure, so we must
+		 * hold ->file_lock.
+		 */
+		spin_lock(&files->file_lock);
 		file = fcheck_files(files, fd);
 		if (file) {
 			*mnt = mntget(file->f_vfsmnt);
 			*dentry = dget(file->f_dentry);
-			rcu_read_unlock();
+			spin_unlock(&files->file_lock);
 			put_files_struct(files);
 			return 0;
 		}
-		rcu_read_unlock();
+		spin_unlock(&files->file_lock);
 		put_files_struct(files);
 	}
 	return -ENOENT;
@@ -1485,7 +1489,12 @@ static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry,
 	if (!files)
 		goto out_unlock;
 	inode->i_mode = S_IFLNK;
-	rcu_read_lock();
+
+	/*
+	 * We are not taking a ref to the file structure, so we must
+	 * hold ->file_lock.
+	 */
+	spin_lock(&files->file_lock);
 	file = fcheck_files(files, fd);
 	if (!file)
 		goto out_unlock2;
@@ -1493,7 +1502,7 @@ static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry,
 		inode->i_mode |= S_IRUSR | S_IXUSR;
 	if (file->f_mode & 2)
 		inode->i_mode |= S_IWUSR | S_IXUSR;
-	rcu_read_unlock();
+	spin_unlock(&files->file_lock);
 	put_files_struct(files);
 	inode->i_op = &proc_pid_link_inode_operations;
 	inode->i_size = 64;
@@ -1503,7 +1512,7 @@ static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry,
 	return NULL;
 
 out_unlock2:
-	rcu_read_unlock();
+	spin_unlock(&files->file_lock);
 	put_files_struct(files);
 out_unlock:
 	iput(inode);