[PATCH] r/o bind mounts: elevate write count for open()s

This is the first really tricky patch in the series.  It elevates the writer
count on a mount each time a non-special file is opened for write.

We used to do this in may_open(), but Miklos pointed out that __dentry_open()
is used as well to create filps.  This will cover even those cases, while a
call in may_open() would not have.

There is also an elevated count around the vfs_create() call in open_namei().
See the comments for more details, but we need this to fix a 'create, remount,
fail r/w open()' race.

Some filesystems forego the use of normal vfs calls to create
struct files.   Make sure that these users elevate the mnt
writer count because they will get __fput(), and we need
to make sure they're balanced.

Acked-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
diff --git a/fs/namei.c b/fs/namei.c
index 83c843b..e179f71 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -1623,8 +1623,7 @@
 			return -EACCES;
 
 		flag &= ~O_TRUNC;
-	} else if (IS_RDONLY(inode) && (acc_mode & MAY_WRITE))
-		return -EROFS;
+	}
 
 	error = vfs_permission(nd, acc_mode);
 	if (error)
@@ -1724,18 +1723,32 @@
 	return flag;
 }
 
+static int open_will_write_to_fs(int flag, struct inode *inode)
+{
+	/*
+	 * We'll never write to the fs underlying
+	 * a device file.
+	 */
+	if (special_file(inode->i_mode))
+		return 0;
+	return (flag & O_TRUNC);
+}
+
 /*
- * Note that the low bits of "flag" aren't the same as in the open
- * system call.  See open_to_namei_flags().
+ * Note that the low bits of the passed in "open_flag"
+ * are not the same as in the local variable "flag". See
+ * open_to_namei_flags() for more details.
  */
 struct file *do_filp_open(int dfd, const char *pathname,
 		int open_flag, int mode)
 {
+	struct file *filp;
 	struct nameidata nd;
 	int acc_mode, error;
 	struct path path;
 	struct dentry *dir;
 	int count = 0;
+	int will_write;
 	int flag = open_to_namei_flags(open_flag);
 
 	acc_mode = ACC_MODE(flag);
@@ -1791,17 +1804,30 @@
 	}
 
 	if (IS_ERR(nd.intent.open.file)) {
-		mutex_unlock(&dir->d_inode->i_mutex);
 		error = PTR_ERR(nd.intent.open.file);
-		goto exit_dput;
+		goto exit_mutex_unlock;
 	}
 
 	/* Negative dentry, just create the file */
 	if (!path.dentry->d_inode) {
-		error = __open_namei_create(&nd, &path, flag, mode);
+		/*
+		 * This write is needed to ensure that a
+		 * ro->rw transition does not occur between
+		 * the time when the file is created and when
+		 * a permanent write count is taken through
+		 * the 'struct file' in nameidata_to_filp().
+		 */
+		error = mnt_want_write(nd.path.mnt);
 		if (error)
+			goto exit_mutex_unlock;
+		error = __open_namei_create(&nd, &path, flag, mode);
+		if (error) {
+			mnt_drop_write(nd.path.mnt);
 			goto exit;
-		return nameidata_to_filp(&nd, open_flag);
+		}
+		filp = nameidata_to_filp(&nd, open_flag);
+		mnt_drop_write(nd.path.mnt);
+		return filp;
 	}
 
 	/*
@@ -1831,11 +1857,40 @@
 	if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode))
 		goto exit;
 ok:
+	/*
+	 * Consider:
+	 * 1. may_open() truncates a file
+	 * 2. a rw->ro mount transition occurs
+	 * 3. nameidata_to_filp() fails due to
+	 *    the ro mount.
+	 * That would be inconsistent, and should
+	 * be avoided. Taking this mnt write here
+	 * ensures that (2) can not occur.
+	 */
+	will_write = open_will_write_to_fs(flag, nd.path.dentry->d_inode);
+	if (will_write) {
+		error = mnt_want_write(nd.path.mnt);
+		if (error)
+			goto exit;
+	}
 	error = may_open(&nd, acc_mode, flag);
-	if (error)
+	if (error) {
+		if (will_write)
+			mnt_drop_write(nd.path.mnt);
 		goto exit;
-	return nameidata_to_filp(&nd, open_flag);
+	}
+	filp = nameidata_to_filp(&nd, open_flag);
+	/*
+	 * It is now safe to drop the mnt write
+	 * because the filp has had a write taken
+	 * on its behalf.
+	 */
+	if (will_write)
+		mnt_drop_write(nd.path.mnt);
+	return filp;
 
+exit_mutex_unlock:
+	mutex_unlock(&dir->d_inode->i_mutex);
 exit_dput:
 	path_put_conditional(&path, &nd);
 exit: